[
https://issues.apache.org/jira/browse/FLINK-8453?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16330684#comment-16330684
]
ASF GitHub Bot commented on FLINK-8453:
---------------------------------------
GitHub user tillrohrmann opened a pull request:
https://github.com/apache/flink/pull/5310
[FLINK-8453] [flip6] Add SerializableExecutionGraphStore to Dispatcher
## What is the purpose of the change
The SerializableExecutionGraphStore is responsible for storing completed
jobs
for historic job requests (e.g. from the web ui or from the client). The
store
is populated by the Dispatcher once a job has terminated.
The FileSerializableExecutionGraphStore implementation persists all
SerializableExecutionGraphs on disk in order to avoid OOM problems. It only
keeps
some of the stored graphs in memory until it reaches a configurable size.
Once
coming close to this size, it will evict the elements and only reload them
if
requested again. Additionally, the FileSerializableExecutionGraphStore
defines
an expiration time after which the execution graphs will be removed from
disk.
This prevents excessive use of disk resources.
This PR is based on #5309.
## Brief change log
- Introduce `SerializableExecutionGraphStore` and
`FileSerializableExecutionGraphStore`
- Add `FileSerializableExecutionGraphStore` to `Dispatcher`
- Store `SerializableExecutionGraphs` in corresponding
`FileSerializableExecutionGraphStore`
- Adapt `Dispatcher` to serve requests for historic jobs
## Verifying this change
- Added `FileSerializableExecutionGraphStoreTest`
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): (no)
- The public API, i.e., is any changed class annotated with
`@Public(Evolving)`: (no)
- The serializers: (no)
- The runtime per-record code paths (performance sensitive): (no)
- Anything that affects deployment or recovery: JobManager (and its
components), Checkpointing, Yarn/Mesos, ZooKeeper: (no)
- The S3 file system connector: (no)
## Documentation
- Does this pull request introduce a new feature? (no)
- If yes, how is the feature documented? (not applicable)
cc @GJL
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/tillrohrmann/flink addHistoricJobView
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/5310.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #5310
----
commit a959b9411833e320065b328ed2fc936b58f911f4
Author: Till Rohrmann <trohrmann@...>
Date: 2018-01-16T17:45:53Z
[FLINK-8449] [flip6] Extend OnCompletionActions to accept an
SerializableExecutionGraph
This commit introduces the SerializableExecutionGraph which extends the
AccessExecutionGraph and adds serializability to it. Moreover, this commit
changes the OnCompletionActions interface such that it accepts a
SerializableExecutionGraph instead of a plain JobResult. This allows to
archive the completed ExecutionGraph for further usage in the container
component of the JobMasterRunner.
commit ca15b076c05ff940a12a240ba385e2434f93790b
Author: Till Rohrmann <trohrmann@...>
Date: 2018-01-18T14:02:36Z
[hotfix] [tests] Let BucketingSink extend TestLogger
commit 21c25502fb6d07c6fb65f18100dc6d4ec23e9d93
Author: Till Rohrmann <trohrmann@...>
Date: 2018-01-17T14:01:57Z
[FLINK-8450] [flip6] Make JobMaster/DispatcherGateway#requestJob type safe
Let JobMasterGateway#requestJob and DispatcherGateway#requestJob return a
CompletableFuture<SerializableExecutionGraph> instead of a
CompletableFuture<AccessExecutionGraph>. In order to support the old code
and the JobManagerGateway implementation we have to keep the return type
in RestfulGateway. Once the old code has been removed, we should change
this as well.
commit 7b7b0692582189b8e540e5ae022d351c45991e43
Author: Till Rohrmann <trohrmann@...>
Date: 2018-01-17T11:22:43Z
[FLINK-8453] [flip6] Add SerializableExecutionGraphStore to Dispatcher
The SerializableExecutionGraphStore is responsible for storing completed
jobs
for historic job requests (e.g. from the web ui or from the client). The
store
is populated by the Dispatcher once a job has terminated.
The FileSerializableExecutionGraphStore implementation persists all
SerializableExecutionGraphs on disk in order to avoid OOM problems. It only
keeps
some of the stored graphs in memory until it reaches a configurable size.
Once
coming close to this size, it will evict the elements and only reload them
if
requested again. Additionally, the FileSerializableExecutionGraphStore
defines
an expiration time after which the execution graphs will be removed from
disk.
This prevents excessive use of disk resources.
----
> Add SerializableExecutionGraphStore to Dispatcher
> -------------------------------------------------
>
> Key: FLINK-8453
> URL: https://issues.apache.org/jira/browse/FLINK-8453
> Project: Flink
> Issue Type: Improvement
> Components: Distributed Coordination, REST
> Affects Versions: 1.5.0
> Reporter: Till Rohrmann
> Assignee: Till Rohrmann
> Priority: Major
> Labels: flip-6
> Fix For: 1.5.0
>
>
> The {{Dispatcher}} should have a {{SerializableExecutionGraphStore}} which it
> can use to store completed jobs. This store can then be used to serve
> historic job requests from the web UI, for example. The default
> implementation should persist the jobs to disk and evict the in memory
> instances once they grow to big in order to avoid memory leaks. Additionally,
> the store should expire elements from disk after a user defined time.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
