[
https://issues.apache.org/jira/browse/FLINK-6206?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16393157#comment-16393157
]
ASF GitHub Bot commented on FLINK-6206:
---------------------------------------
Github user casidiablo commented on the issue:
https://github.com/apache/flink/pull/5399
The reason I proposed the change is that any unexpected behavior, even when
you can recover from it, can be indicative of a potential bug/misconfiguration.
INFO logs are by far noisier than errors, which means it is tough to identify
problems unless you log them with ERROR (or at least WARN).
I'd say we at least demote this to WARN. When one is analyzing logs, trying
to find leads to current problems, it's common to filter out all INFO logs
(otherwise it's mayhem); so anything that caused an exception to be thrown
should be highlighted.
> Log task state transitions as warn/error for FAILURE scenarios
> --------------------------------------------------------------
>
> Key: FLINK-6206
> URL: https://issues.apache.org/jira/browse/FLINK-6206
> Project: Flink
> Issue Type: Improvement
> Components: Core
> Affects Versions: 1.2.0
> Reporter: Dan Bress
> Priority: Critical
>
> If a task fails due to an exception, I would like that to be logged at a warn
> or an error level. currently its info
> {code}
> private boolean transitionState(ExecutionState currentState, ExecutionState
> newState, Throwable cause) {
> if (STATE_UPDATER.compareAndSet(this, currentState, newState)) {
> if (cause == null) {
> LOG.info("{} ({}) switched from {} to {}.",
> taskNameWithSubtask, executionId, currentState, newState);
> } else {
> LOG.info("{} ({}) switched from {} to {}.",
> taskNameWithSubtask, executionId, currentState, newState, cause);
> }
> return true;
> } else {
> return false;
> }
> }
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
