Factory to create client IPC classes. yarn.ipc.client.factory.class Factory to create server IPC classes. yarn.ipc.server.factory.class Factory to create serializeable records. yarn.ipc.record.factory.class RPC class implementation yarn.ipc.rpc.class org.apache.hadoop.yarn.ipc.HadoopYarnProtoRPC The hostname of the RM. yarn.resourcemanager.hostname 0.0.0.0 The address of the applications manager interface in the RM. yarn.resourcemanager.address ${yarn.resourcemanager.hostname}:8032 The actual address the server will bind to. If this optional address is set, the RPC and webapp servers will bind to this address and the port specified in yarn.resourcemanager.address and yarn.resourcemanager.webapp.address, respectively. This is most useful for making RM listen to all interfaces by setting to 0.0.0.0. yarn.resourcemanager.bind-host If set to true, then ALL container updates will be automatically sent to the NM in the next heartbeat yarn.resourcemanager.auto-update.containers false The number of threads used to handle applications manager requests. yarn.resourcemanager.client.thread-count 50 Number of threads used to launch/cleanup AM. yarn.resourcemanager.amlauncher.thread-count 50 Retry times to connect with NM. yarn.resourcemanager.nodemanager-connect-retries 10 Timeout in milliseconds when YARN dispatcher tries to drain the events. Typically, this happens when service is stopping. e.g. RM drains the ATS events dispatcher when stopping. yarn.dispatcher.drain-events.timeout 300000 The threshold used to trigger the logging of event types and counts in RM's main event dispatcher. Default length is 5000, which means RM will print events info when the queue size cumulatively reaches 5000 every time. Such info can be used to reveal what kind of events that RM is stuck at processing mostly, it can help to narrow down certain performance issues. yarn.dispatcher.print-events-info.threshold 5000 Resource manager dispatcher thread cpu monitor sampling rate. Units are samples per minute. This controls how often to sample the cpu utilization of the resource manager dispatcher thread. The cpu utilization is displayed on the RM UI as scheduler busy %. Set this to zero to disable the dispatcher thread monitor. Defaults to 60 samples per minute. yarn.dispatcher.cpu-monitor.samples-per-min 60 Resource manager dispatcher has a thread pool that prints EventQueue, configure the corePoolSize of this thread pool, the meaning of corePoolSize is the number of threads to keep in the pool. the default value is 1. yarn.dispatcher.print-thread-pool.core-pool-size 1 Resource manager dispatcher has a thread pool that prints EventQueue, configure the maximumPoolSize of this thread pool, the meaning of maximumPoolSize is the maximum number of threads to allow in the pool. the default value is 5. yarn.dispatcher.print-thread-pool.maximum-pool-size 5 Resource manager dispatcher has a thread pool that prints EventQueue, configure the keepAliveTime of this thread pool, The meaning of keepAliveTime is as follows when the number of threads is greater than the core, this is the maximum time that excess idle threads will wait for new tasks before terminating. the default value is 10s. yarn.dispatcher.print-thread-pool.keep-alive-time 10s The expiry interval for application master reporting. yarn.am.liveness-monitor.expiry-interval-ms 600000 The Kerberos principal for the resource manager. yarn.resourcemanager.principal The address of the scheduler interface. yarn.resourcemanager.scheduler.address ${yarn.resourcemanager.hostname}:8030 Number of threads to handle scheduler interface. yarn.resourcemanager.scheduler.client.thread-count 50 Specify which handler will be used to process PlacementConstraints. Acceptable values are: `placement-processor`, `scheduler` and `disabled`. For a detailed explanation of these values, please refer to documentation. yarn.resourcemanager.placement-constraints.handler disabled Number of times to retry placing of rejected SchedulingRequests yarn.resourcemanager.placement-constraints.retry-attempts 3 Constraint Placement Algorithm to be used. yarn.resourcemanager.placement-constraints.algorithm.class org.apache.hadoop.yarn.server.resourcemanager.scheduler.constraint.algorithm.DefaultPlacementAlgorithm Placement Algorithm Requests Iterator to be used. yarn.resourcemanager.placement-constraints.algorithm.iterator SERIAL Threadpool size for the Algorithm used for placement constraint processing. yarn.resourcemanager.placement-constraints.algorithm.pool-size 1 Threadpool size for the Scheduler invocation phase of placement constraint processing. yarn.resourcemanager.placement-constraints.scheduler.pool-size 1 Comma separated class names of ApplicationMasterServiceProcessor implementations. The processors will be applied in the order they are specified. yarn.resourcemanager.application-master-service.processors This configures the HTTP endpoint for YARN Daemons.The following values are supported: - HTTP_ONLY : Service is provided only on http - HTTPS_ONLY : Service is provided only on https yarn.http.policy HTTP_ONLY The http address of the RM web application. If only a host is provided as the value, the webapp will be served on a random port. yarn.resourcemanager.webapp.address ${yarn.resourcemanager.hostname}:8088 The https address of the RM web application. If only a host is provided as the value, the webapp will be served on a random port. yarn.resourcemanager.webapp.https.address ${yarn.resourcemanager.hostname}:8090 The Kerberos keytab file to be used for spnego filter for the RM web interface. yarn.resourcemanager.webapp.spnego-keytab-file The Kerberos principal to be used for spnego filter for the RM web interface. yarn.resourcemanager.webapp.spnego-principal Add button to kill application in the RM Application view. yarn.resourcemanager.webapp.ui-actions.enabled true To enable RM web ui2 application. yarn.webapp.ui2.enable false Enable tools section in all ui1 webapp. yarn.webapp.ui1.tools.enable true Explicitly provide WAR file path for ui2 if needed. yarn.webapp.ui2.war-file-path Enable services rest api on ResourceManager. yarn.webapp.api-service.enable false yarn.resourcemanager.resource-tracker.address ${yarn.resourcemanager.hostname}:8031 yarn.resourcemanager.resource-tracker.nm.ip-hostname-check false Are acls enabled. yarn.acl.enable false Are reservation acls enabled. yarn.acl.reservation-enable false ACL of who can be admin of the YARN cluster. yarn.admin.acl * The address of the RM admin interface. yarn.resourcemanager.admin.address ${yarn.resourcemanager.hostname}:8033 Number of threads used to handle RM admin interface. yarn.resourcemanager.admin.client.thread-count 1 Maximum time to wait to establish connection to ResourceManager. yarn.resourcemanager.connect.max-wait.ms 900000 How often to try connecting to the ResourceManager. yarn.resourcemanager.connect.retry-interval.ms 30000 The default maximum number of application attempts, if unset by the user. Each application master can specify its individual maximum number of application attempts via the API, but the individual number cannot be more than the global upper bound in yarn.resourcemanager.am.global.max-attempts. The default number is set to 2, to allow at least one retry for AM. yarn.resourcemanager.am.max-attempts 2 How often to check that containers are still alive. yarn.resourcemanager.container.liveness-monitor.interval-ms 600000 The keytab for the resource manager. yarn.resourcemanager.keytab /etc/krb5.keytab Flag to enable override of the default kerberos authentication filter with the RM authentication filter to allow authentication using delegation tokens(fallback to kerberos if the tokens are missing). Only applicable when the http authentication type is kerberos. yarn.resourcemanager.webapp.delegation-token-auth-filter.enabled true Flag to enable cross-origin (CORS) support in the RM. This flag requires the CORS filter initializer to be added to the filter initializers list in core-site.xml. yarn.resourcemanager.webapp.cross-origin.enabled false How long to wait until a node manager is considered dead. yarn.nm.liveness-monitor.expiry-interval-ms 600000 Path to file with nodes to include. yarn.resourcemanager.nodes.include-path Path to file with nodes to exclude. yarn.resourcemanager.nodes.exclude-path The expiry interval for node IP caching. -1 disables the caching yarn.resourcemanager.node-ip-cache.expiry-interval-secs -1 Number of threads to handle resource tracker calls. yarn.resourcemanager.resource-tracker.client.thread-count 50 The class to use as the resource scheduler. yarn.resourcemanager.scheduler.class org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.CapacityScheduler The minimum allocation for every container request at the RM in MBs. Memory requests lower than this will be set to the value of this property. Additionally, a node manager that is configured to have less memory than this value will be shut down by the resource manager. yarn.scheduler.minimum-allocation-mb 1024 The maximum allocation for every container request at the RM in MBs. Memory requests higher than this will throw an InvalidResourceRequestException. yarn.scheduler.maximum-allocation-mb 8192 The minimum allocation for every container request at the RM in terms of virtual CPU cores. Requests lower than this will be set to the value of this property. Additionally, a node manager that is configured to have fewer virtual cores than this value will be shut down by the resource manager. yarn.scheduler.minimum-allocation-vcores 1 The maximum allocation for every container request at the RM in terms of virtual CPU cores. Requests higher than this will throw an InvalidResourceRequestException. yarn.scheduler.maximum-allocation-vcores 4 Used by node labels. If set to true, the port should be included in the node name. Only usable if your scheduler supports node labels. yarn.scheduler.include-port-in-node-name false Enable RM to recover state after starting. If true, then yarn.resourcemanager.store.class must be specified. yarn.resourcemanager.recovery.enabled false Should RM fail fast if it encounters any errors. By defalt, it points to ${yarn.fail-fast}. Errors include: 1) exceptions when state-store write/read operations fails. yarn.resourcemanager.fail-fast ${yarn.fail-fast} Should YARN fail fast if it encounters any errors. This is a global config for all other components including RM,NM etc. If no value is set for component-specific config (e.g yarn.resourcemanager.fail-fast), this value will be the default. yarn.fail-fast false Enable RM work preserving recovery. This configuration is private to YARN for experimenting the feature. yarn.resourcemanager.work-preserving-recovery.enabled true Set the amount of time RM waits before allocating new containers on work-preserving-recovery. Such wait period gives RM a chance to settle down resyncing with NMs in the cluster on recovery, before assigning new containers to applications. yarn.resourcemanager.work-preserving-recovery.scheduling-wait-ms 10000 The class to use as the persistent store. If org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore is used, the store is implicitly fenced; meaning a single ResourceManager is able to use the store at any point in time. More details on this implicit fencing, along with setting up appropriate ACLs is discussed under yarn.resourcemanager.zk-state-store.root-node.acl. yarn.resourcemanager.store.class org.apache.hadoop.yarn.server.resourcemanager.recovery.FileSystemRMStateStore When automatic failover is enabled, number of zookeeper operation retry times in ActiveStandbyElector yarn.resourcemanager.ha.failover-controller.active-standby-elector.zk.retries The maximum number of completed applications RM state store keeps, less than or equals to ${yarn.resourcemanager.max-completed-applications}. By default, it equals to ${yarn.resourcemanager.max-completed-applications}. This ensures that the applications kept in the state store are consistent with the applications remembered in RM memory. Any values larger than ${yarn.resourcemanager.max-completed-applications} will be reset to ${yarn.resourcemanager.max-completed-applications}. Note that this value impacts the RM recovery performance. Typically, a smaller value indicates better performance on RM recovery. yarn.resourcemanager.state-store.max-completed-applications ${yarn.resourcemanager.max-completed-applications} Full path of the ZooKeeper znode where RM state will be stored. This must be supplied when using org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore as the value for yarn.resourcemanager.store.class yarn.resourcemanager.zk-state-store.parent-path /rmstore ACLs to be used for the root znode when using ZKRMStateStore in an HA scenario for fencing. ZKRMStateStore supports implicit fencing to allow a single ResourceManager write-access to the store. For fencing, the ResourceManagers in the cluster share read-write-admin privileges on the root node, but the Active ResourceManager claims exclusive create-delete permissions. By default, when this property is not set, we use the ACLs from yarn.resourcemanager.zk-acl for shared admin access and rm-address:random-number for username-based exclusive create-delete access. This property allows users to set ACLs of their choice instead of using the default mechanism. For fencing to work, the ACLs should be carefully set differently on each ResourceManager such that all the ResourceManagers have shared admin access and the Active ResourceManager takes over (exclusively) the create-delete access. yarn.resourcemanager.zk-state-store.root-node.acl URI pointing to the location of the FileSystem path where RM state will be stored. This must be supplied when using org.apache.hadoop.yarn.server.resourcemanager.recovery.FileSystemRMStateStore as the value for yarn.resourcemanager.store.class yarn.resourcemanager.fs.state-store.uri ${hadoop.tmp.dir}/yarn/system/rmstore the number of retries to recover from IOException in FileSystemRMStateStore. yarn.resourcemanager.fs.state-store.num-retries 0 Retry interval in milliseconds in FileSystemRMStateStore. yarn.resourcemanager.fs.state-store.retry-interval-ms 1000 Local path where the RM state will be stored when using org.apache.hadoop.yarn.server.resourcemanager.recovery.LeveldbRMStateStore as the value for yarn.resourcemanager.store.class yarn.resourcemanager.leveldb-state-store.path ${hadoop.tmp.dir}/yarn/system/rmstore The time in seconds between full compactions of the leveldb database. Setting the interval to zero disables the full compaction cycles. yarn.resourcemanager.leveldb-state-store.compaction-interval-secs 3600 Enable RM high-availability. When enabled, (1) The RM starts in the Standby mode by default, and transitions to the Active mode when prompted to. (2) The nodes in the RM ensemble are listed in yarn.resourcemanager.ha.rm-ids (3) The id of each RM either comes from yarn.resourcemanager.ha.id if yarn.resourcemanager.ha.id is explicitly specified or can be figured out by matching yarn.resourcemanager.address.{id} with local address (4) The actual physical addresses come from the configs of the pattern - {rpc-config}.