- Refactored nifi-stateless to make use of nifi-framework-components
- Removed requirement for nifi-framework-nar to be provided.
- Refactored stateless nifi into api, engine, nar, and bootstrap modules, with a parent 'bundle' module
- Creation of nifi-stateless-system-tests
- Added unit tests and logging
- Changed flow configuration to use properties file instead of json
- Allow for -p parameter to specify parameters on command line
- Moved implementations of Authorizer, NiFiUser, and UserGroupProviders to new module named nifi-framework-authorization-providers so that those that depend on nifi-framework-authorization don't have to bring in the providers. This way, we can have stateless not bring in those providers, as we otherwise get warnings on startup about the provider already being registered. Additionally, it avoids needing dependencies on spring-security-core
- Updated bin/nifi.sh script to run new stateless bootstrap
- Added Reporting Tasks to stateless.
- Download bundles as necessary on stateless nifi startup
NIFI-7897: Addressing review feedback
NIFI-7897: Fixed typos in README and also addressed issue that caused parameters with spaces not to be parsed properly
This closes#4669.
Signed-off-by: Bryan Bende <bbende@apache.org>
- Make reporting in clustered scope to dependent of expected cluster state in order to prevent unexpected flow file emission
This closes#4642.
Signed-off-by: Mark Payne <markap14@hotmail.com>
NIFI-7549 Refining documentation; Changing explicit HA mode; Smaller review comments
NIFI-7549 Code review responses about license, documentation and dependencies
NIFI-7549 Fixing issue when explicit HA; Some further review based adjustments
NIFI-7549 Response to code review comments
NIFI-7549 Adding extra serialization test
NIFI-7549 Minor changes based on review comments
NIFI-7549 Adding hook point to the shutdown
This closes#4510.
Signed-off-by: Peter Turcsanyi <turcsanyi@apache.org>
* NIFI-7804 Split nifi-security-utils into sub-module for nifi-security-utils-api (no external dependencies).
Separated interface and implementation of TlsConfiguration.
Reabsorbed nifi-security-xml-config into nifi-security-utils.
* NIFI-7804 Resolved failing unit test on Java 8.
Removed accidental module dependency.
* NIFI-7804 Resolved failing unit test.
* NIFI-7804 Removed legacy dependency.
* NIFI-7804 Marked nifi-security-utils-api as provided and overrode with compile scope in specific modules which are not children of nifi-standard-services-api-nar.
Added unit tests.
Refactored shared logic from various algorithm-specific secure hasher implementations to AbstractSecureHasher.
Introduced secure hasher implementations for various KDFs.
Added custom validation to EncryptContent processor.
Implemented logic for EncryptContent to write operational metadata to flowfile attributes.
Added encryption metadata attribute annotations to EncryptContent.
Added Argon2 KDF documentation and Bcrypt key derivation change notes to Admin Guide.
Updated unit tests to calculate default/recommended cost parameters for Argon2.
This closes#4421.
Co-authored-by: mtien <mtien.apache@gmail.com>
- Updated InstanceClassLoader to resolve files that are in the instance urls or additional urls
- Updated nifi-mock to support KerberosContext and removeProperty for ControllerServices
- Added unit test for HadoopDBCPConnectionPool
- Addressing review feedback
This closes#4149.
Kerberos Password property should not support EL, this includes a change to KerberosProperties which is also used by the HDFS processors (AbstractHadoopProcessor)
Added wiring in a KerberosContext to a TestRunner's MockProcessorInitializationContext
Removed synchronization blocks around KerberosUser.checkTGTAndRelogin, since that method is already synchronized
Updated AbstractHadoopProcessor to have a boolean accessor method to determine if explicit keytab configuration is allowed
Removed synchronization block from HiveConnectionPool's getConnection method (in Hive, Hive_1_1, Hive3 modules), since new TGT ticket acquisition is handled by the KerberosUser implementation. If UGI is used to relogin, synchronization is handled internally by UGI.
