[
https://issues.apache.org/jira/browse/FLINK-18312?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17138276#comment-17138276
]
Chesnay Schepler commented on FLINK-18312:
------------------------------------------
I suppose for synchronizing between rest servers we would need to put all
caching/persistent data structures/operations behind the RPC layer, i.e., into
the Dispatcher (or some component thereof).
This may not be too bad in the long run, and would solve a number of issues.
I don't really want to start doing redirections in the REST layer; then we have
2 separate of redirecting requests. IIRC that is what we wanted to avoid
initially.
> SavepointStatusHandler and StaticFileServerHandler not redirect
> ----------------------------------------------------------------
>
> Key: FLINK-18312
> URL: https://issues.apache.org/jira/browse/FLINK-18312
> Project: Flink
> Issue Type: Bug
> Components: Runtime / REST
> Affects Versions: 1.8.0, 1.9.0, 1.10.0
> Environment: 1. Deploy flink cluster in standlone mode on kubernetes
> and use two Jobmanagers for HA.
> 2. Deploy a kubernetes service for the two jobmanagers to provide a unified
> url.
> Reporter: Yu Wang
> Priority: Major
>
> Savepoint:
> 1. Deploy our flink cluster in standlone mode on kubernetes and use two
> Jobmanagers for HA.
> 2. Deploy a kubernetes service for the two jobmanagers to provide a unified
> url.
> 3. Send a savepoint trigger request to the leader Jobmanager.
> 4. Query the savepoint status from leader Jobmanager, get correct response.
> 5. Query the savepoint status from standby Jobmanager, the response will be
> 404.
> Jobmanager log:
> 1. Query log from leader Jobmanager, get leader log.
> 2. Query log from standby Jobmanager, get standby log.
>
> Both these two requests will be redirect to the leader in 1.7.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