{id} yarn.resourcemanager.ha.enabled false Enable automatic failover. By default, it is enabled only when HA is enabled yarn.resourcemanager.ha.automatic-failover.enabled true Enable embedded automatic failover. By default, it is enabled only when HA is enabled. The embedded elector relies on the RM state store to handle fencing, and is primarily intended to be used in conjunction with ZKRMStateStore. yarn.resourcemanager.ha.automatic-failover.embedded true The base znode path to use for storing leader information, when using ZooKeeper based leader election. yarn.resourcemanager.ha.automatic-failover.zk-base-path /yarn-leader-election Index at which last section of application id (with each section separated by _ in application id) will be split so that application znode stored in zookeeper RM state store will be stored as two different znodes (parent-child). Split is done from the end. For instance, with no split, appid znode will be of the form application_1352994193343_0001. If the value of this config is 1, the appid znode will be broken into two parts application_1352994193343_000 and 1 respectively with former being the parent node. application_1352994193343_0002 will then be stored as 2 under the parent node application_1352994193343_000. This config can take values from 0 to 4. 0 means there will be no split. If configuration value is outside this range, it will be treated as config value of 0(i.e. no split). A value larger than 0 (up to 4) should be configured if you are storing a large number of apps in ZK based RM state store and state store operations are failing due to LenError in Zookeeper. yarn.resourcemanager.zk-appid-node.split-index 0 Index at which the RM Delegation Token ids will be split so that the delegation token znodes stored in the zookeeper RM state store will be stored as two different znodes (parent-child). The split is done from the end. For instance, with no split, a delegation token znode will be of the form RMDelegationToken_123456789. If the value of this config is 1, the delegation token znode will be broken into two parts: RMDelegationToken_12345678 and 9 respectively with former being the parent node. This config can take values from 0 to 4. 0 means there will be no split. If the value is outside this range, it will be treated as 0 (i.e. no split). A value larger than 0 (up to 4) should be configured if you are running a large number of applications, with long-lived delegation tokens and state store operations (e.g. failover) are failing due to LenError in Zookeeper. yarn.resourcemanager.zk-delegation-token-node.split-index 0 Specifies the maximum size of the data that can be stored in a znode. Value should be same or less than jute.maxbuffer configured in zookeeper. Default value configured is 1MB. yarn.resourcemanager.zk-max-znode-size.bytes 1048576 Name of the cluster. In a HA setting, this is used to ensure the RM participates in leader election for this cluster and ensures it does not affect other clusters yarn.resourcemanager.cluster-id The range of values above base epoch that the RM will use before wrapping around yarn.resourcemanager.epoch.range 0 The list of RM nodes in the cluster when HA is enabled. See description of yarn.resourcemanager.ha .enabled for full details on how this is used. yarn.resourcemanager.ha.rm-ids The id (string) of the current RM. When HA is enabled, this is an optional config. The id of current RM can be set by explicitly specifying yarn.resourcemanager.ha.id or figured out by matching yarn.resourcemanager.address.{id} with local address See description of yarn.resourcemanager.ha.enabled for full details on how this is used. yarn.resourcemanager.ha.id When HA is enabled, the class to be used by Clients, AMs and NMs to failover to the Active RM. It should extend org.apache.hadoop.yarn.client.RMFailoverProxyProvider yarn.client.failover-proxy-provider org.apache.hadoop.yarn.client.ConfiguredRMFailoverProxyProvider When HA is not enabled, the class to be used by Clients, AMs and NMs to retry connecting to the Active RM. It should extend org.apache.hadoop.yarn.client.RMFailoverProxyProvider yarn.client.failover-no-ha-proxy-provider org.apache.hadoop.yarn.client.DefaultNoHARMFailoverProxyProvider When HA is enabled, the max number of times FailoverProxyProvider should attempt failover. When set, this overrides the yarn.resourcemanager.connect.max-wait.ms. When not set, this is inferred from yarn.resourcemanager.connect.max-wait.ms. yarn.client.failover-max-attempts When HA is enabled, the sleep base (in milliseconds) to be used for calculating the exponential delay between failovers. When set, this overrides the yarn.resourcemanager.connect.* settings. When not set, yarn.resourcemanager.connect.retry-interval.ms is used instead. yarn.client.failover-sleep-base-ms When HA is enabled, the maximum sleep time (in milliseconds) between failovers. When set, this overrides the yarn.resourcemanager.connect.* settings. When not set, yarn.resourcemanager.connect.retry-interval.ms is used instead. yarn.client.failover-sleep-max-ms When HA is enabled, the number of retries per attempt to connect to a ResourceManager. In other words, it is the ipc.client.connect.max.retries to be used during failover attempts yarn.client.failover-retries 0 When HA is enabled, the number of retries per attempt to connect to a ResourceManager on socket timeouts. In other words, it is the ipc.client.connect.max.retries.on.timeouts to be used during failover attempts yarn.client.failover-retries-on-socket-timeouts 0 The maximum number of completed applications RM keeps. yarn.resourcemanager.max-completed-applications 1000 Interval at which the delayed token removal thread runs yarn.resourcemanager.delayed.delegation-token.removal-interval-ms 30000 Maximum size in bytes for configurations that can be provided by application to RM for delegation token renewal. By experiment, its roughly 128 bytes per key-value pair. The default value 12800 allows roughly 100 configs, may be less. yarn.resourcemanager.delegation-token.max-conf-size-bytes 12800 If true, ResourceManager will always try to cancel delegation tokens after the application completes, even if the client sets shouldCancelAtEnd false. References to delegation tokens are tracked, so they will not be canceled until all sub-tasks are done using them. yarn.resourcemanager.delegation-token.always-cancel false If true, ResourceManager will have proxy-user privileges. Use case: In a secure cluster, YARN requires the user hdfs delegation-tokens to do localization and log-aggregation on behalf of the user. If this is set to true, ResourceManager is able to request new hdfs delegation tokens on behalf of the user. This is needed by long-running-service, because the hdfs tokens will eventually expire and YARN requires new valid tokens to do localization and log-aggregation. Note that to enable this use case, the corresponding HDFS NameNode has to configure ResourceManager as the proxy-user so that ResourceManager can itself ask for new tokens on behalf of the user when tokens are past their max-life-time. yarn.resourcemanager.proxy-user-privileges.enabled false Interval for the roll over for the master key used to generate application tokens yarn.resourcemanager.am-rm-tokens.master-key-rolling-interval-secs 86400 Interval for the roll over for the master key used to generate container tokens. It is expected to be much greater than yarn.nm.liveness-monitor.expiry-interval-ms and yarn.resourcemanager.rm.container-allocation.expiry-interval-ms. Otherwise the behavior is undefined. yarn.resourcemanager.container-tokens.master-key-rolling-interval-secs 86400 The heart-beat interval in milliseconds for every NodeManager in the cluster. yarn.resourcemanager.nodemanagers.heartbeat-interval-ms 1000 Enables heart-beat interval scaling. The NodeManager heart-beat interval will scale based on the difference between the CPU utilization on the node and the cluster-wide average CPU utilization. yarn.resourcemanager.nodemanagers.heartbeat-interval-scaling-enable false If heart-beat interval scaling is enabled, this is the minimum heart-beat interval in milliseconds yarn.resourcemanager.nodemanagers.heartbeat-interval-min-ms 1000 If heart-beat interval scaling is enabled, this is the maximum heart-beat interval in milliseconds yarn.resourcemanager.nodemanagers.heartbeat-interval-max-ms 1000 If heart-beat interval scaling is enabled, this controls the degree of adjustment when speeding up heartbeat intervals. At 1.0, 20% less than average CPU utilization will result in a 20% decrease in heartbeat interval. yarn.resourcemanager.nodemanagers.heartbeat-interval-speedup-factor 1.0 If heart-beat interval scaling is enabled, this controls the degree of adjustment when slowing down heartbeat intervals. At 1.0, 20% greater than average CPU utilization will result in a 20% increase in heartbeat interval. yarn.resourcemanager.nodemanagers.heartbeat-interval-slowdown-factor 1.0 The Number of consecutive missed heartbeats after which node will be skipped from scheduling yarn.scheduler.skip.node.multiplier 2 The minimum allowed version of a connecting nodemanager. The valid values are NONE (no version checking), EqualToRM (the nodemanager's version is equal to or greater than the RM version), or a Version String. yarn.resourcemanager.nodemanager.minimum.version NONE Enable a set of periodic monitors (specified in yarn.resourcemanager.scheduler.monitor.policies) that affect the scheduler. yarn.resourcemanager.scheduler.monitor.enable false The list of SchedulingEditPolicy classes that interact with the scheduler. A particular module may be incompatible with the scheduler, other policies, or a configuration of either. yarn.resourcemanager.scheduler.monitor.policies org.apache.hadoop.yarn.server.resourcemanager.monitor.capacity.ProportionalCapacityPreemptionPolicy The class to use as the configuration provider. If org.apache.hadoop.yarn.LocalConfigurationProvider is used, the local configuration will be loaded. If org.apache.hadoop.yarn.FileSystemBasedConfigurationProvider is used, the configuration which will be loaded should be uploaded to remote File system first. yarn.resourcemanager.configuration.provider-class org.apache.hadoop.yarn.LocalConfigurationProvider The value specifies the file system (e.g. HDFS) path where ResourceManager loads configuration if yarn.resourcemanager.configuration.provider-class is set to org.apache.hadoop.yarn.FileSystemBasedConfigurationProvider. yarn.resourcemanager.configuration.file-system-based-store /yarn/conf The setting that controls whether yarn system metrics is published to the Timeline server (version one) or not, by RM. This configuration is now deprecated in favor of yarn.system-metrics-publisher.enabled. yarn.resourcemanager.system-metrics-publisher.enabled false The setting that controls whether yarn system metrics is published on the Timeline service or not by RM And NM. yarn.system-metrics-publisher.enabled false The setting that controls whether yarn container events are published to the timeline service or not by RM. This configuration setting is for ATS V2. yarn.rm.system-metrics-publisher.emit-container-events false Number of worker threads that send the yarn system metrics data. yarn.resourcemanager.system-metrics-publisher.dispatcher.pool-size 10 This setting enables/disables timeline server v1 publisher to publish timeline events in batch. yarn.resourcemanager.system-metrics-publisher.timeline-server-v1.enable-batch false The size of timeline server v1 publisher sending events in one request. yarn.resourcemanager.system-metrics-publisher.timeline-server-v1.batch-size 1000 When enable batch publishing in timeline server v1, we must avoid that the publisher waits for a batch to be filled up and hold events in buffer for long time. So we add another thread which send event's in the buffer periodically. This config sets the interval of the cyclical sending thread. yarn.resourcemanager.system-metrics-publisher.timeline-server-v1.interval-seconds 60 Number of diagnostics/failure messages can be saved in RM for log aggregation. It also defines the number of diagnostics/failure messages can be shown in log aggregation web ui. yarn.resourcemanager.max-log-aggregation-diagnostics-in-memory 10 RM DelegationTokenRenewer thread count yarn.resourcemanager.delegation-token-renewer.thread-count 50 RM secret key update interval in ms yarn.resourcemanager.delegation.key.update-interval 86400000 RM delegation token maximum lifetime in ms yarn.resourcemanager.delegation.token.max-lifetime 604800000 RM delegation token update interval in ms yarn.resourcemanager.delegation.token.renew-interval 86400000 This configuration is used for how often the tokens are scanned for expired tokens in milliseconds. the background thread(delegation token remover thread) will delete expired tokens after the configured time. the default value is 1h. yarn.resourcemanager.delegation.token.remove-scan-interval 1h RM DelegationTokenRenewer thread timeout yarn.resourcemanager.delegation-token-renewer.thread-timeout 60s Default maximum number of retries for each RM DelegationTokenRenewer thread yarn.resourcemanager.delegation-token-renewer.thread-retry-max-attempts 10 Time interval between each RM DelegationTokenRenewer thread retry attempt yarn.resourcemanager.delegation-token-renewer.thread-retry-interval 60s Thread pool size for RMApplicationHistoryWriter. yarn.resourcemanager.history-writer.multi-threaded-dispatcher.pool-size 10 Comma-separated list of values (in minutes) for schedule queue related metrics. yarn.resourcemanager.metrics.runtime.buckets 60,300,1440 Interval for the roll over for the master key used to generate NodeManager tokens. It is expected to be set to a value much larger than yarn.nm.liveness-monitor.expiry-interval-ms. yarn.resourcemanager.nm-tokens.master-key-rolling-interval-secs 86400 Flag to enable the ResourceManager reservation system. yarn.resourcemanager.reservation-system.enable false The Java class to use as the ResourceManager reservation system. By default, is set to org.apache.hadoop.yarn.server.resourcemanager.reservation.CapacityReservationSystem when using CapacityScheduler and is set to org.apache.hadoop.yarn.server.resourcemanager.reservation.FairReservationSystem when using FairScheduler. yarn.resourcemanager.reservation-system.class The plan follower policy class name to use for the ResourceManager reservation system. By default, is set to org.apache.hadoop.yarn.server.resourcemanager.reservation.CapacitySchedulerPlanFollower is used when using CapacityScheduler, and is set to org.apache.hadoop.yarn.server.resourcemanager.reservation.FairSchedulerPlanFollower when using FairScheduler. yarn.resourcemanager.reservation-system.plan.follower Step size of the reservation system in ms yarn.resourcemanager.reservation-system.planfollower.time-step 1000 The expiry interval for a container yarn.resourcemanager.rm.container-allocation.expiry-interval-ms 600000 Flag to enable/disable resource profiles yarn.resourcemanager.resource-profiles.enabled false If resource profiles is enabled, source file for the profiles yarn.resourcemanager.resource-profiles.source-file resource-profiles.json The hostname of the NM. yarn.nodemanager.hostname 0.0.0.0 The address of the container manager in the NM. yarn.nodemanager.address ${yarn.nodemanager.hostname}:0 The actual address the server will bind to. If this optional address is set, the RPC and webapp servers will bind to this address and the port specified in yarn.nodemanager.address and yarn.nodemanager.webapp.address, respectively. This is most useful for making NM listen to all interfaces by setting to 0.0.0.0. yarn.nodemanager.bind-host Environment variables that should be forwarded from the NodeManager's environment to the container's, specified as a comma separated list of VARNAME=value pairs. To define environment variables individually, you can specify multiple properties of the form yarn.nodemanager.admin-env.VARNAME, where VARNAME is the name of the environment variable. This is the only way to add a variable when its value contains