[
https://issues.apache.org/jira/browse/FLINK-11352?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16749776#comment-16749776
]
Congxian Qiu commented on FLINK-11352:
--------------------------------------
Hi, [~till.rohrmann] I did not find any existing test case to cover the
`testCheckpointRecoveryFailure()`,
should I add it in `BlobLibraryCacheRecoveryITCase.java` or
`DispatcherHATest.java` or anywhere else?
`testCheckpointRecoveryFailure()` wants to verify `logged failures during
recovery
properly`([FLINK-3185|https://issues.apache.org/jira/browse/FLINK-3185]). The
test starts two jobmanagers(a leader and a slave) _—>_ Wait for the job to be
running —> delete the directory contains blob and jobgraph files such as below
—> kill the leader jobmanager process —> wait until timeout or the condition
reached (the specified log message logged by the previous slave) — test
completed.
the deleted directory
/var/folders/px/yygqbw9s0c399k5xzk2yzww40000gn/T/junit4831596921752388878/junit8832725465506871161/recovery
├── default
│ └── blob
└── submittedJobGraph0d503d83da70
> Check and port JobManagerHACheckpointRecoveryITCase to new code base if
> necessary
> ---------------------------------------------------------------------------------
>
> Key: FLINK-11352
> URL: https://issues.apache.org/jira/browse/FLINK-11352
> Project: Flink
> Issue Type: Sub-task
> Components: Tests
> Reporter: Till Rohrmann
> Assignee: Congxian Qiu
> Priority: Major
>
> Check and port {{JobManagerHACheckpointRecoveryITCase}} to new code base if
> necessary
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