Added Kerberos Principal and Kerberos Password properties to Hive, Hive_1_1, and Hive3 components
Hive, Hive_1_1, and Hive3 components now use KerberosUser implementations to authenticate with a KDC
Updated handling of the NIFI_ALLOW_EXPLICIT_KEYTAB environment variable in Hive and Hive3 components. An accessor method has been added that uses Boolean.parseBoolean, which returns true if the environment variable is set to true, and false otherwise (including when the environment variable is unset).
Addressing PR feedback
Addressing PR feedback
This closes#4102.
- Create a mock Validator to track count of calls to validate().
We cannot use Mockito for this, because it can't mock all the
StandardValidators
refactor based on review comments
fix naming in comments
moved to main based on review
Signed-off-by: Matthew Burgess <mattyb149@apache.org>
This closes#4043
* NIFI-6925: Fixed JoltTransformRecord for RecordReaders, improved MockProcessSession
* Fixed logic for no records, added unit test
* Fixed PutElasticsearchHttpRecord and PutHive3Streaming, same bug as JoltTransformRecord
* Added null checks
Updated description for "Full Path" filter mode to state that it will ignore scheme and authority
Added tests to TestListHDFS for listing an empty and nonexistent dirs
Updated TestListHDFS' mock file system to track state properly when FileStatus instances are added, and updated listStatus to work properly with the underlying Map that contains FileStatus instances
Updated ListHDFS' additional details to document "Full Path" filter mode ignoring scheme and authority, with an example
Updated TestRunners, StandardProcessorTestRunner, MockProcessorInitializationContext to support passing in a logger.
NIFI-6275 Updated the "Full Path" filter mode to check the full path of a file with and without its scheme and authority against the filter regex
Added additional documentation for how ListHDFS handles scheme and authority when "Full Path" filter mode is used
Added test case for "Full Path" filter mode with a regular expression that includes scheme and authority
This closes#3483.
Signed-off-by: Koji Kawamura <ijokarumawak@apache.org>
Added AzureStorageCredentialsService controller service interface
with 2 implementations:
- AzureStorageCredentialsControllerService: holds the credential properties
(account name, account key, sas token)
- AzureStorageCredentialsControllerServiceLookup: can be used to dynamically
look up another AzureStorageCredentialsService (similar to
DBCPConnectionPoolLookup)
The controller service can be used by the Azure Storage processors:
{List|Fetch|Put|Delete}AzureBlobStorage, {Get|Put}AzureQueueStorage
NIFI-6550: Review changes.
NIFI-6550: Review changes #2.
This closes#3742.
Signed-off-by: Koji Kawamura <ijokarumawak@apache.org>
NIFI-5816 Switching SFTP processors from JSCH to SSHJ
NIFI-5816 LICENSE/NOTICIE updates to reflect changing from JSch to SSHJ
Signed-off-by: Joe Witt <joewitt@apache.org>
<h1>Changes</h1>
- Instead of keeping a Set of referencing `ComponentNode`, keep instead a Map which stores, for each referencing component, also the name of the property for which the service is referenced
- This is done to keep the `addReference()` idempotent
- `StandardControllerService.referencingComponents` changed from HashSet to HashMap
- `Key` is the combination of `ComponentNode.hashCode()` and property name
- `Value` is the referencing `ComponentNode`
- `ControllerServiceNode.addReference()` now requires a tuple (referring `ComponentNode`, property name)
- `ControllerServiceNode.removeReference()` now requires a tuple (referring `ComponentNode`, property name)
- `ControllerServiceNode.getReferences()` signature is left untouched, when it's called, the values of the `referencingComponents` map are turned into a Set, meaning a referencing component will keep being returned by this method until at least one of its properties still reference the `ControllerService`
- `StandardSchedulingContext.leaseControllerService()` uses a dedicated constant
- `FrameworkIntegrationTest` has a test which create a processor with two optional controller service properties and then verifies that having them referencing and dereferencing the same controller service doesn't cause the reported bug anymore
Removed SchedulingContext from ClojureScriptEngineConfigurator
Co-Authored-By: Marco Gaido <marcogaido91@gmail.com>
Signed-off-by: Mark Payne <markap14@hotmail.com>
This closes#3600.