commas. yarn.nodemanager.admin-env MALLOC_ARENA_MAX=$MALLOC_ARENA_MAX * PATH components that will be prepended to the user's path. * If this is defined and the user does not define PATH, NM will also * append ":$PATH" to prevent this from eclipsing the PATH defined in * the container. This feature is only available for Linux. yarn.nodemanager.force.path Environment variables that containers may override rather than use NodeManager's default. yarn.nodemanager.env-whitelist JAVA_HOME,HADOOP_COMMON_HOME,HADOOP_HDFS_HOME,HADOOP_CONF_DIR,CLASSPATH_PREPEND_DISTCACHE,HADOOP_YARN_HOME,HADOOP_HOME,PATH,LANG,TZ who will execute(launch) the containers. yarn.nodemanager.container-executor.class org.apache.hadoop.yarn.server.nodemanager.DefaultContainerExecutor Comma separated List of container state transition listeners. yarn.nodemanager.container-state-transition-listener.classes Number of threads container manager uses. yarn.nodemanager.container-manager.thread-count 20 Number of threads collector service uses. yarn.nodemanager.collector-service.thread-count 5 Number of threads used in cleanup. yarn.nodemanager.delete.thread-count 4 How long the container executor should wait for the exit code file to appear after a reacquired container has exited. yarn.nodemanager.container-executor.exit-code-file.timeout-ms 2000 At the NM, the policy to determine whether to queue an OPPORTUNISTIC container or not. If set to BY_QUEUE_LEN, uses the queue capacity, as set by yarn.nodemanager.opportunistic-containers-max-queue-length to limit how many containers to accept/queue. If set to BY_RESOURCES, limits the number of containers accepted based on the resource capacity of the node. yarn.nodemanager.opportunistic-containers-queue-policy BY_QUEUE_LEN Max number of OPPORTUNISTIC containers to queue at the nodemanager (NM). If the value is 0 or negative, NMs do not allow any OPPORTUNISTIC containers. If the value is positive, the NM caps the number of OPPORTUNISTIC containers that can be queued at the NM. yarn.nodemanager.opportunistic-containers-max-queue-length 0 Number of seconds after an application finishes before the nodemanager's DeletionService will delete the application's localized file directory and log directory. To diagnose YARN application problems, set this property's value large enough (for example, to 600 = 10 minutes) to permit examination of these directories. After changing the property's value, you must restart the nodemanager in order for it to have an effect. The roots of YARN applications' work directories is configurable with the yarn.nodemanager.local-dirs property (see below), and the roots of the YARN applications' log directories is configurable with the yarn.nodemanager.log-dirs property (see also below). yarn.nodemanager.delete.debug-delay-sec 0 Keytab for NM. yarn.nodemanager.keytab /etc/krb5.keytab List of directories to store localized files in. An application's localized file directory will be found in: ${yarn.nodemanager.local-dirs}/usercache/${user}/appcache/application_${appid}. Individual containers' work directories, called container_${contid}, will be subdirectories of this. yarn.nodemanager.local-dirs ${hadoop.tmp.dir}/nm-local-dir It limits the maximum number of files which will be localized in a single local directory. If the limit is reached then sub-directories will be created and new files will be localized in them. If it is set to a value less than or equal to 36 [which are sub-directories (0-9 and then a-z)] then NodeManager will fail to start. For example; [for public cache] if this is configured with a value of 40 ( 4 files + 36 sub-directories) and the local-dir is "/tmp/local-dir1" then it will allow 4 files to be created directly inside "/tmp/local-dir1/filecache". For files that are localized further it will create a sub-directory "0" inside "/tmp/local-dir1/filecache" and will localize files inside it until it becomes full. If a file is removed from a sub-directory that is marked full, then that sub-directory will be used back again to localize files. yarn.nodemanager.local-cache.max-files-per-directory 8192 Address where the localizer IPC is. yarn.nodemanager.localizer.address ${yarn.nodemanager.hostname}:8040 Address where the collector service IPC is. yarn.nodemanager.collector-service.address ${yarn.nodemanager.hostname}:8048 The setting that controls whether yarn container events are published to the timeline service or not by NM. This configuration setting is for ATS V2. yarn.nodemanager.emit-container-events true Interval in between cache cleanups. yarn.nodemanager.localizer.cache.cleanup.interval-ms 600000 Target size of localizer cache in MB, per nodemanager. It is a target retention size that only includes resources with PUBLIC and PRIVATE visibility and excludes resources with APPLICATION visibility yarn.nodemanager.localizer.cache.target-size-mb 10240 Number of threads to handle localization requests. yarn.nodemanager.localizer.client.thread-count 5 Number of threads to use for localization fetching. yarn.nodemanager.localizer.fetch.thread-count 4 yarn.nodemanager.container-localizer.java.opts -Xmx256m The log level for container localizer while it is an independent process. yarn.nodemanager.container-localizer.log.level INFO Where to store container logs. An application's localized log directory will be found in ${yarn.nodemanager.log-dirs}/application_${appid}. Individual containers' log directories will be below this, in directories named container_{$contid}. Each container directory will contain the files stderr, stdin, and syslog generated by that container. yarn.nodemanager.log-dirs ${yarn.log.dir}/userlogs The permissions settings used for the creation of container directories when using DefaultContainerExecutor. This follows standard user/group/all permissions format. yarn.nodemanager.default-container-executor.log-dirs.permissions 710 Whether to enable log aggregation. Log aggregation collects each container's logs and moves these logs onto a file-system, for e.g. HDFS, after the application completes. Users can configure the "yarn.nodemanager.remote-app-log-dir" and "yarn.nodemanager.remote-app-log-dir-suffix" properties to determine where these logs are moved to. Users can access the logs via the Application Timeline Server. yarn.log-aggregation-enable false Whether to enable application placement based on user ID passed via application tags. When it is enabled, userid=<userId> pattern will be checked and if found, the application will be placed onto the found user's queue, if the original user has enough rights on the passed user's queue. yarn.resourcemanager.application-tag-based-placement.enable false Comma separated list of users who can use the application tag based placement, if it is enabled. yarn.resourcemanager.application-tag-based-placement.username.whitelist How long to keep aggregation logs before deleting them. -1 disables. Be careful set this too small and you will spam the name node. yarn.log-aggregation.retain-seconds -1 How long to wait between aggregated log retention checks. If set to 0 or a negative value then the value is computed as one-tenth of the aggregated log retention time. Be careful set this too small and you will spam the name node. yarn.log-aggregation.retain-check-interval-seconds -1 The log files created under NM Local Directories will be logged if it exceeds the configured bytes. This only takes effect if log4j level is at least Debug. yarn.log-aggregation.debug.filesize 104857600 Specify which log file controllers we will support. The first file controller we add will be used to write the aggregated logs. This comma separated configuration will work with the configuration: yarn.log-aggregation.file-controller.%s.class which defines the supported file controller's class. By default, the TFile controller would be used. The user could override this configuration by adding more file controllers. To support back-ward compatibility, make sure that we always add TFile file controller. yarn.log-aggregation.file-formats TFile Class that supports TFile read and write operations. yarn.log-aggregation.file-controller.TFile.class org.apache.hadoop.yarn.logaggregation.filecontroller.tfile.LogAggregationTFileController How long for ResourceManager to wait for NodeManager to report its log aggregation status. If waiting time of which the log aggregation status is reported from NodeManager exceeds the configured value, RM will report log aggregation status for this NodeManager as TIME_OUT. This configuration will be used in NodeManager as well to decide whether and when to delete the cached log aggregation status. yarn.log-aggregation-status.time-out.ms 600000 Whether to clean up nodemanager logs when log aggregation is enabled. Setting to false disables the cleanup nodemanager logging, and it causes disk full in the long run. Users can set to false for test-only purpose. yarn.log-aggregation.enable-local-cleanup true Time in seconds to retain user logs. Only applicable if log aggregation is disabled yarn.nodemanager.log.retain-seconds 10800 Where to aggregate logs to. yarn.nodemanager.remote-app-log-dir /tmp/logs The remote log dir will be created at {yarn.nodemanager.remote-app-log-dir}/${user}/{thisParam} yarn.nodemanager.remote-app-log-dir-suffix logs If set to true, the older application log directory will be considered while fetching application logs. yarn.nodemanager.remote-app-log-dir-include-older true If the NodeManager creates the remote-app-log-dir folder, it will be created with this groupname. yarn.nodemanager.remote-app-log-dir.groupname Generate additional logs about container launches. Currently, this creates a copy of the launch script and lists the directory contents of the container work dir. When listing directory contents, we follow symlinks to a max-depth of 5(including symlinks which point to outside the container work dir) which may lead to a slowness in launching containers. yarn.nodemanager.log-container-debug-info.enabled true Amount of physical memory, in MB, that can be allocated for containers. If set to -1 and yarn.nodemanager.resource.detect-hardware-capabilities is true, it is automatically calculated(in case of Windows and Linux). In other cases, the default is 8192MB. yarn.nodemanager.resource.memory-mb -1 Amount of physical memory, in MB, that is reserved for non-YARN processes. This configuration is only used if yarn.nodemanager.resource.detect-hardware-capabilities is set to true and yarn.nodemanager.resource.memory-mb is -1. If set to -1, this amount is calculated as 20% of (system memory - 2*HADOOP_HEAPSIZE) yarn.nodemanager.resource.system-reserved-memory-mb -1 Whether YARN CGroups memory tracking is enabled. yarn.nodemanager.resource.memory.enabled false Whether YARN CGroups strict memory enforcement is enabled. yarn.nodemanager.resource.memory.enforced true If memory limit is enforced, this the percentage of soft limit compared to the memory assigned to the container. If there is memory pressure container memory usage will be pushed back to its soft limit by swapping out memory. yarn.nodemanager.resource.memory.cgroups.soft-limit-percentage 90.0 Container swappiness is the likelihood a page will be swapped out compared to be kept in memory. Value is between 0-100. yarn.nodemanager.resource.memory.cgroups.swappiness 0 Whether physical memory limits will be enforced for containers. yarn.nodemanager.pmem-check-enabled true Whether virtual memory limits will be enforced for containers. yarn.nodemanager.vmem-check-enabled true Ratio between virtual memory to physical memory when setting memory limits for containers. Container allocations are expressed in terms of physical memory, and virtual memory usage is allowed to exceed this allocation by this ratio. yarn.nodemanager.vmem-pmem-ratio 2.1 Number of vcores that can be allocated for containers. This is used by the RM scheduler when allocating resources for containers. This is not used to limit the number of CPUs used by YARN containers. If it is set to -1 and yarn.nodemanager.resource.detect-hardware-capabilities is true, it is automatically determined from the hardware in case of Windows and Linux. In other cases, number of vcores is 8 by default. yarn.nodemanager.resource.cpu-vcores -1 Flag to determine if logical processors(such as hyperthreads) should be counted as cores. Only applicable on Linux when yarn.nodemanager.resource.cpu-vcores is set to -1 and yarn.nodemanager.resource.detect-hardware-capabilities is true. yarn.nodemanager.resource.count-logical-processors-as-cores false Multiplier to determine how to convert phyiscal cores to vcores. This value is used if yarn.nodemanager.resource.cpu-vcores is set to -1(which implies auto-calculate vcores) and yarn.nodemanager.resource.detect-hardware-capabilities is set to true. The number of vcores will be calculated as number of CPUs * multiplier. yarn.nodemanager.resource.pcores-vcores-multiplier 1.0 Thread pool size for LogAggregationService in Node Manager. yarn.nodemanager.logaggregation.threadpool-size-max 100 Percentage of CPU that can be allocated for containers. This setting allows users to limit the amount of CPU that YARN containers use. Currently functional only on Linux using cgroups. The default is to use 100% of CPU. yarn.nodemanager.resource.percentage-physical-cpu-limit 100 Enable auto-detection of node capabilities such as memory and CPU. yarn.nodemanager.resource.detect-hardware-capabilities false NM Webapp address. yarn.nodemanager.webapp.address ${yarn.nodemanager.hostname}:8042 The https adddress of the NM web application. yarn.nodemanager.webapp.https.address 0.0.0.0:8044 The Kerberos keytab file to be used for spnego filter for the NM web interface. yarn.nodemanager.webapp.spnego-keytab-file The Kerberos principal to be used for spnego filter for the NM web interface. yarn.nodemanager.webapp.spnego-principal How often to monitor the node and the containers. If 0 or negative, monitoring is disabled. yarn.nodemanager.resource-monitor.interval-ms 3000 Class that calculates current resource utilization. yarn.nodemanager.resource-calculator.class Enable container monitor yarn.nodemanager.container-monitor.enabled true How often to monitor containers. If not set, the value for yarn.nodemanager.resource-monitor.interval-ms will be used. If 0 or negative, container monitoring is disabled. yarn.nodemanager.container-monitor.interval-ms Flag to enable the container log monitor which enforces container log directory size limits. yarn.nodemanager.container-log-monitor.enable false How often to check the usage of a container's log directories in milliseconds yarn.nodemanager.container-log-monitor.interval-ms 60000 The disk space limit, in bytes, for a single container log directory yarn.nodemanager.container-log-monitor.dir-size-limit-bytes 1000000000 The disk space limit, in bytes, for all of a container's logs yarn.nodemanager.container-log-monitor.total-size-limit-bytes 10000000000 Class that calculates containers current resource utilization. If not set, the value for yarn.nodemanager.resource-calculator.class will be used. yarn.nodemanager.container-monitor.resource-calculator.class The nodemanager health check scripts to run. yarn.nodemanager.health-checker.scripts script Health check script time out period. yarn.nodemanager.health-checker.timeout-ms 1200000 Whether or not to run the node health script before the NM starts up. yarn.nodemanager.health-checker.run-before-startup false Frequency of running node health scripts. yarn.nodemanager.health-checker.interval-ms 600000 Frequency of running disk health checker code. yarn.nodemanager.disk-health-checker.interval-ms 120000 The