[ https://issues.apache.org/jira/browse/FLINK-8057?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17027344#comment-17027344 ]
Till Rohrmann commented on FLINK-8057: -------------------------------------- Are we planning to fix this issue [~kkl0u]? If not, then please unassign and close the issue as won't do. > KvStateRegistry#registerKvState may randomly fail > ------------------------------------------------- > > Key: FLINK-8057 > URL: https://issues.apache.org/jira/browse/FLINK-8057 > Project: Flink > Issue Type: Improvement > Components: Runtime / Queryable State > Affects Versions: 1.4.0 > Reporter: Chesnay Schepler > Assignee: Kostas Kloudas > Priority: Major > > The {{KvStateRegistry#registerKvState}} method is a bit weird. On each call, > a new KvStateID is generated, and we then check whether a state for this ID > was already registered. Realistically, this is never the case making the > check unnecessary, and on the off-chance that it does happen the job will > fail for no good reason. > {code} > KvStateID kvStateId = new KvStateID(); > if (registeredKvStates.putIfAbsent(kvStateId, kvState) == null) { > KvStateRegistryListener listener = this.listener.get(); > if (listener != null) { > listener.notifyKvStateRegistered( > jobId, > jobVertexId, > keyGroupRange, > registrationName, > kvStateId); > } > return kvStateId; > } else { > throw new IllegalStateException(kvStateId + " is already registered."); > } > {code} -- This message was sent by Atlassian Jira (v8.3.4#803005)