- Added nifi-parameter module
- Added Parameter to nifi-api
- Added ParameterContext and ParameterLookup
- Updated EL to support Parameters
- Updated backend so that any property can make use of Parameters
- Added web endpoint for Parameter Contexts
- Updated Templates to make use of Parameter Contexts
- Updated Versioned Flows to make use of Parameter Contexts
- Updated Stateless NiFi to support Parameters and take Parameters as part of the configuration, not Variables.
NIFI-6380: Addressed review feedback and fixed bugs; added additional unit and integration tests to verify
NIFI-6380: Added Description to Parameter Context
NIFI-6380: Fixed checkstyle violations
NIFI-6380: Fixed bug that caused updating a Parameter Context to fail if the name is provided and unchanged
NIFI-6380: If parameter is being deleted, don't worry about its sensitivity flag. This addresses a bug where the deletion of a Sensitive Parameter would require that the Parameter be submitted with a value of null and a sensitivity flag of true; else it would provide an error indicating that the parameter can't be changed from sensitive to non-sensitive. Now, the sensitivity flag is ignored.
NIFI-6380: Fixed bug around unsetting Process Group's Parameter Context
NIFI-6380: Moved lastRefreshed timestamp from ParameterContextDTO to ParameterContextsEntity and renamed to currentTime to match the pattern of ControllerServicesEntity rather than FlowHistory. Added parameterContextId to ProcessGroupFlowDTO
NIFI-6380: Added additional integration tests around escaped parameter references
NIFI-6380: Additional tests and bug fix for referencing EL from within another EL Expression
NIFI-6380: Created ParameterEntity to house a 'canWrite' flag for parameters and updated ParameterContext to use it. Updated ParameterContextUpdateRequestDTO to include a Set<AffectedComponentEntity> indicating the components that are affected by the update
NIFI-6380: Addressed review feedback
NIFI-6380: Addressed additional review feedback, mostly around code cleanup
NIFI-6380: Bug fix
NIFI-6380: Addressed more review feedback; fixed a couple of minor bugs encountered when testing
NIFI-6380: Bug fix around Parameter escaping for properties that support Expression Language. Fixed inconcsistency in Authorization hierarchy between /parameter-contexts/1234 and /parameter-contexts to ensure that the parent of /parameter-contexts is /controller, regardless of how we arrive at the /parameter-contexts resource
NIFI-6380: Fixed but around using ProcessContext#newPropertyValue(String) that previously resulted in causing the Parameters to be evaluated a second time, thereby ignoring escape characters. Updated Integration Test to verify behavior. Also fixed bug in LoadBalanceQueueIT as it was a one-liner and noticed it while running tests to verify other behavior
NIFI-6380: Ensure that if Processor or Controller Services references or de-references a Parameter that user making the change has READ policy on the Parameter Context
NIFI-6380: Verify permissions on Parameter Context when creating processor/controller service and when moving between groups, insantiating template, reverting/change flow version
NIFI-6380: Addressed problem that resulted from rebasing against master; fixed bug in showing the Affected Components for a particular Parameter Context Update
NIFI-6380: Fixed NPE that occurs when attempting to update controller-level Controller Service
NIFI-6380: Updated validation error text when referencing parameter from controller-level controller service or reporting task
NIFI-6380: Ensuring permissions for copy/paste when components reference parameters.
NIFI-6380: Do not require READ policy on Parameter Context in order to un-reference a parameter, only to reference one.
NIFI-6380: Restricted parameter names to match those in the Feature Proposal. Fixed bug that allowed copy & paste of a PG for which user does not have READ policy for a child/descendant PG's parameter context
This closes#3536
CSVReader and CSVRecordSetWriter controller services and also ConvertExcelToCSVProcessor
support EL for Value Separator, Quote Character and Escape Character properties.
NIFI-6318: Fixed null checks and compound OR expression.
NIFI-6318: RecordSetWriterFactory.createWriter() changes.