minimum fraction of number of disks to be healthy for the nodemanager to launch new containers. This correspond to both yarn.nodemanager.local-dirs and yarn.nodemanager.log-dirs. i.e. If there are less number of healthy local-dirs (or log-dirs) available, then new containers will not be launched on this node. yarn.nodemanager.disk-health-checker.min-healthy-disks 0.25 Enable/Disable the disk utilisation percentage threshold for disk health checker. yarn.nodemanager.disk-health-checker.disk-utilization-threshold.enabled true Enable/Disable the minimum disk free space threshold for disk health checker. yarn.nodemanager.disk-health-checker.disk-free-space-threshold.enabled true The maximum percentage of disk space utilization allowed after which a disk is marked as bad. Values can range from 0.0 to 100.0. If the value is greater than or equal to 100, the nodemanager will check for full disk. This applies to yarn.nodemanager.local-dirs and yarn.nodemanager.log-dirs when yarn.nodemanager.disk-health-checker.disk-utilization-threshold.enabled is true. yarn.nodemanager.disk-health-checker.max-disk-utilization-per-disk-percentage 90.0 The low threshold percentage of disk space used when a bad disk is marked as good. Values can range from 0.0 to 100.0. This applies to yarn.nodemanager.local-dirs and yarn.nodemanager.log-dirs. Note that if its value is more than yarn.nodemanager.disk-health-checker. max-disk-utilization-per-disk-percentage or not set, it will be set to the same value as yarn.nodemanager.disk-health-checker.max-disk-utilization-per-disk-percentage. yarn.nodemanager.disk-health-checker.disk-utilization-watermark-low-per-disk-percentage The minimum space in megabytes that must be available on a disk for it to be used. If space on a disk falls below this threshold, it will be marked as bad. This applies to yarn.nodemanager.local-dirs and yarn.nodemanager.log-dirs when yarn.nodemanager.disk-health-checker.disk-free-space-threshold.enabled is true. yarn.nodemanager.disk-health-checker.min-free-space-per-disk-mb 0 The minimum space in megabytes that must be available on a bad disk for it to be marked as good. This value should not be less than yarn.nodemanager.disk-health-checker.min-free-space-per-disk-mb. If it is less than yarn.nodemanager.disk-health-checker.min-free-space-per-disk-mb, or it is not set, it will be set to the same value as yarn.nodemanager.disk-health-checker.min-free-space-per-disk-mb. This applies to yarn.nodemanager.local-dirs and yarn.nodemanager.log-dirs. yarn.nodemanager.disk-health-checker.min-free-space-per-disk-watermark-high-mb 0 The path to the Linux container executor. yarn.nodemanager.linux-container-executor.path The class which should help the LCE handle resources. yarn.nodemanager.linux-container-executor.resources-handler.class org.apache.hadoop.yarn.server.nodemanager.util.DefaultLCEResourcesHandler The cgroups hierarchy under which to place YARN proccesses (cannot contain commas). If yarn.nodemanager.linux-container-executor.cgroups.mount is false (that is, if cgroups have been pre-configured) and the YARN user has write access to the parent directory, then the directory will be created. If the directory already exists, the administrator has to give YARN write permissions to it recursively. This property only applies when the LCE resources handler is set to CgroupsLCEResourcesHandler. yarn.nodemanager.linux-container-executor.cgroups.hierarchy /hadoop-yarn Whether the LCE should attempt to mount cgroups if not found. This property only applies when the LCE resources handler is set to CgroupsLCEResourcesHandler. yarn.nodemanager.linux-container-executor.cgroups.mount false This property sets the path from which YARN will read the CGroups configuration. YARN has built-in functionality to discover the system CGroup mount paths, so use this property only if YARN's automatic mount path discovery does not work. The path specified by this property must exist before the NodeManager is launched. If yarn.nodemanager.linux-container-executor.cgroups.mount is set to true, YARN will first try to mount the CGroups at the specified path before reading them. If yarn.nodemanager.linux-container-executor.cgroups.mount is set to false, YARN will read the CGroups at the specified path. If this property is empty, YARN tries to detect the CGroups location. Please refer to NodeManagerCgroups.html in the documentation for further details. This property only applies when the LCE resources handler is set to CgroupsLCEResourcesHandler. yarn.nodemanager.linux-container-executor.cgroups.mount-path Delay in ms between attempts to remove linux cgroup yarn.nodemanager.linux-container-executor.cgroups.delete-delay-ms 20 This determines which of the two modes that LCE should use on a non-secure cluster. If this value is set to true, then all containers will be launched as the user specified in yarn.nodemanager.linux-container-executor.nonsecure-mode.local-user. If this value is set to false, then containers will run as the user who submitted the application. yarn.nodemanager.linux-container-executor.nonsecure-mode.limit-users true The UNIX user that containers will run as when Linux-container-executor is used in nonsecure mode (a use case for this is using cgroups) if the yarn.nodemanager.linux-container-executor.nonsecure-mode.limit-users is set to true. yarn.nodemanager.linux-container-executor.nonsecure-mode.local-user nobody The allowed pattern for UNIX user names enforced by Linux-container-executor when used in nonsecure mode (use case for this is using cgroups). The default value is taken from /usr/sbin/adduser yarn.nodemanager.linux-container-executor.nonsecure-mode.user-pattern ^[_.A-Za-z0-9][-@_.A-Za-z0-9]{0,255}?[$]?$ This flag determines whether apps should run with strict resource limits or be allowed to consume spare resources if they need them. For example, turning the flag on will restrict apps to use only their share of CPU, even if the node has spare CPU cycles. The default value is false i.e. use available resources. Please note that turning this flag on may reduce job throughput on the cluster. This setting does not apply to other subsystems like memory. yarn.nodemanager.linux-container-executor.cgroups.strict-resource-usage false Comma separated list of runtimes that are allowed when using LinuxContainerExecutor. The allowed values are default, docker, runc, and javasandbox. yarn.nodemanager.runtime.linux.allowed-runtimes default Default container runtime to use. yarn.nodemanager.runtime.linux.type This configuration setting determines the capabilities assigned to docker containers when they are launched. While these may not be case-sensitive from a docker perspective, it is best to keep these uppercase. To run without any capabilities, set this value to "none" or "NONE" yarn.nodemanager.runtime.linux.docker.capabilities CHOWN,DAC_OVERRIDE,FSETID,FOWNER,MKNOD,NET_RAW,SETGID,SETUID,SETFCAP,SETPCAP,NET_BIND_SERVICE,SYS_CHROOT,KILL,AUDIT_WRITE Default docker image to be used when the docker runtime is selected. yarn.nodemanager.runtime.linux.docker.image-name Default option to decide whether to pull the latest image or not. yarn.nodemanager.runtime.linux.docker.image-update false This configuration setting determines if privileged docker containers are allowed on this cluster. Privileged containers are granted the complete set of capabilities and are not subject to the limitations imposed by the device cgroup controller. In other words, privileged containers can do almost everything that the host can do. Use with extreme care. yarn.nodemanager.runtime.linux.docker.privileged-containers.allowed false This configuration setting determines who is allowed to run privileged docker containers on this cluster. Use with extreme care. yarn.nodemanager.runtime.linux.docker.privileged-containers.acl The set of networks allowed when launching containers using the DockerContainerRuntime. yarn.nodemanager.runtime.linux.docker.allowed-container-networks host,none,bridge The network used when launching containers using the DockerContainerRuntime when no network is specified in the request . This network must be one of the (configurable) set of allowed container networks. yarn.nodemanager.runtime.linux.docker.default-container-network host The set of runtimes allowed when launching containers using the DockerContainerRuntime. yarn.nodemanager.runtime.linux.docker.allowed-container-runtimes runc This configuration setting determines whether the host's PID namespace is allowed for docker containers on this cluster. Use with care. yarn.nodemanager.runtime.linux.docker.host-pid-namespace.allowed false Property to enable docker user remapping yarn.nodemanager.runtime.linux.docker.enable-userremapping.allowed true lower limit for acceptable uids of user remapped user yarn.nodemanager.runtime.linux.docker.userremapping-uid-threshold 1 lower limit for acceptable gids of user remapped user yarn.nodemanager.runtime.linux.docker.userremapping-gid-threshold 1 Whether or not users are allowed to request that Docker containers honor the debug deletion delay. This is useful for troubleshooting Docker container related launch failures. yarn.nodemanager.runtime.linux.docker.delayed-removal.allowed false A configurable value to pass to the Docker Stop command. This value defines the number of seconds between the docker stop command sending a SIGTERM and a SIGKILL. yarn.nodemanager.runtime.linux.docker.stop.grace-period 10 The default list of read-only mounts to be bind-mounted into all Docker containers that use DockerContainerRuntime. yarn.nodemanager.runtime.linux.docker.default-ro-mounts The default list of read-write mounts to be bind-mounted into all Docker containers that use DockerContainerRuntime. yarn.nodemanager.runtime.linux.docker.default-rw-mounts The default list of tmpfs mounts to be mounted into all Docker containers that use DockerContainerRuntime. yarn.nodemanager.runtime.linux.docker.default-tmpfs-mounts The runC image tag to manifest plugin class to be used. yarn.nodemanager.runtime.linux.runc.image-tag-to-manifest-plugin org.apache.hadoop.yarn.server.nodemanager.containermanager.linux.runtime.runc.ImageTagToManifestPlugin The runC manifest to resources plugin class to be used. yarn.nodemanager.runtime.linux.runc.manifest-to-resources-plugin org.apache.hadoop.yarn.server.nodemanager.containermanager.linux.runtime.runc.HdfsManifestToResourcesPlugin The HDFS location under which the oci image manifests, layers, and configs directories exist. yarn.nodemanager.runtime.linux.runc.image-toplevel-dir /runc-root Target count of layer mounts that we should keep on disk at one time. yarn.nodemanager.runtime.linux.runc.layer-mounts-to-keep 100 The interval in seconds between executions of reaping layer mounts. yarn.nodemanager.runtime.linux.runc.layer-mounts-interval-secs 600 Image to be used if no other image is specified. yarn.nodemanager.runtime.linux.runc.image-name Allow or disallow privileged containers. yarn.nodemanager.runtime.linux.runc.privileged-containers.allowed false The set of networks allowed when launching containers using the RuncContainerRuntime. yarn.nodemanager.runtime.linux.runc.allowed-container-networks host,none,bridge The set of runtimes allowed when launching containers using the RuncContainerRuntime. yarn.nodemanager.runtime.linux.runc.allowed-container-runtimes runc ACL list for users allowed to run privileged containers. yarn.nodemanager.runtime.linux.runc.privileged-containers.acl Allow host pid namespace for runC containers. Use with care. yarn.nodemanager.runtime.linux.runc.host-pid-namespace.allowed false The default list of read-only mounts to be bind-mounted into all runC containers that use RuncContainerRuntime. yarn.nodemanager.runtime.linux.runc.default-ro-mounts The default list of read-write mounts to be bind-mounted into all runC containers that use RuncContainerRuntime. yarn.nodemanager.runtime.linux.runc.default-rw-mounts Path to the seccomp profile to use with runC containers yarn.nodemanager.runtime.linux.runc.seccomp-profile The HDFS location where the runC image tag to hash file exists. yarn.nodemanager.runtime.linux.runc.image-tag-to-manifest-plugin.hdfs-hash-file /runc-root/image-tag-to-hash The local file system location where the runC image tag to hash file exists. yarn.nodemanager.runtime.linux.runc.image-tag-to-manifest-plugin.local-hash-file The interval in seconds between refreshing the hdfs image tag to hash cache. yarn.nodemanager.runtime.linux.runc.image-tag-to-manifest-plugin.cache-refresh-interval-secs 60 The number of manifests to cache in the image tag to hash cache. yarn.nodemanager.runtime.linux.runc.image-tag-to-manifest-plugin.num-manifests-to-cache 10 The timeout value in seconds for the values in the stat cache. yarn.nodemanager.runtime.linux.runc.hdfs-manifest-to-resources-plugin.stat-cache-timeout-interval-secs 360 The size of the stat cache which stores stats of the layers and config. yarn.nodemanager.runtime.linux.runc.hdfs-manifest-to-resources-plugin.stat-cache-size 500 The mode in which the Java Container Sandbox should run detailed by the JavaSandboxLinuxContainerRuntime. yarn.nodemanager.runtime.linux.sandbox-mode disabled Permissions for application local directories. yarn.nodemanager.runtime.linux.sandbox-mode.local-dirs.permissions read Location for non-default java policy file. yarn.nodemanager.runtime.linux.sandbox-mode.policy The group which will run by default without the java security manager. yarn.nodemanager.runtime.linux.sandbox-mode.whitelist-group This flag determines whether memory limit will be set for the Windows Job Object of the containers launched by the default container executor. yarn.nodemanager.windows-container.memory-limit.enabled false This flag determines whether CPU limit will be set for the Windows Job Object of the containers launched by the default container executor. yarn.nodemanager.windows-container.cpu-limit.enabled false Interval of time the linux container executor should try cleaning up cgroups entry when cleaning up a container. yarn.nodemanager.linux-container-executor.cgroups.delete-timeout-ms 1000 T-file compression types used to compress aggregated logs. yarn.nodemanager.log-aggregation.compression-type none The kerberos principal for the node manager. yarn.nodemanager.principal A comma separated list of services where service name should only contain a-zA-Z0-9_ and can not start with numbers yarn.nodemanager.aux-services Boolean indicating whether loading aux services from a manifest is enabled. If enabled, aux services may be dynamically modified through reloading the manifest via filesystem changes or a REST API. When enabled, aux services configuration properties unrelated to the manifest will be ignored. yarn.nodemanager.aux-services.manifest.enabled false A file containing auxiliary service specifications. yarn.nodemanager.aux-services.manifest Length of time in ms to wait between reloading aux services manifest. If 0 or less, manifest will not be reloaded. yarn.nodemanager.aux-services.manifest.reload-ms 0 Provide local directory which includes the related jar file as well as all the dependencies’ jar file. We could specify the single jar file or use ${local_dir_to_jar}/* to load all jars under the dep directory. yarn.nodemanager.aux-services.%s.classpath NONE Normally this configuration shouldn't be manually set. The class would be loaded from customized classpath if it does not belong to system-classes. By default, the package org.apache.hadoop is part of the system-classes. If for example the class CustomAuxService is in the package org.apache.hadoop, it will not be loaded from the customized classpath. To solve this, either the package of CustomAuxService could be changed, or a separate system-classes could be configured which excludes the package org.apache.hadoop. yarn.nodemanager.aux-services.