org.apache.flink.runtime.highavailability.zookeeper

Class ZooKeeperHaServices

java.lang.Object

org.apache.flink.runtime.highavailability.AbstractHaServices

org.apache.flink.runtime.highavailability.zookeeper.ZooKeeperHaServices

All Implemented Interfaces:

AutoCloseable, ClientHighAvailabilityServices, HighAvailabilityServices

public class ZooKeeperHaServices
extends AbstractHaServices

An implementation of the AbstractHaServices using Apache ZooKeeper. The services store data in ZooKeeper's nodes as illustrated by the following tree structure:

 /flink
      +/cluster_id_1/leader/resource_manager/latch
      |            |                        /connection_info
      |            |       /dispatcher/latch
      |            |                  /connection_info
      |            |       /rest_server/latch
      |            |                   /connection_info
      |            |
      |            |
      |            +jobgraphs/job-id-1
      |            |         /job-id-2
      |            +jobs/job-id-1/leader/latch
      |                 |               /connection_info
      |                 |        /checkpoints/latest
      |                 |                    /latest-1
      |                 |                    /latest-2
      |                 |       /checkpoint_id_counter
      |
      +/cluster_id_2/leader/resource_manager/latch
      |            |                        /connection_info
      |            |       /dispatcher/latch
      |            |                  /connection_info
      |            |       /rest_server/latch
      |            |                   /connection_info
      |            |
      |            +jobgraphs/job-id-2
      |            +jobs/job-id-2/leader/latch
      |                 |               /connection_info
      |                 |        /checkpoints/latest
      |                 |                    /latest-1
      |                 |                    /latest-2
      |                 |       /checkpoint_id_counter
 

The root path "/flink" is configurable via the option HighAvailabilityOptions.HA_ZOOKEEPER_ROOT. This makes sure Flink stores its data under specific subtrees in ZooKeeper, for example to accommodate specific permission.

The "cluster_id" part identifies the data stored for a specific Flink "cluster". This "cluster" can be either a standalone or containerized Flink cluster, or it can be job on a framework like YARN (in a "per-job-cluster" mode).

In case of a "per-job-cluster" on YARN, the cluster-id is generated and configured automatically by the client or dispatcher that submits the Job to YARN.

In the case of a standalone cluster, that cluster-id needs to be configured via HighAvailabilityOptions.HA_CLUSTER_ID. All nodes with the same cluster id will join the same cluster and participate in the execution of the same set of jobs.

Field Summary

Fields inherited from class org.apache.flink.runtime.highavailability.AbstractHaServices

configuration, ioExecutor, logger

Fields inherited from interface org.apache.flink.runtime.highavailability.HighAvailabilityServices

DEFAULT_JOB_ID, DEFAULT_LEADER_ID

Constructor Summary

Constructors

Constructor and Description
ZooKeeperHaServices(CuratorFrameworkWithUnhandledErrorListener curatorFrameworkWrapper, Executor executor, org.apache.flink.configuration.Configuration configuration, BlobStoreService blobStoreService)

Method Summary

Modifier and Type	Method and Description
CheckpointRecoveryFactory	createCheckpointRecoveryFactory() Create the checkpoint recovery factory for the job manager.
JobGraphStore	createJobGraphStore() Create the submitted job graph store for the job manager.
protected LeaderElectionService	createLeaderElectionService(String leaderPath) Create leader election service with specified leaderName.
protected LeaderRetrievalService	createLeaderRetrievalService(String leaderPath) Create leader retrieval service with specified leaderName.
RunningJobsRegistry	createRunningJobsRegistry() Create the registry that holds information about whether jobs are currently running.
protected String	getLeaderPathForDispatcher() Get the leader path for Dispatcher.
String	getLeaderPathForJobManager(org.apache.flink.api.common.JobID jobID) Get the leader path for specific JobManager.
protected String	getLeaderPathForResourceManager() Get the leader path for ResourceManager.
protected String	getLeaderPathForRestServer() Get the leader path for RestServer.
void	internalCleanup() Clean up the meta data in the distributed system(e.g.
void	internalCleanupJobData(org.apache.flink.api.common.JobID jobID) Clean up the meta data in the distributed system(e.g.
void	internalClose() Closes the components which is used for external operations(e.g.

Methods inherited from class org.apache.flink.runtime.highavailability.AbstractHaServices

cleanupJobData, close, closeAndCleanupAllData, createBlobStore, getCheckpointRecoveryFactory, getClusterRestEndpointLeaderElectionService, getClusterRestEndpointLeaderRetriever, getDispatcherLeaderElectionService, getDispatcherLeaderRetriever, getJobGraphStore, getJobManagerLeaderElectionService, getJobManagerLeaderRetriever, getJobManagerLeaderRetriever, getResourceManagerLeaderElectionService, getResourceManagerLeaderRetriever, getRunningJobsRegistry

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface org.apache.flink.runtime.highavailability.HighAvailabilityServices

getWebMonitorLeaderElectionService, getWebMonitorLeaderRetriever

Constructor Detail

ZooKeeperHaServices

public ZooKeeperHaServices(CuratorFrameworkWithUnhandledErrorListener curatorFrameworkWrapper,
                           Executor executor,
                           org.apache.flink.configuration.Configuration configuration,
                           BlobStoreService blobStoreService)

Method Detail

createCheckpointRecoveryFactory

public CheckpointRecoveryFactory createCheckpointRecoveryFactory()
                                                          throws Exception

Description copied from class: AbstractHaServices

Create the checkpoint recovery factory for the job manager.