NIFI-6318: Initialize CSVFormat in onEnabled() if there are no dynamic formatting properties.
NIFI-6318: Comment Marker supports EL.
NIFI-6318: Various review changes.
This closes#3504.
Signed-off-by: Koji Kawamura <ijokarumawak@apache.org>
NIFI-6323 Changed URLs for splunk.artifactoryonline.com to use HTTPS (certificate validity warning in browsers, but command-line connection using openssl s_client is successful).
NIFI-6323 Changed URLs for XMLNS schema locations to use HTTPS (the XMLNS and schema identifier remain http:// because they are not designed to be resolvable).
NIFI-6323 Fixed Maven XML schema descriptor URLs.
This closes#3497
- Updates to make UpdateRecord and RecordPath automatically update Record schema when performing update and perform the updates on the first record in UpdateRecord before obtaining Writer Schema. This allows the Writer to to inherit the Schema of the updated Record instead of the Schema of the Record as it was when it was read.
- Updated JoltTransformRecord so that schema is inferred on the first transformed object before passing the schema to the Record Writer, so that if writer inherits schema from record, the schema that is inherited is the trans transformed schema
- Updated LookupRecord to allow for Record fields to be arbitrarily added
- Implemented ContentClaimInputStream
- Added controller service for caching schemas
- UpdatedQueryRecord to cache schemas automatically up to some number of schemas, which will significantly inprove throughput in many cases, especially with inferred schemas.
NIFI-5938: Updated AvroTypeUtil so that if creating an Avro Schema using a field name that is not valid for Avro, it creates a Schema that uses a different, valid field name and adds an alias for the given field name so that the fields still are looked up appropriately. Fixed a bug in finding the appropriate Avro field when aliases are used. Updated ContentClaimInputStream so that if mark() is called followed by multiple calls to reset(), that each reset() call is successful instead of failing after the first one (the JavaDoc for InputStream appears to indicate that the InputStream is free to do either and in fact the InputStream is even free to allow reset() to reset to the beginning of file if mark() is not even called, if it chooses to do so instead of requiring a call to mark()).
NIFI-5938: Added another unit test for AvroTypeUtil
NIFI-5938: If using inferred schema in CSV Reader, do not consider first record as a header line. Also addressed a bug in StandardConfigurationContext that was exposed by CSVReader, in which calling getProperty(PropertyDescriptor) did not properly lookup the canonical representation of the Property Descriptor from the component before attempting to get a default value
Signed-off-by: Matthew Burgess <mattyb149@apache.org>
This closes#3253
Deprecated HashContent processor.
Added documentation to HashAttribute processor.
Added shared HashService and HashAlgorithm enum.
Added unit tests.
Added #clearProperties() to TestRunner, StandardProcessorTestRunner, and MockProcessContext.
Updated processor manifest.
Updated Javadoc.
Added documentation about deprecated/renamed components.
Added logic handling for UTF-16 encoding where UTF-16BE is overridden so the BOM is not inserted prior to hashing.
This closes#2983.
Co-authored-by: Otto Fowler <ottobackwards@gmail.com>
Signed-off-by: Kevin Doran <kdoran@apache.org>
NIFI-950: Still seeing some slow response times when instantiating a large template in cluster mode so making some minor tweaks based on the results of CPU profiling
NIFI-5112: Refactored FlowSerializer so that it creates the desired intermediate data model that can be serialized, separate from serializing. This allows us to hold the FlowController's Read Lock only while creating the data model, not while actually serializing the data. Configured Jersey Client in ThreadPoolRequestReplicator not to look for features using the Service Loader for every request. Updated Template object to hold a DOM Node that represents the template contents instead of having to serialize the DTO, then parse the serialized form as a DOM object each time that it needs to be serialized.
NIFI-5112: Change ThreadPoolRequestReplicator to use OkHttp client instead of Jersey Client
NIFI-5111: Ensure that if a node is no longer cluster coordinator, that it clears any stale heartbeats.
