[jira] [Commented] (FLINK-18955) Add snapshot path to job startup message

["Yuan Mei (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-18955?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17178920#comment-17178920 ] 

Yuan Mei commented on FLINK-18955:
----------------------------------

Hey [~NicoK], just to confirm, do you mean something similar to the following:

{{8004 [flink-akka.actor.default-dispatcher-3] INFO org.apache.flink.runtime.checkpoint.CheckpointCoordinator [] - Restoring job 36681198e38a2c91932eb131f47e3641 from latest valid checkpoint: Checkpoint 1 @ 1597663438566 for 36681198e38a2c91932eb131f47e3641 located at file:/var/folders/dm/5xn_h6n9135dwy4j27sr65zh0000gp/T/junit492898918775365121/junit2392379173263601046/checkpoints/36681198e38a2c91932eb131f47e3641/chk-1.}}

 

The fix is simple, but not sure whether we can test the log easily. Do you have a suggestion how to test the log?

> Add snapshot path to job startup message
> ----------------------------------------
>
>                 Key: FLINK-18955
>                 URL: https://issues.apache.org/jira/browse/FLINK-18955
>             Project: Flink
>          Issue Type: Improvement
>          Components: Runtime / Checkpointing
>    Affects Versions: 1.10.1, 1.12.0, 1.11.1
>            Reporter: Nico Kruber
>            Priority: Major
>              Labels: usability
>             Fix For: 1.12.0, 1.11.2, 1.10.3
>
>
> When a job is started from a checkpoint or savepoint (I'm using snapshot as the unanimous term below), the {{CheckpointCoordinator}} prints a log line like this:
> {code}
> 2020-08-13 13:50:51,418 INFO  org.apache.flink.runtime.checkpoint.CheckpointCoordinator    [] - Restoring job 220d8a4953cd40198b6eb3b1ec0cece0 from latest valid checkpoint: Checkpoint 357 @ 1597326576925 for 220d8a4953cd40198b6eb3b1ec0cece0.
> {code}
> I propose to add the path to the snapshot to this message because which snapshot is taken for restore may actually not be that obvious for the user: even if a savepoint was specified in the job start command, e.g. in a Kubernetes pod spec, an HA store could overrule the decision and take a more recent snapshot instead. If that snapshot is a savepoint, it is not that easy to map this to checkpoint IDs and find out which savepoint the job actually started from.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)