Specified by:

createCheckpointRecoveryFactory in class AbstractHaServices

Returns:

Checkpoint recovery factory

Throws:

Exception

createJobGraphStore

public JobGraphStore createJobGraphStore()
                                  throws Exception

Description copied from class: AbstractHaServices

Create the submitted job graph store for the job manager.

Specified by:

createJobGraphStore in class AbstractHaServices

Returns:

Submitted job graph store

Throws:

Exception - if the submitted job graph store could not be created

createRunningJobsRegistry

public RunningJobsRegistry createRunningJobsRegistry()

Description copied from class: AbstractHaServices

Create the registry that holds information about whether jobs are currently running.

Specified by:

createRunningJobsRegistry in class AbstractHaServices

Returns:

Running job registry to retrieve running jobs

createLeaderElectionService

protected LeaderElectionService createLeaderElectionService(String leaderPath)

Description copied from class: AbstractHaServices

Create leader election service with specified leaderName.

Specified by:

createLeaderElectionService in class AbstractHaServices

Parameters:

leaderPath - ConfigMap name in Kubernetes or child node path in Zookeeper.

Returns:

Return LeaderElectionService using Zookeeper or Kubernetes.

createLeaderRetrievalService

protected LeaderRetrievalService createLeaderRetrievalService(String leaderPath)

Description copied from class: AbstractHaServices

Create leader retrieval service with specified leaderName.

Specified by:

createLeaderRetrievalService in class AbstractHaServices

Parameters:

leaderPath - ConfigMap name in Kubernetes or child node path in Zookeeper.

Returns:

Return LeaderRetrievalService using Zookeeper or Kubernetes.

internalClose

public void internalClose()

Description copied from class: AbstractHaServices

Closes the components which is used for external operations(e.g. Zookeeper Client, Kubernetes Client).

Specified by:

internalClose in class AbstractHaServices

internalCleanup

public void internalCleanup()
                     throws Exception

Description copied from class: AbstractHaServices

Clean up the meta data in the distributed system(e.g. Zookeeper, Kubernetes ConfigMap).

If an exception occurs during internal cleanup, we will continue the cleanup in AbstractHaServices.closeAndCleanupAllData() and report exceptions only after all cleanup steps have been attempted.

Specified by:

internalCleanup in class AbstractHaServices

Throws:

Exception - when do the cleanup operation on external storage.

internalCleanupJobData

public void internalCleanupJobData(org.apache.flink.api.common.JobID jobID)
                            throws Exception

Description copied from class: AbstractHaServices

Clean up the meta data in the distributed system(e.g. Zookeeper, Kubernetes ConfigMap) for the specified Job.

Specified by:

internalCleanupJobData in class AbstractHaServices

Parameters:

jobID - The identifier of the job to cleanup.

Throws:

Exception - when do the cleanup operation on external storage.

getLeaderPathForResourceManager

protected String getLeaderPathForResourceManager()

Description copied from class: AbstractHaServices

Get the leader path for ResourceManager.

Specified by:

getLeaderPathForResourceManager in class AbstractHaServices

Returns:

Return the ResourceManager leader name. It is ConfigMap name in Kubernetes or child node path in Zookeeper.

getLeaderPathForDispatcher

protected String getLeaderPathForDispatcher()

Description copied from class: AbstractHaServices

Get the leader path for Dispatcher.

Specified by:

getLeaderPathForDispatcher in class AbstractHaServices

Returns:

Return the Dispatcher leader name. It is ConfigMap name in Kubernetes or child node path in Zookeeper.

getLeaderPathForJobManager

public String getLeaderPathForJobManager(org.apache.flink.api.common.JobID jobID)

Description copied from class: AbstractHaServices

Get the leader path for specific JobManager.

Specified by:

getLeaderPathForJobManager in class AbstractHaServices

Parameters:

jobID - job id

Returns:

Return the JobManager leader name for specified job id. It is ConfigMap name in Kubernetes or child node path in Zookeeper.

getLeaderPathForRestServer

protected String getLeaderPathForRestServer()

Description copied from class: AbstractHaServices

Get the leader path for RestServer.

Specified by:

getLeaderPathForRestServer in class AbstractHaServices

Returns:

Return the RestServer leader name. It is ConfigMap name in Kubernetes or child node path in Zookeeper.