%s.system-classes No. of ms to wait between sending a SIGTERM and SIGKILL to a container yarn.nodemanager.sleep-delay-before-sigkill.ms 250 Max time to wait for a process to come up when trying to cleanup a container yarn.nodemanager.process-kill-wait.ms 5000 The minimum allowed version of a resourcemanager that a nodemanager will connect to. The valid values are NONE (no version checking), EqualToNM (the resourcemanager's version is equal to or greater than the NM version), or a Version String. yarn.nodemanager.resourcemanager.minimum.version NONE Maximum size of contain's diagnostics to keep for relaunching container case. yarn.nodemanager.container-diagnostics-maximum-size 10000 Minimum container restart interval in milliseconds. yarn.nodemanager.container-retry-minimum-interval-ms 1000 Max number of threads in NMClientAsync to process container management events yarn.client.nodemanager-client-async.thread-pool-max-size 500 Max time to wait to establish a connection to NM yarn.client.nodemanager-connect.max-wait-ms 180000 Time interval between each attempt to connect to NM yarn.client.nodemanager-connect.retry-interval-ms 10000 Max time to wait for NM to connect to RM. When not set, proxy will fall back to use value of yarn.resourcemanager.connect.max-wait.ms. yarn.nodemanager.resourcemanager.connect.max-wait.ms Time interval between each NM attempt to connect to RM. When not set, proxy will fall back to use value of yarn.resourcemanager.connect.retry-interval.ms. yarn.nodemanager.resourcemanager.connect.retry-interval.ms Maximum number of proxy connections to cache for node managers. If set to a value greater than zero then the cache is enabled and the NMClient and MRAppMaster will cache the specified number of node manager proxies. There will be at max one proxy per node manager. Ex. configuring it to a value of 5 will make sure that client will at max have 5 proxies cached with 5 different node managers. These connections for these proxies will be timed out if idle for more than the system wide idle timeout period. Note that this could cause issues on large clusters as many connections could linger simultaneously and lead to a large number of connection threads. The token used for authentication will be used only at connection creation time. If a new token is received then the earlier connection should be closed in order to use the new token. This and (yarn.client.nodemanager-client-async.thread-pool-max-size) are related and should be in sync (no need for them to be equal). If the value of this property is zero then the connection cache is disabled and connections will use a zero idle timeout to prevent too many connection threads on large clusters. yarn.client.max-cached-nodemanagers-proxies 0 Enable the node manager to recover after starting yarn.nodemanager.recovery.enabled false The local filesystem directory in which the node manager will store state when recovery is enabled. yarn.nodemanager.recovery.dir ${hadoop.tmp.dir}/yarn-nm-recovery The time in seconds between full compactions of the NM state database. Setting the interval to zero disables the full compaction cycles. yarn.nodemanager.recovery.compaction-interval-secs 3600 Whether the nodemanager is running under supervision. A nodemanager that supports recovery and is running under supervision will not try to cleanup containers as it exits with the assumption it will be immediately be restarted and recover containers. yarn.nodemanager.recovery.supervised false Adjustment to the container OS scheduling priority. In Linux, passed directly to the nice command. If unspecified then containers are launched without any explicit OS priority. yarn.nodemanager.container-executor.os.sched.priority.adjustment Flag to enable container metrics yarn.nodemanager.container-metrics.enable true Container metrics flush period in ms. Set to -1 for flush on completion. yarn.nodemanager.container-metrics.period-ms -1 The delay time ms to unregister container metrics after completion. yarn.nodemanager.container-metrics.unregister-delay-ms 10000 Class used to calculate current container resource utilization. yarn.nodemanager.container-monitor.process-tree.class Flag to enable NodeManager disk health checker yarn.nodemanager.disk-health-checker.enable true Number of threads to use in NM log cleanup. Used when log aggregation is disabled. yarn.nodemanager.log.deletion-threads-count 4 The Windows group that the windows-container-executor should run as. yarn.nodemanager.windows-secure-container-executor.group yarn.nodemanager.aux-services.mapreduce_shuffle.class org.apache.hadoop.mapred.ShuffleHandler The kerberos principal for the proxy, if the proxy is not running as part of the RM. yarn.web-proxy.principal Keytab for WebAppProxy, if the proxy is not running as part of the RM. yarn.web-proxy.keytab The address for the web proxy as HOST:PORT, if this is not given then the proxy will run as part of the RM yarn.web-proxy.address The actual address the web proxy will bind to. If this optional address is set, it overrides only the hostname portion of yarn.web-proxy.address. This is useful for making the web proxy server listen on all interfaces by setting it to 0.0.0.0 yarn.web-proxy.bind-host Enable the web proxy connection timeout, default is enabled. yarn.resourcemanager.proxy.timeout.enabled true The web proxy connection timeout. yarn.resourcemanager.proxy.connection.timeout 60000 CLASSPATH for YARN applications. A comma-separated list of CLASSPATH entries. When this value is empty, the following default CLASSPATH for YARN applications would be used. For Linux: $HADOOP_CONF_DIR, $HADOOP_COMMON_HOME/share/hadoop/common/*, $HADOOP_COMMON_HOME/share/hadoop/common/lib/*, $HADOOP_HDFS_HOME/share/hadoop/hdfs/*, $HADOOP_HDFS_HOME/share/hadoop/hdfs/lib/*, $HADOOP_YARN_HOME/share/hadoop/yarn/*, $HADOOP_YARN_HOME/share/hadoop/yarn/lib/* For Windows: %HADOOP_CONF_DIR%, %HADOOP_COMMON_HOME%/share/hadoop/common/*, %HADOOP_COMMON_HOME%/share/hadoop/common/lib/*, %HADOOP_HDFS_HOME%/share/hadoop/hdfs/*, %HADOOP_HDFS_HOME%/share/hadoop/hdfs/lib/*, %HADOOP_YARN_HOME%/share/hadoop/yarn/*, %HADOOP_YARN_HOME%/share/hadoop/yarn/lib/* yarn.application.classpath Indicate what is the current version of the running timeline service. For example, if "yarn.timeline-service.version" is 1.5, and "yarn.timeline-service.enabled" is true, it means the cluster will and should bring up the timeline service v.1.5 (and nothing else). On the client side, if the client uses the same version of timeline service, it should succeed. If the client chooses to use a smaller version in spite of this, then depending on how robust the compatibility story is between versions, the results may vary. yarn.timeline-service.version 1.0f In the server side it indicates whether timeline service is enabled or not. And in the client side, users can enable it to indicate whether client wants to use timeline service. If its enabled in the client side along with security, then yarn client tries to fetch the delegation tokens for the timeline server. yarn.timeline-service.enabled false The hostname of the timeline service web application. yarn.timeline-service.hostname 0.0.0.0 This is default address for the timeline server to start the RPC server. yarn.timeline-service.address ${yarn.timeline-service.hostname}:10200 The http address of the timeline service web application. yarn.timeline-service.webapp.address ${yarn.timeline-service.hostname}:8188 The https address of the timeline service web application. yarn.timeline-service.webapp.https.address ${yarn.timeline-service.hostname}:8190 The actual address the server will bind to. If this optional address is set, the RPC and webapp servers will bind to this address and the port specified in yarn.timeline-service.address and yarn.timeline-service.webapp.address, respectively. This is most useful for making the service listen to all interfaces by setting to 0.0.0.0. yarn.timeline-service.bind-host Defines the max number of applications could be fetched using REST API or application history protocol and shown in timeline server web ui. yarn.timeline-service.generic-application-history.max-applications 10000 Store class name for timeline store. yarn.timeline-service.store-class org.apache.hadoop.yarn.server.timeline.LeveldbTimelineStore Enable age off of timeline store data. yarn.timeline-service.ttl-enable true Time to live for timeline store data in milliseconds. yarn.timeline-service.ttl-ms 604800000 Store file name for leveldb timeline store. yarn.timeline-service.leveldb-timeline-store.path ${hadoop.tmp.dir}/yarn/timeline Length of time to wait between deletion cycles of leveldb timeline store in milliseconds. yarn.timeline-service.leveldb-timeline-store.ttl-interval-ms 300000 Size of read cache for uncompressed blocks for leveldb timeline store in bytes. yarn.timeline-service.leveldb-timeline-store.read-cache-size 104857600 Size of cache for recently read entity start times for leveldb timeline store in number of entities. yarn.timeline-service.leveldb-timeline-store.start-time-read-cache-size 10000 Size of cache for recently written entity start times for leveldb timeline store in number of entities. yarn.timeline-service.leveldb-timeline-store.start-time-write-cache-size 10000 Handler thread count to serve the client RPC requests. yarn.timeline-service.handler-thread-count 10 yarn.timeline-service.http-authentication.type simple Defines authentication used for the timeline server HTTP endpoint. Supported values are: simple | kerberos | #AUTHENTICATION_HANDLER_CLASSNAME# yarn.timeline-service.http-authentication.simple.anonymous.allowed true Indicates if anonymous requests are allowed by the timeline server when using 'simple' authentication. The Kerberos principal for the timeline server. yarn.timeline-service.principal The Kerberos keytab for the timeline server. yarn.timeline-service.keytab /etc/krb5.keytab Comma separated list of UIs that will be hosted yarn.timeline-service.ui-names Default maximum number of retries for timeline service client and value -1 means no limit. yarn.timeline-service.client.max-retries 30 Client policy for whether timeline operations are non-fatal. Should the failure to obtain a delegation token be considered an application failure (option = false), or should the client attempt to continue to publish information without it (option=true) yarn.timeline-service.client.best-effort false Default retry time interval for timeline servive client. yarn.timeline-service.client.retry-interval-ms 1000 The time period for which timeline v2 client will wait for draining leftover entities after stop. yarn.timeline-service.client.drain-entities.timeout.ms 2000 Enable timeline server to recover state after starting. If true, then yarn.timeline-service.state-store-class must be specified. yarn.timeline-service.recovery.enabled false Store class name for timeline state store. yarn.timeline-service.state-store-class org.apache.hadoop.yarn.server.timeline.recovery.LeveldbTimelineStateStore Store file name for leveldb state store. yarn.timeline-service.leveldb-state-store.path ${hadoop.tmp.dir}/yarn/timeline yarn.timeline-service.entity-group-fs-store.cache-store-class org.apache.hadoop.yarn.server.timeline.MemoryTimelineStore Caching storage timeline server v1.5 is using. yarn.timeline-service.entity-group-fs-store.active-dir /tmp/entity-file-history/active HDFS path to store active application’s timeline data yarn.timeline-service.entity-group-fs-store.done-dir /tmp/entity-file-history/done/ HDFS path to store done application’s timeline data yarn.timeline-service.entity-group-fs-store.group-id-plugin-classes Plugins that can translate a timeline entity read request into a list of timeline entity group ids, separated by commas. yarn.timeline-service.entity-group-fs-store.group-id-plugin-classpath Classpath for all plugins defined in yarn.timeline-service.entity-group-fs-store.group-id-plugin-classes. yarn.timeline-service.entity-group-fs-store.summary-store Summary storage for ATS v1.5 org.apache.hadoop.yarn.server.timeline.LeveldbTimelineStore yarn.timeline-service.entity-group-fs-store.scan-interval-seconds Scan interval for ATS v1.5 entity group file system storage reader.This value controls how frequent the reader will scan the HDFS active directory for application status. 60 yarn.timeline-service.entity-group-fs-store.cleaner-interval-seconds Scan interval for ATS v1.5 entity group file system storage cleaner.This value controls how frequent the reader will scan the HDFS done directory for stale application data. 3600 yarn.timeline-service.entity-group-fs-store.retain-seconds How long the ATS v1.5 entity group file system storage will keep an application's data in the done directory. 604800 yarn.timeline-service.entity-group-fs-store.leveldb-cache-read-cache-size Read cache size for the leveldb cache storage in ATS v1.5 plugin storage. 10485760 yarn.timeline-service.entity-group-fs-store.app-cache-size Size of the reader cache for ATS v1.5 reader. This value controls how many entity groups the ATS v1.5 server should cache. If the number of active read entity groups is greater than the number of caches items, some reads may return empty data. This value must be greater than 0. 10 yarn.timeline-service.client.fd-flush-interval-secs Flush interval for ATS v1.5 writer. This value controls how frequent the writer will flush the HDFS FSStream for the entity/domain. 10 yarn.timeline-service.client.fd-clean-interval-secs Scan interval for ATS v1.5 writer. This value controls how frequent the writer will scan the HDFS FSStream for the entity/domain. If the FSStream is stale for a long time, this FSStream will be close. 60 yarn.timeline-service.client.fd-retain-secs How long the ATS v1.5 writer will keep an FSStream open. If this fsstream does not write anything for this configured time, it will be close. 300 yarn.timeline-service.writer.class Storage implementation ATS v2 will use for the TimelineWriter service. org.apache.hadoop.yarn.server.timelineservice.storage.HBaseTimelineWriterImpl yarn.timeline-service.reader.class Storage implementation ATS v2 will use for the TimelineReader service. org.apache.hadoop.yarn.server.timelineservice.storage.HBaseTimelineReaderImpl yarn.timeline-service.client.internal-timers-ttl-secs How long the internal Timer Tasks can be alive in writer. If there is no write operation for this configured time, the internal timer tasks will be close. 