Apache/hadoop
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

yarn.resourcemanager.fail-fast is used inconsistently. Contributed by Yuanbo Liu.

Browse Source 
(cherry picked from commit d9ba6f3656e8dc97d2813181e27d12e52dca4328)
(cherry picked from commit 3d6ba2dd4e4f1c42522c837ca96ee5d13f1491a4)
This commit is contained in:
Junping Du 2018-07-03 14:46:44 +08:00
parent 3042aa2685
commit 0a6942d58c
4 changed files with 24 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/conf/capacity-scheduler.xml
View File

@ -207,4 +207,14 @@
</description>
</property>
<property>
<name>yarn.scheduler.capacity.application.fail-fast</name>
<value>false</value>
<description>
Whether RM should fail during recovery if previous applications'
queue is no longer valid.
</description>
</property>
</configuration>

4
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/CapacityScheduler.java
View File

@ -727,7 +727,7 @@ private void addApplicationOnRecovery(
if (queue == null) {
//During a restart, this indicates a queue was removed, which is
//not presently supported
if (!YarnConfiguration.shouldRMFailFast(getConfig())) {
if (!getConfiguration().shouldAppFailFast(getConfig())) {
this.rmContext.getDispatcher().getEventHandler().handle(
new RMAppEvent(applicationId, RMAppEventType.KILL,
"Application killed on recovery as it was submitted to queue "
@ -747,7 +747,7 @@ private void addApplicationOnRecovery(
if (!(queue instanceof LeafQueue)) {
// During RM restart, this means leaf queue was converted to a parent
// queue, which is not supported for running apps.
if (!YarnConfiguration.shouldRMFailFast(getConfig())) {
if (!getConfiguration().shouldAppFailFast(getConfig())) {
this.rmContext.getDispatcher().getEventHandler().handle(
new RMAppEvent(applicationId, RMAppEventType.KILL,
"Application killed on recovery as it was submitted to queue "

10
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/capacity/CapacitySchedulerConfiguration.java
View File

@ -243,6 +243,12 @@ public class CapacitySchedulerConfiguration extends ReservationSchedulerConfigur
public static final String SCHEDULE_ASYNCHRONOUSLY_MAXIMUM_PENDING_BACKLOGS =
SCHEDULE_ASYNCHRONOUSLY_PREFIX + ".maximum-pending-backlogs";
@Private
public static final String APP_FAIL_FAST = PREFIX + "application.fail-fast";
@Private
public static final boolean DEFAULT_APP_FAIL_FAST = false;
@Private
public static final Integer
DEFAULT_SCHEDULE_ASYNCHRONOUSLY_MAXIMUM_PENDING_BACKLOGS = 100;
@ -1193,6 +1199,10 @@ public boolean getLazyPreemptionEnabled() {
return getBoolean(LAZY_PREEMPTION_ENABLED, DEFAULT_LAZY_PREEMPTION_ENABLED);
}
public boolean shouldAppFailFast(Configuration conf) {
return conf.getBoolean(APP_FAIL_FAST, DEFAULT_APP_FAIL_FAST);
}
private static final String PREEMPTION_CONFIG_PREFIX =
"yarn.resourcemanager.monitor.capacity.preemption.";

2
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/test/java/org/apache/hadoop/yarn/server/resourcemanager/TestWorkPreservingRMRestart.java
View File

@ -741,6 +741,7 @@ private void verifyAppRecoveryWithWrongQueueConfig(
MockMemoryRMStateStore memStore, RMState state) throws Exception {
// Restart RM with fail-fast as false. App should be killed.
csConf.setBoolean(YarnConfiguration.RM_FAIL_FAST, false);
csConf.setBoolean(CapacitySchedulerConfiguration.APP_FAIL_FAST, false);
rm2 = new MockRM(csConf, memStore);
rm2.start();
@ -775,6 +776,7 @@ private void verifyAppRecoveryWithWrongQueueConfig(
// Now restart RM with fail-fast as true. QueueException should be thrown.
csConf.setBoolean(YarnConfiguration.RM_FAIL_FAST, true);
csConf.setBoolean(CapacitySchedulerConfiguration.APP_FAIL_FAST, true);
MockRM rm = new MockRM(csConf, memStore2);
try {
rm.start();