NIFI-5110: Notify StandardProcessScheduler when a component is removed so that it will clean up any resource related to component lifecycle.
NIFI-950: Avoid gathering the Status objects for entire flow when we don't need them; removed unnecessary code
NIFI-950: Bug fixes
NIFI-950: Bug fix; added validation status to ProcessorDTO, ControllerServiceDTO, ReportingTaskDTO; updated DebugFlow to allow for pause time to be set in the customValidate method for testing functionality
NIFI-950: Addressing test failures
NIFI-950: Bug fixes
NIFI-950: Addressing review feedback
NIFI-950: Fixed validation logic in mock framework
This closes#2693
NIFI-5145 Fixed a mistake with evaluateAttributeExpressions found in a code review.
NIFI-5145 Removed a property descriptor from GetHBase that was a duplication of the one in VisibilityFetchSupport.
NIFI-5145 Added comments in the code that were requested in the review.
NIFI-5145 Fixed check style issue.
NIFI-5145 Fixed a few improperly scoped items in ListFileTransfer that impacted the FTP processor(s).
NIFI-5145 Changed which override gets called based on code review feedback.
NIFI-5145: Reverted changes to ListFileTransfer re ExpressionLanguageScope
NIFI-5145: Removed TODO from MockPropertyValue javadoc
Signed-off-by: Matthew Burgess <mattyb149@apache.org>
This closes#2672
- take into account input requirement for documentation rendering
- Renamed variable registry scope and added comments
- Doc + change in mock framework to check scope + update of components + UI
Implemented review feedback. Refactored data model to make the API cleaner and delineate more along the lines of what permissions are required in order to see which details
Implementing review feedback
Removed sensitive information from the diagnostics reports
Fixed bug in merging logic for GCDiagnosticsSnapshots
This closes#2468
This commit includes changes to DeleteHDFS to report REMOTE_INVOCATION
event. In order to do so, the processor had to be changed to create
output FlowFile because a provenance event needs a FlowFile it
associates with.
Signed-off-by: Pierre Villard <pierre.villard.fr@gmail.com>
This closes#2234.
NIFI-4061 Initial version of RedisStateProvider
- Adding PropertyContext and updating existing contexts to extend it
- Added embedded Redis for unit testing
- Added wrapped StateProvider with NAR ClassLoader in StandardStateManagerProvider
- Updating state-management.xml with config for RedisStateProvider
- Renaming tests that use RedisServer to be IT tests so they don't run all the time
This closes#1918.
Signed-off-by: Mark Payne <markap14@hotmail.com>
Signed-off-by: Matt Burgess <mattyb149@apache.org>
NIFI-1280: Updated SimpleKeyValueSchemaRegistry to make use of new CHOICE RecordFieldType - Update Record Readers to use SchemaRegistry controller service. Moved SchemaRegistry api into its own maven module and added to standard-services-api so that we can properly add dependencies on it. Code cleanup and bug fixes
Signed-off-by: Matt Burgess <mattyb149@apache.org>
NIFI-1280: Fixed checkstyle violations and license exclusions for RAT plugin
Signed-off-by: Matt Burgess <mattyb149@apache.org>
NIFI-1280: Addressed feedback from PR Review
Signed-off-by: Matt Burgess <mattyb149@apache.org>
NIFI-1280: Additional changes/doc to support QueryFlowFile and Record Readers/Writers
This closes#1652
Use it with FirstInFirstOutPrioritizer, it can enforce original ordering
of 'out-of-order' FlowFiles.
nifi-mock is modified to support FlowFile assertion using Prioritizer.
Signed-off-by: Matt Burgess <mattyb149@apache.org>
NIFI-3414: Added EnforceOrder processor
Incorporated review comments, added displayNames.
Signed-off-by: Matt Burgess <mattyb149@apache.org>
NIFI-3414: Added EnforceOrder processor
Incorporate review comments:
- Moved nifi-standard-prioritizers dependency to top level nifi/pom.xml.
- Changed default initial order from 1 to 0.
- Fixed typos.
- Use session.get(batchCount).