420 The setting that controls how often the timeline collector flushes the timeline writer. yarn.timeline-service.writer.flush-interval-seconds 60 The setting that decides the capacity of the queue to hold asynchronous timeline entities. yarn.timeline-service.writer.async.queue.capacity 100 Time period till which the application collector will be alive in NM, after the application master container finishes. yarn.timeline-service.app-collector.linger-period.ms 60000 Time line V2 client tries to merge these many number of async entities (if available) and then call the REST ATS V2 API to submit. yarn.timeline-service.timeline-client.number-of-async-entities-to-merge 10 The setting that controls how long the final value of a metric of a completed app is retained before merging into the flow sum. Up to this time after an application is completed out-of-order values that arrive can be recognized and discarded at the cost of increased storage. yarn.timeline-service.hbase.coprocessor.app-final-value-retention-milliseconds 259200000 The setting that controls how often in-memory app level aggregation is kicked off in timeline collector. yarn.timeline-service.app-aggregation-interval-secs 15 The default hdfs location for flowrun coprocessor jar. yarn.timeline-service.hbase.coprocessor.jar.hdfs.location /hbase/coprocessor/hadoop-yarn-server-timelineservice.jar The value of this parameter sets the prefix for all tables that are part of timeline service in the hbase storage schema. It can be set to "dev." or "staging." if it is to be used for development or staging instances. This way the data in production tables stays in a separate set of tables prefixed by "prod.". yarn.timeline-service.hbase-schema.prefix prod. Optional URL to an hbase-site.xml configuration file to be used to connect to the timeline-service hbase cluster. If empty or not specified, then the HBase configuration will be loaded from the classpath. When specified the values in the specified configuration file will override those from the ones that are present on the classpath. yarn.timeline-service.hbase.configuration.file Removes the UUID if represent and limit the the flowname length with the given value for ATSv2. In case the value is negative or 0, it only removes the UUID and does not limit the flow name. yarn.timeline-service.flowname.max-size 0 Whether the shared cache is enabled yarn.sharedcache.enabled false The root directory for the shared cache yarn.sharedcache.root-dir /sharedcache The level of nested directories before getting to the checksum directories. It must be non-negative. yarn.sharedcache.nested-level 3 The implementation to be used for the SCM store yarn.sharedcache.store.class org.apache.hadoop.yarn.server.sharedcachemanager.store.InMemorySCMStore The implementation to be used for the SCM app-checker yarn.sharedcache.app-checker.class org.apache.hadoop.yarn.server.sharedcachemanager.RemoteAppChecker A resource in the in-memory store is considered stale if the time since the last reference exceeds the staleness period. This value is specified in minutes. yarn.sharedcache.store.in-memory.staleness-period-mins 10080 Initial delay before the in-memory store runs its first check to remove dead initial applications. Specified in minutes. yarn.sharedcache.store.in-memory.initial-delay-mins 10 The frequency at which the in-memory store checks to remove dead initial applications. Specified in minutes. yarn.sharedcache.store.in-memory.check-period-mins 720 The address of the admin interface in the SCM (shared cache manager) yarn.sharedcache.admin.address 0.0.0.0:8047 The number of threads used to handle SCM admin interface (1 by default) yarn.sharedcache.admin.thread-count 1 The address of the web application in the SCM (shared cache manager) yarn.sharedcache.webapp.address 0.0.0.0:8788 The frequency at which a cleaner task runs. Specified in minutes. yarn.sharedcache.cleaner.period-mins 1440 Initial delay before the first cleaner task is scheduled. Specified in minutes. yarn.sharedcache.cleaner.initial-delay-mins 10 The time to sleep between processing each shared cache resource. Specified in milliseconds. yarn.sharedcache.cleaner.resource-sleep-ms 0 The address of the node manager interface in the SCM (shared cache manager) yarn.sharedcache.uploader.server.address 0.0.0.0:8046 The number of threads used to handle shared cache manager requests from the node manager (50 by default) yarn.sharedcache.uploader.server.thread-count 50 The address of the client interface in the SCM (shared cache manager) yarn.sharedcache.client-server.address 0.0.0.0:8045 The number of threads used to handle shared cache manager requests from clients (50 by default) yarn.sharedcache.client-server.thread-count 50 The algorithm used to compute checksums of files (SHA-256 by default) yarn.sharedcache.checksum.algo.impl org.apache.hadoop.yarn.sharedcache.ChecksumSHA256Impl The replication factor for the node manager uploader for the shared cache (10 by default) yarn.sharedcache.nm.uploader.replication.factor 10 The number of threads used to upload files from a node manager instance (20 by default) yarn.sharedcache.nm.uploader.thread-count 20 ACL protocol for use in the Timeline server. security.applicationhistory.protocol.acl Set to true for MiniYARNCluster unit tests yarn.is.minicluster false Set for MiniYARNCluster unit tests to control resource monitoring yarn.minicluster.control-resource-monitoring false Set to false in order to allow MiniYARNCluster to run tests without port conflicts. yarn.minicluster.fixed.ports false Set to false in order to allow the NodeManager in MiniYARNCluster to use RPC to talk to the RM. yarn.minicluster.use-rpc false As yarn.nodemanager.resource.memory-mb property but for the NodeManager in a MiniYARNCluster. yarn.minicluster.yarn.nodemanager.resource.memory-mb 4096 Enable node labels feature yarn.node-labels.enabled false URI for NodeLabelManager. The default value is /tmp/hadoop-yarn-${user}/node-labels/ in the local filesystem. yarn.node-labels.fs-store.root-dir Set configuration type for node labels. Administrators can specify "centralized", "delegated-centralized" or "distributed". yarn.node-labels.configuration-type centralized When "yarn.node-labels.configuration-type" is configured with "distributed" in RM, Administrators can configure in NM the provider for the node labels by configuring this parameter. Administrators can configure "config", "script" or the class name of the provider. Configured class needs to extend org.apache.hadoop.yarn.server.nodemanager.nodelabels.NodeLabelsProvider. If "config" is configured, then "ConfigurationNodeLabelsProvider" and if "script" is configured, then "ScriptNodeLabelsProvider" will be used. yarn.nodemanager.node-labels.provider When "yarn.nodemanager.node-labels.provider" is configured with "config", "Script" or the configured class extends AbstractNodeLabelsProvider, then periodically node labels are retrieved from the node labels provider. This configuration is to define the interval period. If -1 is configured then node labels are retrieved from provider only during initialization. Defaults to 10 mins. yarn.nodemanager.node-labels.provider.fetch-interval-ms 600000 Interval at which NM syncs its node labels with RM. NM will send its loaded labels every x intervals configured, along with heartbeat to RM. yarn.nodemanager.node-labels.resync-interval-ms 120000 When "yarn.nodemanager.node-labels.provider" is configured with "config" then ConfigurationNodeLabelsProvider fetches the partition label from this parameter. yarn.nodemanager.node-labels.provider.configured-node-partition When "yarn.nodemanager.node-labels.provider" is configured with "Script" then this configuration provides the timeout period after which it will interrupt the script which queries the Node labels. Defaults to 20 mins. yarn.nodemanager.node-labels.provider.fetch-timeout-ms 1200000 When node labels "yarn.node-labels.configuration-type" is of type "delegated-centralized", administrators should configure the class for fetching node labels by ResourceManager. Configured class needs to extend org.apache.hadoop.yarn.server.resourcemanager.nodelabels. RMNodeLabelsMappingProvider. yarn.resourcemanager.node-labels.provider When "yarn.node-labels.configuration-type" is configured with "delegated-centralized", then node labels of all nodes are updated by periodically retrieving node labels from the provider. If -1 is configured then node labels are retrieved from provider only once for each node after it registers. Defaults to 30 mins. yarn.resourcemanager.node-labels.provider.fetch-interval-ms 1800000 When "yarn.node-labels.configuration-type" is configured with "delegated-centralized", then node labels of newly registered nodes are updated by periodically retrieving node labels from the provider. Defaults to 30 secs. yarn.resourcemanager.node-labels.provider.update-newly-registered-nodes-interval-ms 30000 Overwrites default-node-label-expression only for the ApplicationMaster container. It is disabled by default. yarn.resourcemanager.node-labels.am.default-node-label-expression Flag to indicate whether the AM can be allocated to non-exclusive nodes or not. Default is false. yarn.resourcemanager.node-labels.am.allow-non-exclusive-allocation false This property determines which provider will be plugged by the node manager to collect node-attributes. Administrators can configure "config", "script" or the class name of the provider. Configured class needs to extend org.apache.hadoop.yarn.server.nodemanager.nodelabels.NodeAttributesProvider. If "config" is configured, then "ConfigurationNodeLabelsProvider" and if "script" is configured, then "ScriptBasedNodeAttributesProvider" will be used. yarn.nodemanager.node-attributes.provider The node attribute script NM runs to collect node attributes. Script output Line starting with "NODE_ATTRIBUTE:" will be considered as a record of node attribute, attribute name, type and value should be delimited by comma. Each of such lines will be parsed to a node attribute. yarn.nodemanager.node-attributes.provider.script.path Command arguments passed to the node attribute script. yarn.nodemanager.node-attributes.provider.script.opts Time interval that determines how long NM fetches node attributes from a given provider. If -1 is configured then node labels are retrieved from provider only during initialization. Defaults to 10 mins. yarn.nodemanager.node-attributes.provider.fetch-interval-ms 600000 Timeout period after which NM will interrupt the node attribute provider script which queries node attributes. Defaults to 20 mins. yarn.nodemanager.node-attributes.provider.fetch-timeout-ms 1200000 When "yarn.nodemanager.node-attributes.provider" is configured with "config" then ConfigurationNodeAttributesProvider fetches node attributes from this parameter. yarn.nodemanager.node-attributes.provider.configured-node-attributes Interval at which NM syncs its node attributes with RM. NM will send its loaded attributes every x intervals configured, along with heartbeat to RM. yarn.nodemanager.node-attributes.resync-interval-ms 120000 Timeout in seconds for YARN node graceful decommission. This is the maximal time to wait for running containers and applications to complete before transition a DECOMMISSIONING node into DECOMMISSIONED. yarn.resourcemanager.nodemanager-graceful-decommission-timeout-secs 3600 Timeout in seconds of DecommissioningNodesWatcher internal polling. yarn.resourcemanager.decommissioning-nodes-watcher.poll-interval-secs 20 Used to specify custom web services for Resourcemanager. Value can be classnames separated by comma. Ex: org.apache.hadoop.yarn.server.resourcemanager.webapp.RMWebServices, org.apache.hadoop.yarn.server.resourcemanager.webapp.DummyClass yarn.http.rmwebapp.external.classes Used to specify custom scheduler page yarn.http.rmwebapp.scheduler.page.class Used to specify custom DAO classes used by custom web services. yarn.http.rmwebapp.custom.dao.classes Used to specify custom DAO classes used by custom web services which requires root unwrapping. yarn.http.rmwebapp.custom.unwrapped.dao.classes Used to specify custom WebServices class to bind with RMWebApp overriding the default RMWebServices. yarn.webapp.custom.webservice.class The Node Label script to run. Script output Line starting with "NODE_PARTITION:" will be considered as Node Label Partition. In case of multiple lines have this pattern, then last one will be considered yarn.nodemanager.node-labels.provider.script.path The arguments to pass to the Node label script. yarn.nodemanager.node-labels.provider.script.opts Flag to indicate whether the RM is participating in Federation or not. yarn.federation.enabled false Initial delay for federation state-store heartbeat service. Value is followed by a unit specifier: ns, us, ms, s, m, h, d for nanoseconds, microseconds, milliseconds, seconds, minutes, hours, days respectively. Values should provide units, but seconds are assumed yarn.federation.state-store.heartbeat.initial-delay 30s Machine list file to be loaded by the FederationSubCluster Resolver yarn.federation.machine-list Class name for SubClusterResolver yarn.federation.subcluster-resolver.class org.apache.hadoop.yarn.server.federation.resolver.DefaultSubClusterResolverImpl Store class name for federation state store yarn.federation.state-store.class org.apache.hadoop.yarn.server.federation.store.impl.MemoryFederationStateStore The time in seconds after which the federation state store local cache will be refreshed periodically yarn.federation.cache-ttl.secs 300 The registry base directory for federation. yarn.federation.registry.base-dir yarnfederation/ The number of retries to clear the app in the FederationStateStore, the default value is 1, that is, after the app fails to clean up, it will retry the cleanup again. yarn.federation.state-store.clean-up-retry-count 1 Clear the sleep time of App retry in FederationStateStore. When the app fails to clean up, it will sleep for a period of time and then try to clean up. The default value is 1s. yarn.federation.state-store.clean-up-retry-sleep-time 1s The registry implementation to use. yarn.registry.class org.apache.hadoop.registry.client.impl.FSRegistryOperationsService The interval that the yarn client library uses to poll the completion status of the asynchronous API of application client protocol. yarn.client.application-client-protocol.poll-interval-ms 200 The duration (in ms) the YARN client waits for an expected state change to occur. -1 means unlimited wait time. yarn.client.application-client-protocol.poll-timeout-ms -1 RSS usage of a process computed via /proc/pid/stat is not very accurate as it includes shared pages of a process. /proc/pid/smaps provides useful information like Private_Dirty, Private_Clean, Shared_Dirty, Shared_Clean which can be used for computing more accurate RSS. When this flag is enabled, RSS is computed as Min(Shared_Dirty, Pss) + Private_Clean + Private_Dirty. It excludes read-only shared mappings in RSS computation. yarn.nodemanager.container-monitor.procfs-tree.smaps-based-rss.enabled false URL for log aggregation server yarn.log.server.url URL for log aggregation server web service yarn.log.server.web-service.url RM Application Tracking URL yarn.tracking.url.generator Class to be used for YarnAuthorizationProvider yarn.authorization-provider Defines how often NMs wake up to upload log files. The default value is -1. By default, the logs will be uploaded when the application is finished. By setting this configuration logs can be uploaded periodically while the application is running. The minimum positive accepted value can be configured by the setting "yarn.nodemanager.log-aggregation.roll-monitoring-interval-seconds.min". yarn.nodemanager.log-aggregation.roll-monitoring-interval-seconds -1 Defines the positive minimum hard limit for "yarn.nodemanager.log-aggregation.roll-monitoring-interval-seconds". If this configuration has been set less than its default value (3600) the NodeManager may raise a warning. yarn.nodemanager.log-aggregation.roll-monitoring-interval-seconds.min 3600 Define how many aggregated log files per application per NM we can have in remote file system. By default, the total number of aggregated log files per application per NM is 30. yarn.nodemanager.log-aggregation.num-log-files-per-app 30 Enable/disable intermediate-data encryption at YARN level. For now, this only is used by the FileSystemRMStateStore to setup right file-system security attributes. yarn.intermediate-data-encryption.enable false Flag to enable cross-origin (CORS) support in the NM. This flag requires