Chanhyeong Cho created FLINK-37559:
--------------------------------------

             Summary: Produce warn log whenever CommitterOperator#snapshotState 
called
                 Key: FLINK-37559
                 URL: https://issues.apache.org/jira/browse/FLINK-37559
             Project: Flink
          Issue Type: Bug
    Affects Versions: 1.20.1
            Reporter: Chanhyeong Cho


Too many 'pendingCommittables' Name collision logs are produced due to the 
reasons mentioned in the title.

 
{code:java}
Name collision: Group already contains a Metric with the name 
'pendingCommittables'. Metric will not be reported.[<ip address>, taskmanager, 
<tm host>, <job name>, <task name>: Committer, <task index>] {code}
We've started use flink 1.20.1, and don't know about previous versions.
I guess this occurs after change of 
https://issues.apache.org/jira/browse/FLINK-36455.

 

CommitterOperator#snapshotState raise CommittableCollector#copy, and that calls 
 
SinkCommitterMetricGroup#setCurrentPendingCommittablesGauge which registers 
'pendingCommittables' gauge into metricGroup.
Because the 'pendingCommittables' gauge was already registered at creation 
time, it produces warn message at every 'copy()' called.
 



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to