Signed-off-by: Matt Burgess <mattyb149@apache.org>
NIFI-3414: Added EnforceOrder processor
When a FlowFile is transferred to success, remove attributes previously set when it was transferred to wait or failure.
Signed-off-by: Matt Burgess <mattyb149@apache.org>
This closes#1496
- Refactoring NarDetails to include all info from MANIFEST
- Adding the concept of a Bundle and refactoring NarClassLoaders to pass Bundles to ExtensionManager
- Adding logic to fail start-up when multiple NARs with same coordinates exist, moving Bundle classes to framework API
- Refactoring bundle API to classes and creating BundleCoordinate
- Updating FlowController to use BundleCoordinate
- Updating the UI and DTO model to support showing bundle details that loaded an extension type.
- Adding bundle details for processor canvas node, processor dialogs, controller service dialogs, and reporting task dialogs.
- Updating the formating of the bundle coordinates.
- Addressing text overflow in the configuration/details dialog.
- Fixing self referencing functions.
- Updating extension UI mapping to incorporate bundle coordinates.
- Discovering custom UIs through the supplied bundles.
- Adding verification methods for creating extensions through the rest api.
- Only returning extensions that are common amongst all nodes.
- Rendering the ghost processors using a dotted border.
- Adding bundle details to the flow.xml.
- Loading NiFi build and version details from the framework NAR.
- Removing properties for build and version details.
- Wiring together front end and back end changes.
- Including bundle coordinates in the component data model.
- Wiring together component data model and flow.xml.
- Addressing issue when resolve unvesioned dependent NARs.
Updating unit tests to pass based on framework changes
- Fixing logging of extension types during start up
- Allowing the application to start if there is a compatible bundle found. - Reporting missing bundle when the a compatible bundle is not found. - Fixing table height in new component dialogs.
Fixing chechstyle error and increasing test timeout for TestStandardControllerServiceProvider
- Adding ability to change processor type at runtime
- Adding backend code to change type for controller services
- Cleaning up instance classloaders for temp components.
- Creating a dialog for changing the version of a component.
- Updating the formatting of the component type and bundle throughout.
- Updating the new component dialogs to support selecting source group.
- Cleaning up new component dialogs.
- Cleaning up documentation in the cluster node endpoint.
Adding missing include in nifi-web-ui pom compressor plugin
- Refactoring so ConfigurableComponent provides getLogger() and so the nodes provide the ConfigurableComponent
- Creating LoggableComponent to pass around the component, logger, and coordinate with in the framework
- Finishing clean up following rebase.
Calling lifecycle methods for add and remove when changing versions of a component
- Introducing verifyCanUpdateBundle(coordinate) to ConfiguredComponent, and adding unit tests
- Ensuring documentation is available for all components. Including those of the same type that are loaded from different bundles.
Adding lookup from ClassLoader to Bundle, adding fix for instance class loading to include all parent NARs, and adding additional unit tests for FlowController
- Adding validation to ensure referenced controller services implement the required API
- Fixing template instantiation to look up compatible bundle
- Requiring services/reporting tasks to be disabled/stopped.
- Only supporting a change version option when the item has multiple versions available.
- Limiting the possible new controller services to the applicable API version.
- Showing the implemented API versions for Controller Services.
- Updating the property descriptor tooltip to indicate the required service requirements.
- Introducing version based sorting in the new component dialog, change version dialog, and new controller service dialog.
- Addressing remainder of the issues from recent rebase.
Ensuring bundles have been added to the flow before proposing a flow, and incorporating bundle information into flow fingerprinting
- Refactoring the way missing bundles work to retain the desired bundle if available
- Fixing logger.isDebugEnabled to be logger.isTraceEnabled
- Auditing when user changes the bundle. - Ensuring bundle details are present in templates.
Moving standard prioritizers to framework NAR and refactoring ExtensionManager logic to handle cases where an extension is in a JAR directly in the lib directory
- Ensuring all nodes attempt to instantiate the same template instance when the available bundles may differ. - Fixing the auditing of copy/paste and template instantiation. - Running addtional verification methods when running standalone.