the CORS filter initializer to be added to the filter initializers list in core-site.xml. yarn.nodemanager.webapp.cross-origin.enabled false Defines maximum application priority in a cluster. If an application is submitted with a priority higher than this value, it will be reset to this maximum value. yarn.cluster.max-application-priority 0 The default log aggregation policy class. Applications can override it via LogAggregationContext. This configuration can provide some cluster-side default behavior so that if the application doesn't specify any policy via LogAggregationContext administrators of the cluster can adjust the policy globally. yarn.nodemanager.log-aggregation.policy.class org.apache.hadoop.yarn.server.nodemanager.containermanager.logaggregation.AllContainerLogAggregationPolicy The default parameters for the log aggregation policy. Applications can override it via LogAggregationContext. This configuration can provide some cluster-side default behavior so that if the application doesn't specify any policy via LogAggregationContext administrators of the cluster can adjust the policy globally. yarn.nodemanager.log-aggregation.policy.parameters Enable/Disable AMRMProxyService in the node manager. This service is used to intercept calls from the application masters to the resource manager. yarn.nodemanager.amrmproxy.enabled false The address of the AMRMProxyService listener. yarn.nodemanager.amrmproxy.address 0.0.0.0:8049 The number of threads used to handle requests by the AMRMProxyService. yarn.nodemanager.amrmproxy.client.thread-count 25 The comma separated list of class names that implement the RequestInterceptor interface. This is used by the AMRMProxyService to create the request processing pipeline for applications. yarn.nodemanager.amrmproxy.interceptor-class.pipeline org.apache.hadoop.yarn.server.nodemanager.amrmproxy.DefaultRequestInterceptor Whether AMRMProxy HA is enabled. yarn.nodemanager.amrmproxy.ha.enable false Setting that controls whether distributed scheduling is enabled. yarn.nodemanager.distributed-scheduling.enabled false Setting that controls whether opportunistic container allocation is enabled. yarn.resourcemanager.opportunistic-container-allocation.enabled false Maximum number of opportunistic containers to be allocated per Application Master heartbeat. yarn.resourcemanager.opportunistic.max.container-allocation.per.am.heartbeat -1 Number of nodes to be used by the Opportunistic Container Allocator for dispatching containers during container allocation. yarn.resourcemanager.opportunistic-container-allocation.nodes-used 10 Frequency for computing least loaded NMs. yarn.resourcemanager.nm-container-queuing.sorting-nodes-interval-ms 1000 Comparator for determining node load for Distributed Scheduling. yarn.resourcemanager.nm-container-queuing.load-comparator QUEUE_LENGTH Value of standard deviation used for calculation of queue limit thresholds. yarn.resourcemanager.nm-container-queuing.queue-limit-stdev 1.0f Min length of container queue at NodeManager. yarn.resourcemanager.nm-container-queuing.min-queue-length 5 Max length of container queue at NodeManager. yarn.resourcemanager.nm-container-queuing.max-queue-length 15 Min queue wait time for a container at a NodeManager. yarn.resourcemanager.nm-container-queuing.min-queue-wait-time-ms 10 Max queue wait time for a container queue at a NodeManager. yarn.resourcemanager.nm-container-queuing.max-queue-wait-time-ms 100 Use container pause as the preemption policy over kill in the container queue at a NodeManager. yarn.nodemanager.opportunistic-containers-use-pause-for-preemption false Error filename pattern, to identify the file in the container's Log directory which contain the container's error log. As error file redirection is done by client/AM and yarn will not be aware of the error file name. YARN uses this pattern to identify the error file and tail the error log as diagnostics when the container execution returns non zero value. Filename patterns are case sensitive and should match the specifications of FileSystem.globStatus(Path) api. If multiple filenames matches the pattern, first file matching the pattern will be picked. yarn.nodemanager.container.stderr.pattern {*stderr*,*STDERR*} Size of the container error file which needs to be tailed, in bytes. yarn.nodemanager.container.stderr.tail.bytes 4096 Choose different implementation of node label's storage yarn.node-labels.fs-store.impl.class org.apache.hadoop.yarn.nodelabels.FileSystemNodeLabelsStore The replication factor for the FS store files. Default value is 0, means it will use file system default replication. yarn.fs-store.file.replication 0 Enable the CSRF filter for the RM web app yarn.resourcemanager.webapp.rest-csrf.enabled false Optional parameter that indicates the custom header name to use for CSRF protection. yarn.resourcemanager.webapp.rest-csrf.custom-header X-XSRF-Header Optional parameter that indicates the list of HTTP methods that do not require CSRF protection yarn.resourcemanager.webapp.rest-csrf.methods-to-ignore GET,OPTIONS,HEAD Enable the CSRF filter for the NM web app yarn.nodemanager.webapp.rest-csrf.enabled false Optional parameter that indicates the custom header name to use for CSRF protection. yarn.nodemanager.webapp.rest-csrf.custom-header X-XSRF-Header Optional parameter that indicates the list of HTTP methods that do not require CSRF protection yarn.nodemanager.webapp.rest-csrf.methods-to-ignore GET,OPTIONS,HEAD The name of disk validator. yarn.nodemanager.disk-validator basic Enable the CSRF filter for the timeline service web app yarn.timeline-service.webapp.rest-csrf.enabled false Optional parameter that indicates the custom header name to use for CSRF protection. yarn.timeline-service.webapp.rest-csrf.custom-header X-XSRF-Header Optional parameter that indicates the list of HTTP methods that do not require CSRF protection yarn.timeline-service.webapp.rest-csrf.methods-to-ignore GET,OPTIONS,HEAD Enable the XFS filter for YARN yarn.webapp.xfs-filter.enabled true Property specifying the xframe options value. yarn.resourcemanager.webapp.xfs-filter.xframe-options SAMEORIGIN Property specifying the xframe options value. yarn.nodemanager.webapp.xfs-filter.xframe-options SAMEORIGIN Property specifying the xframe options value. yarn.timeline-service.webapp.xfs-filter.xframe-options SAMEORIGIN The least amount of time(msec.) an inactive (decommissioned or shutdown) node can stay in the nodes list of the resourcemanager after being declared untracked. A node is marked untracked if and only if it is absent from both include and exclude nodemanager lists on the RM. All inactive nodes are checked twice per timeout interval or every 10 minutes, whichever is lesser, and marked appropriately. The same is done when refreshNodes command (graceful or otherwise) is invoked. yarn.resourcemanager.node-removal-untracked.timeout-ms 60000 The RMAppLifetimeMonitor Service uses this value as monitor interval yarn.resourcemanager.application-timeouts.monitor.interval-ms 3000 Specifies what the RM does regarding HTTPS enforcement for communication with AM Web Servers, as well as generating and providing certificates. Possible values are: - NONE - the RM will do nothing special. - LENIENT - the RM will generate and provide a keystore and truststore to the AM, which it is free to use for HTTPS in its tracking URL web server. The RM proxy will still allow HTTP connections to AMs that opt not to use HTTPS. - STRICT - this is the same as LENIENT, except that the RM proxy will only allow HTTPS connections to AMs; HTTP connections will be blocked and result in a warning page to the user. yarn.resourcemanager.application-https.policy NONE Defines the limit of the diagnostics message of an application attempt, in kilo characters (character count * 1024). When using ZooKeeper to store application state behavior, its important to limit the size of the diagnostic messages to prevent YARN from overwhelming ZooKeeper. In cases where yarn.resourcemanager.state-store.max-completed-applications is set to a large number, it may be desirable to reduce the value of this property to limit the total data stored. yarn.app.attempt.diagnostics.limit.kc 64 Flag to enable cross-origin (CORS) support for timeline service v1.x or Timeline Reader in timeline service v2. For timeline service v2, also add org.apache.hadoop.security.HttpCrossOriginFilterInitializer to the configuration hadoop.http.filter.initializers in core-site.xml. yarn.timeline-service.http-cross-origin.enabled false The comma separated list of class names that implement the RequestInterceptor interface. This is used by the RouterClientRMService to create the request processing pipeline for users. yarn.router.clientrm.interceptor-class.pipeline org.apache.hadoop.yarn.server.router.clientrm.DefaultClientRequestInterceptor The thread pool executor size of per user in Router ClientRM Service FederationClientInterceptor. yarn.router.interceptor.user.threadpool-size 5 Size of LRU cache for Router ClientRM Service and RMAdmin Service. yarn.router.pipeline.cache-max-size 25 The comma separated list of class names that implement the RequestInterceptor interface. This is used by the RouterRMAdminService to create the request processing pipeline for users. yarn.router.rmadmin.interceptor-class.pipeline org.apache.hadoop.yarn.server.router.rmadmin.DefaultRMAdminRequestInterceptor The actual address the server will bind to. If this optional address is set, the RPC and webapp servers will bind to this address and the port specified in yarn.router.address and yarn.router.webapp.address, respectively. This is most useful for making Router listen to all interfaces by setting to 0.0.0.0. yarn.router.bind-host Comma-separated list of PlacementRules to determine how applications submitted by certain users get mapped to certain queues. Default is user-group, which corresponds to UserGroupMappingPlacementRule. yarn.scheduler.queue-placement-rules user-group The comma separated list of class names that implement the RequestInterceptor interface. This is used by the RouterWebServices to create the request processing pipeline for users. yarn.router.webapp.interceptor-class.pipeline org.apache.hadoop.yarn.server.router.webapp.DefaultRequestInterceptorREST The http address of the Router web application. If only a host is provided as the value, the webapp will be served on a random port. yarn.router.webapp.address 0.0.0.0:8089 The https address of the Router web application. If only a host is provided as the value, the webapp will be served on a random port. yarn.router.webapp.https.address 0.0.0.0:8091 It is TimelineClient 1.5 configuration whether to store active application’s timeline data with in user directory i.e ${yarn.timeline-service.entity-group-fs-store.active-dir}/${user.name} yarn.timeline-service.entity-group-fs-store.with-user-dir false yarn.resource-types The resource types to be used for scheduling. Use resource-types.xml to specify details about the individual resource types. yarn.webapp.filter-entity-list-by-user false Flag to enable display of applications per user as an admin configuration. yarn.webapp.filter-invalid-xml-chars false Flag to enable filter of invalid xml 1.0 characters present in the value of diagnostics field of apps output from RM WebService. The type of configuration store to use for scheduler configurations. Default is "file", which uses file based capacity-scheduler.xml to retrieve and change scheduler configuration. To enable API based scheduler configuration, use either "memory" (in memory storage, no persistence across restarts), "leveldb" (leveldb based storage), or "zk" (zookeeper based storage). API based configuration is only useful when using a scheduler which supports mutable configuration. Currently only capacity scheduler supports this. yarn.scheduler.configuration.store.class file The class to use for configuration mutation ACL policy if using a mutable configuration provider. Controls whether a mutation request is allowed. The DefaultConfigurationMutationACLPolicy checks if the requestor is a YARN admin. yarn.scheduler.configuration.mutation.acl-policy.class org.apache.hadoop.yarn.server.resourcemanager.scheduler.DefaultConfigurationMutationACLPolicy The storage path for LevelDB implementation of configuration store, when yarn.scheduler.configuration.store.class is configured to be "leveldb". yarn.scheduler.configuration.leveldb-store.path ${hadoop.tmp.dir}/yarn/system/confstore The compaction interval for LevelDB configuration store in secs, when yarn.scheduler.configuration.store.class is configured to be "leveldb". Default is one day. yarn.scheduler.configuration.leveldb-store.compaction-interval-secs 86400 The max number of configuration change log entries kept in config store, when yarn.scheduler.configuration.store.class is configured to be "leveldb" or "zk". Default is 1000 for either. yarn.scheduler.configuration.store.max-logs 1000 The file system directory to store the configuration files. The path can be any format as long as it follows hadoop compatible schema, for example value "file:///path/to/dir" means to store files on local file system, value "hdfs:///path/to/dir" means to store files on HDFS. If resource manager HA is enabled, recommended to use hdfs schema so it works in fail-over scenario. yarn.scheduler.configuration.fs.path file://${hadoop.tmp.dir}/yarn/system/schedconf The max number of configuration file in filesystem. Default is 100 for either. yarn.scheduler.configuration.max.version 100 ZK root node path for configuration store when using zookeeper-based configuration store. yarn.scheduler.configuration.zk-store.parent-path /confstore Provides an option for client to load supported resource types from RM instead of depending on local resource-types.xml file. yarn.client.load.resource-types.from-server false This setting controls if pluggable device framework is enabled. Disabled by default yarn.nodemanager.pluggable-device-framework.enabled false Configure vendor device plugin class name here. Comma separated. The class must be found in CLASSPATH. The pluggable device framework will load these classes. yarn.nodemanager.pluggable-device-framework.device-classes When yarn.nodemanager.resource.gpu.allowed-gpu-devices=auto specified, YARN NodeManager needs to run GPU discovery binary (now only support nvidia-smi) to get GPU-related information. When value is empty (default), YARN NodeManager will try to locate discovery executable itself. An example of the config value is: /usr/local/bin/nvidia-smi yarn.nodemanager.resource-plugins.gpu.path-to-discovery-executables Enable additional discovery/isolation of resources on the NodeManager, split by comma. By default, this is empty. Acceptable values: { "yarn.io/gpu", "yarn.io/fpga"}. yarn.nodemanager.resource-plugins Specifies whether the initialization of the Node Manager should continue if a certain device (GPU, FPGA, etc) was not found in the system. If set to "true", then an exception will be thrown if a device is missing or an error occurred during discovery. yarn.nodemanager.resource-plugins.fail-fast Specify GPU devices which can be managed by YARN NodeManager, split by comma Number of GPU devices will be reported to RM to make scheduling decisions. Set to auto (default) let YARN automatically discover GPU resource from system. Manually specify GPU devices if auto detect GPU device failed or admin only want subset of GPU devices managed by YARN. GPU device is identified by their minor device number and index. A common approach to get minor device number of GPUs is using "nvidia-smi -q" and search "Minor Number" output. When manual specify minor numbers, admin needs to include indice of GPUs as well, format is index:minor_number[,index:minor_number...]