[
https://issues.apache.org/jira/browse/FLINK-2354?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14906491#comment-14906491
]
ASF GitHub Bot commented on FLINK-2354:
---------------------------------------
Github user uce commented on the pull request:
https://github.com/apache/flink/pull/1153#issuecomment-142962514
Currently, ZooKeeper only stores the latest added job graph (existing ones
are overwritten), whereas in the JobManager the JobGraph is attached to the
existing ExecutionGraph. In case of recovery, only the latest "attached"
JobGraph will be recovered.
> Recover running jobs on JobManager failure
> ------------------------------------------
>
> Key: FLINK-2354
> URL: https://issues.apache.org/jira/browse/FLINK-2354
> Project: Flink
> Issue Type: Sub-task
> Components: JobManager
> Affects Versions: master
> Reporter: Ufuk Celebi
> Assignee: Ufuk Celebi
> Fix For: 0.10
>
>
> tl;dr Persist JobGraphs in state backend and coordinate reference to state
> handle via ZooKeeper.
> Problem: When running multiple JobManagers in high availability mode, the
> leading job manager looses all running jobs when it fails. After a new
> leading job manager is elected, it is not possible to recover any previously
> running jobs.
> Solution: The leading job manager, which receives the job graph writes 1) the
> job graph to a state backend, and 2) a reference to the respective state
> handle to ZooKeeper. In general, job graphs can become large (multiple MBs,
> because they include closures etc.). ZooKeeper is not designed for data of
> this size. The level of indirection via the reference to the state backend
> keeps the data in ZooKeeper small.
> Proposed ZooKeeper layout:
> /flink (default)
> +- currentJobs
> +- job id i
> +- state handle reference of job graph i
> The 'currentJobs' node needs to be persistent to allow recovery of jobs
> between job managers. The currentJobs node needs to satisfy the following
> invariant: There is a reference to a job graph with id i IFF the respective
> job graph needs to be recovered by a newly elected job manager leader.
> With this in place, jobs will be recovered from their initial state (as if
> resubmitted). The next step is to backup the runtime state handles of
> checkpoints in a similar manner.
> ---
> This work will be based on [~trohrm...@apache.org]'s implementation of
> FLINK-2291. The leader election service notifies the job manager about
> granted/revoked leadership. This notification happens via Akka and thus
> serially *per* job manager, but results in eventually consistent state
> between job managers. For some snapshots of time it is possible to have a new
> leader granted leadership, before the old one has been revoked its leadership.
> [~trohrm...@apache.org], can you confirm that leadership does not guarantee
> mutually exclusive access to the shared 'currentJobs' state?
> For example, the following can happen:
> - JM 1 is leader, JM 2 is standby
> - JOB i is running (and hence /flink/currentJobs/i exists)
> - ZK notifies leader election service (LES) of JM 1 and JM 2
> - LES 2 immediately notifies JM 2 about granted leadership, but LES 1
> notification revoking leadership takes longer
> - JOB i finishes (TMs don't notice leadership change yet) and JM 1 receives
> final JobStatusChange
> - JM 2 resubmits the job /flink/currentJobs/i
> - JM 1 removes /flink/currentJobs/i, because it is now finished
> => inconsistent state (wrt the specified invariant above)
> If it is indeed a problem, we can circumvent this with a Curator recipe for
> [shared locks|http://curator.apache.org/curator-recipes/shared-lock.html] to
> coordinate the access to currentJobs. The lock needs to be acquired on
> leadership.
> ---
> Minimum required tests:
> - Unit tests for job graph serialization and writing to state backend and
> ZooKeeper with expected nodes
> - Unit tests for job submission to job manager in leader/non-leader state
> - Unit tests for leadership granting/revoking and job submission/restarting
> interleavings
> - Process failure integration tests with single and multiple running jobs
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