Refactoring controller service invocation handler to allow updating the node used by the invocation handler
- Ensuring the bundles in a proposed flow are compatible with the current instance when the current instance has no flow is going to accept the proposed flow
- Merging whether multiple versions of the component are available
- Setting NAR plugin back to current released version
- Cleaning up DocGenerator to not process multiple times
Addressing incorrect usage of nf.Common. - Using formatType in the new component type dialogs.
Improving error messages when looking for bundles
Addressing comments from PR. - Fixing references to global nf namespace. - Fixing injection of nfProcessGroupConfiguration in nfComponentVersion. - Fixing web api integration tests.
Not rendering unversioned in help documentation. - Ensuring the isExtentionMissing flag is correct after changing the component type.
Adding synchronization in node classes to ensure changing component can't occur when component is running, introducing MissingBundleException for better reporting when a node can't join cluster due to a missing bundle, and bumping NAR plugin to released version 1.2.0
Adding concept of missing components to fingerprinting to ensure nodes agree on missing components when joining a cluster
NIFI-3380: NIFI-3520: - Fixing hive nar dependency. - Marking DBCPService as provided. - Skipping services that require instance classloading and are cobundled with their service API. - Skipping components that require instance classloading and reference service APIs that are cobundled. - Addressing UI issues in the new component dialogs when re-opening with a filter applied.
Fixing checkstyles issue and adding back assume checks to distributed cache server test
Ensuring new component types are sorted correctly when shown initially.
This closes#1585.
- The idea behind NIFI-3356 was to improve the efficiency and throughput of the Provenance Repository, as it is often the bottleneck. While testing the newly designed repository,
a handful of other, fairly minor, changes were made to improve efficiency as well, as these came to light when testing the new repository:
- Use a BufferedOutputStream within StandardProcessSession (via a ClaimCache abstraction) in order to avoid continually writing to FileOutputStream when writing many small FlowFiles
- Updated threading model of MinimalLockingWriteAheadLog - now performs serialization outside of lock and writes to a 'synchronized' OutputStream
- Change minimum scheduling period for components from 30 microseconds to 1 nanosecond. ScheduledExecutor is very inconsistent with timing of task scheduling. With the bored.yield.duration
now present, this value doesn't need to be set to 30 microseconds. This was originally done to avoid processors that had no work from dominating the CPU. However, now that we will yield
when processors have no work, this results in slowing down processors that are able to perform work.
- Allow nifi.properties to specify multiple directories for FlowFile Repository
- If backpressure is engaged while running a batch of sessions, then stop batch processing earlier. This helps FlowFiles to move through the system much more smoothly instead of the
herky-jerky queuing that we previously saw at very high rates of FlowFiles.
- Added NiFi PID to log message when starting nifi. This was simply an update to the log message that provides helpful information.
NIFI-3356: Fixed bug in ContentClaimWriteCache that resulted in data corruption and fixed bug in RepositoryConfiguration that threw exception if cache warm duration was set to empty string
NIFI-3356: Fixed NPE
NIFI-3356: Added debug-level performance monitoring
NIFI-3356: Updates to unit tests that failed after rebasing against master
NIFI-3356: Incorporated PR review feedback
NIFI-3356: Fixed bug where we would delete index directories that are still in use; also added additional debug logging and a simple util class that can be used to textualize provenance event files - useful in debugging
This closes#1493
Joe Witt
Peter Turcsanyi
Mark Payne
Bence Simon
Matthew Burgess
Andy LoPresto
Andy LoPresto
Niels Basjes
KovalevIV
Bryan Bende
Pierre Villard
jstorck
karthik-kadajji
Otto Fowler
Alessandro D'Armiento
Peter Turcsanyi
Peter Wicks
Alessandro D'Armiento
SavtechSolutions
Martin Payne
Mark Bean
Mike Thomsen
jlwt90
Koji Kawamura
Omer Hadari
m-hogue
Tony Kurc
Matt Gilman
Oleg Zhurakousky