. An example of manual specification is "0:0,1:1,2:2,3:4" to allow YARN NodeManager to manage GPU devices with indice 0/1/2/3 and minor number 0/1/2/4. numbers . yarn.nodemanager.resource-plugins.gpu.allowed-gpu-devices auto Specify docker command plugin for GPU. By default uses Nvidia docker V1. yarn.nodemanager.resource-plugins.gpu.docker-plugin nvidia-docker-v1 Specify end point of nvidia-docker-plugin. Please find documentation: https://github.com/NVIDIA/nvidia-docker/wiki For more details. yarn.nodemanager.resource-plugins.gpu.docker-plugin.nvidia-docker-v1.endpoint http://localhost:3476/v1.0/docker/cli Specify one vendor plugin to handle FPGA devices discovery/IP download/configure. Only IntelFpgaOpenclPlugin is supported by default. We only allow one NM configured with one vendor FPGA plugin now since the end user can put the same vendor's cards in one host. And this also simplify our design. yarn.nodemanager.resource-plugins.fpga.vendor-plugin.class org.apache.hadoop.yarn.server.nodemanager.containermanager.resourceplugin.fpga.IntelFpgaOpenclPlugin When yarn.nodemanager.resource.fpga.allowed-fpga-devices=auto specified, YARN NodeManager needs to run FPGA discovery binary (now only support IntelFpgaOpenclPlugin) to get FPGA information. When value is empty (default), YARN NodeManager will try to locate discovery executable from vendor plugin's preference yarn.nodemanager.resource-plugins.fpga.path-to-discovery-executables Specify FPGA devices which can be managed by YARN NodeManager, split by comma Number of FPGA devices will be reported to RM to make scheduling decisions. Set to auto (default) let YARN automatically discover FPGA resource from system. Manually specify FPGA devices if admin only want subset of FPGA devices managed by YARN. At present, since we can only configure one major number in c-e.cfg, FPGA device is identified by their minor device number. A common approach to get minor device number of FPGA is using "aocl diagnose" and check uevent with device name. A sample manual value for this is like "0,1" yarn.nodemanager.resource-plugins.fpga.allowed-fpga-devices auto Absolute path to a script or executable that returns the available FPGA cards. The returned string must be a single line and follow the format: "deviceA/N:M,deviceB/X:Y". Example: "acl0/243:0,acl1/243:1". The numbers after the "/" character are the device major and minor numbers. When the script is enabled, auto-discovery is disabled the "aocl" command is not invoked to verify the available cards. yarn.nodemanager.resource-plugins.fpga.device-discovery-script List of FPGA available devices in the given node. The value must follow the format: "deviceA/N:M,deviceB/X:Y". Example: "acl0/243:0,acl1/243:1". The numbers after the "/" character are the device major and minor numbers. When this property is used, both auto-discovery and external script are ignored. yarn.nodemanager.resource-plugins.fpga.available-devices The http address of the timeline reader web application. yarn.timeline-service.reader.webapp.address ${yarn.timeline-service.webapp.address} The https address of the timeline reader web application. yarn.timeline-service.reader.webapp.https.address ${yarn.timeline-service.webapp.https.address} The actual address timeline reader will bind to. If this optional address is set, the reader server will bind to this address and the port specified in yarn.timeline-service.reader.webapp.address. This is most useful for making the service listen to all interfaces by setting to 0.0.0.0. yarn.timeline-service.reader.bind-host Whether to enable the NUMA awareness for containers in Node Manager. yarn.nodemanager.numa-awareness.enabled false Whether to read the NUMA topology from the system or from the configurations. If the value is true then NM reads the NUMA topology from system using the command 'numactl --hardware'. If the value is false then NM reads the topology from the configurations 'yarn.nodemanager.numa-awareness.node-ids'(for node id's), 'yarn.nodemanager.numa-awareness.<NODE_ID>.memory'(for each node memory), 'yarn.nodemanager.numa-awareness.<NODE_ID>.cpus'(for each node cpus). yarn.nodemanager.numa-awareness.read-topology false NUMA node id's in the form of comma separated list. Memory and No of CPUs will be read using the properties 'yarn.nodemanager.numa-awareness.<NODE_ID>.memory' and 'yarn.nodemanager.numa-awareness.<NODE_ID>.cpus' for each id specified in this value. This property value will be read only when 'yarn.nodemanager.numa-awareness.read-topology=false'. For example, if yarn.nodemanager.numa-awareness.node-ids=0,1 then need to specify memory and cpus for node id's '0' and '1' like below, yarn.nodemanager.numa-awareness.0.memory=73717 yarn.nodemanager.numa-awareness.0.cpus=4 yarn.nodemanager.numa-awareness.1.memory=73727 yarn.nodemanager.numa-awareness.1.cpus=4 yarn.nodemanager.numa-awareness.node-ids The numactl command path which controls NUMA policy for processes or shared memory. yarn.nodemanager.numa-awareness.numactl.cmd /usr/bin/numactl Enable elastic memory control. This is a Linux only feature. When enabled, the node manager adds a listener to receive an event, if all the containers exceeded a limit. The limit is specified by yarn.nodemanager.resource.memory-mb. If this is not set, the limit is set based on the capabilities. See yarn.nodemanager.resource.detect-hardware-capabilities for details. The limit applies to the physical or virtual (rss+swap) memory depending on whether yarn.nodemanager.pmem-check-enabled or yarn.nodemanager.vmem-check-enabled is set. yarn.nodemanager.elastic-memory-control.enabled false The name of a JVM class. The class must implement the Runnable interface. It is called, if yarn.nodemanager.elastic-memory-control.enabled is set and the system reaches its memory limit. When called the handler must preempt a container, since all containers are frozen by cgroups. Once preempted some memory is released, so that the kernel can resume all containers. Because of this the handler has to act quickly. yarn.nodemanager.elastic-memory-control.oom-handler org.apache.hadoop.yarn.server.nodemanager.containermanager.linux.resources.DefaultOOMHandler The path to the oom-listener tool. Elastic memory control is only supported on Linux. It relies on kernel events. The tool forwards these kernel events to the standard input, so that the node manager can preempt containers, in and out-of-memory scenario. You rarely need to update this setting. yarn.nodemanager.elastic-memory-control.oom-listener.path Maximum time to wait for an OOM situation to get resolved before bringing down the node. yarn.nodemanager.elastic-memory-control.timeout-sec 5 URI for NodeAttributeManager. The default value is /tmp/hadoop-yarn-${user}/node-attribute/ in the local filesystem. yarn.node-attribute.fs-store.root-dir Choose different implementation of node attribute's storage yarn.node-attribute.fs-store.impl.class org.apache.hadoop.yarn.server.resourcemanager.nodelabels.FileSystemNodeAttributeStore CSI driver adaptor addresses on a node manager. This configuration will be loaded by the resource manager to initiate a client for each adaptor in order to communicate with CSI drivers. Note, these addresses should be mapped to the adaptor addresses which runs the controller plugin. yarn.nodemanager.csi-driver-adaptor.addresses CSI driver names running on this node, multiple driver names need to be delimited by comma. The driver name should be same value returned by the getPluginInfo call.For each of the CSI driver name, it must to define following two corresponding properties: "yarn.nodemanager.csi-driver.${NAME}.endpoint" "yarn.nodemanager.csi-driver-adaptor.${NAME}.address" The 1st property defines where the driver's endpoint is; 2nd property defines where the mapping csi-driver-adaptor's address is. What's more, an optional csi-driver-adaptor class can be defined for each csi-driver: "yarn.nodemanager.csi-driver.${NAME}.class" once given, the adaptor will be initiated with the given class instead of the default implementation org.apache.hadoop.yarn.csi.adaptor.DefaultCsiAdaptorImpl. User can plug customized adaptor code for csi-driver with this configuration if necessary. yarn.nodemanager.csi-driver.names The cleanup interval for activities in milliseconds. yarn.resourcemanager.activities-manager.cleanup-interval-ms 5000 Time to live for scheduler activities in milliseconds. yarn.resourcemanager.activities-manager.scheduler-activities.ttl-ms 600000 Time to live for app activities in milliseconds. yarn.resourcemanager.activities-manager.app-activities.ttl-ms 600000 Max queue length for app activities. yarn.resourcemanager.activities-manager.app-activities.max-queue-length 100 Containers launcher implementation for determining how containers are launched within NodeManagers. yarn.nodemanager.containers-launcher.class org.apache.hadoop.yarn.server.nodemanager.containermanager.launcher.ContainersLauncher Enable the Pre processing of Application Submission context with server side configuration yarn.resourcemanager.submission-preprocessor.enabled false Path to file with hosts for the submission processor to handle. yarn.resourcemanager.submission-preprocessor.file-path Submission processor refresh interval yarn.resourcemanager.submission-preprocessor.file-refresh-interval-ms 60000 Comma-separated list of partitions. If a label P is in this list, then the RM will enforce that an app has resource requests with label P iff that app's node label expression is P. yarn.node-labels.exclusive-enforced-partitions Prefix used to identify the YARN tag which contains workflow ID. If a tag coming in application submission context has this prefix, whatever follows the prefix will be considered as workflow ID associated with the application. This configuration is used by features such as workflow priority for identifying the workflow associated with an application. yarn.workflow-id.tag-prefix workflowid: Whether or not to allow application submissions via REST. Default is true. yarn.webapp.enable-rest-app-submissions true The maximum number of application attempts. It's a global setting for all application masters. Each application master can specify its individual maximum number of application attempts via the API, but the individual number cannot be more than the global upper bound. If it is, the resourcemanager will override it. The default number value is set to yarn.resourcemanager.am.max-attempts. yarn.resourcemanager.am.global.max-attempts Max number of application tags set by user in ApplicationSubmissionContext while submitting application yarn.resourcemanager.application.max-tags 10 Max length of each application tag set by user in ApplicationSubmissionContext while submitting application. yarn.resourcemanager.application.max-tag.length 100 Specifies whether application tags should be converted to lowercase or not. yarn.resourcemanager.application-tag-based-placement.force-lowercase true Whether to enable RM to mark inactive nodes as untracked after the timeout specified by yarn.resourcemanager.node-removal-untracked.timeout-ms and then remove them from nodes list for the YARN cluster without configured include path, which means RM can periodically clear inactive nodes to avoid increasing memory to store these data when enabled, most desired by elastic cloud environment with frequent auto-scaling operations. It works only when the YARN cluster doesn't utilize include file, the key configurations are as follows: yarn.resourcemanager.nodes.exclude-path=/path-to-exclude-file yarn.resourcemanager.nodes.include-path= yarn.resourcemanager.node-removal-untracked.timeout-ms=60000 In this situation, the inactive nodes will never be marked as untracked and removed from the nodes list unless this configuration is enabled: yarn.resourcemanager.enable-node-untracked-without-include-path=true yarn.resourcemanager.enable-node-untracked-without-include-path false yarn.scheduler.app-placement-allocator.class In the absence of APPLICATION_PLACEMENT_TYPE_CLASS from the RM application scheduling environments, the value of this config is used to determine the default implementation of AppPlacementAllocator. If APPLICATION_PLACEMENT_TYPE_CLASS is absent from the application scheduling env and this config also has no value present, then default implementation LocalityAppPlacementAllocator is used. yarn.router.keytab.file The keytab file used by router to login as its service principal. The principal name is configured with dfs.federation.router.kerberos.principal. yarn.router.kerberos.principal The Router service principal. This is typically set to router/_HOST@REALM.TLD. Each Router will substitute _HOST with its own fully qualified hostname at startup. The _HOST placeholder allows using the same configuration setting on both Router setup. yarn.router.kerberos.principal.hostname Optional. The hostname for the Router containing this configuration file. Will be different for each machine. Defaults to current hostname. yarn.router.webapp.appsinfo-enabled false This configuration is used to enable the cache of AppsInfo. If it is set to true, the cache is enabled. If it is set to false, the cache is not enabled. yarn.router.webapp.appsinfo-cached-count 100 When yarn.router.appsinfo-enabled is set to true, the number of cached appsInfo. Default is 100 yarn.router.webapp.cross-origin.enabled false Flag to enable cross-origin (CORS) support for Yarn Router. For Yarn Router, also add org.apache.hadoop.security.HttpCrossOriginFilterInitializer to the configuration hadoop.http.filter.initializers in core-site.xml. yarn.federation.state-store.max-applications 1000 Yarn federation state-store supports querying the maximum number of apps. Default is 1000. yarn.router.interceptor.user-thread-pool.minimum-pool-size 5 This configurable is used to set the corePoolSize(minimumPoolSize) of the thread pool of the interceptor. Default is 5. yarn.router.interceptor.user-thread-pool.maximum-pool-size 5 This configuration is used to set the default value of maximumPoolSize of the thread pool of the interceptor. Default is 5. yarn.router.interceptor.user-thread-pool.keep-alive-time 0s This configurable is used to set the keepAliveTime of the thread pool of the interceptor. Default is 0s. yarn.router.submit.interval.time 10ms The interval Time between calling different subCluster requests. Default is 10ms. yarn.router.interceptor.allow-partial-result.enable false This configuration represents whether to allow the interceptor to return partial SubCluster results. If true, we will ignore the exception to some subClusters during the calling process, and return result. If false, if an exception occurs in a subCluster during the calling process, an exception will be thrown directly. Default is false. yarn.router.asc-interceptor-max-size 1MB We define the size limit of ApplicationSubmissionContext. If the size of the ApplicationSubmissionContext is larger than this value, We will throw an exception. the default value is 1MB. The number of threads to use for the Router scheduled executor service. yarn.router.subcluster.cleaner.interval.time 1 The interval at which the subClusterCleaner runs. Default is 60s. yarn.router.subcluster.cleaner.interval.time 60s SubCluster heartbeat timeout. Default is 30mins. yarn.router.subcluster.heartbeat.expiration.time 30m Whether to enable deregisterSubCluster. Default is true. yarn.router.deregister.subcluster.enabled true Number of Router Scheduler Threads. yarn.router.scheduled.executor.threads 1