You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Check the Flink currentFetchEventTimeLag metric for a completely blocked datastream. It will not grow over time. But the currentEmitEventTimeLag will continue to grow.
What doesn't meet your expectations?
The currentFetchEventTimeLag metric should grow over time even if the datastream is completely blocked.
Anything else?
The lastSplitUpdateTime will not update for a completely blocked datastream.
In flink, the currentEmitEventTimeLag is calculated by idleTime or currentTime, so this metric can keep grow for a completely blocked datastream.
Are you willing to submit a PR?
I'm willing to submit a PR!
The text was updated successfully, but these errors were encountered:
xiangyuf
changed the title
[Bug] CURRENT_FETCH_EVENT_TIME_LAG metric calculation is wrong when the datastream is completely blocked
[Bug] Flink currentFetchEventTimeLag metric calculation is wrong when the datastream is completely blocked
Sep 24, 2024
Hi @xiangyuf, I'm wondering if currentFetchEventTimeLag need to increase when datastream is blocked, as another pr #4281 update the logic of emitEventTimeLag, the metrics will not increase when no new records are in. I'm not sure if the logic should be synced between this two metrics, currentFetchEventTimeLag and emitEventTimeLag.
Search before asking
Paimon version
master(8ac121a)
Compute Engine
Flink
Minimal reproduce step
Check the Flink currentFetchEventTimeLag metric for a completely blocked datastream. It will not grow over time. But the currentEmitEventTimeLag will continue to grow.
What doesn't meet your expectations?
The currentFetchEventTimeLag metric should grow over time even if the datastream is completely blocked.
Anything else?
The lastSplitUpdateTime will not update for a completely blocked datastream.

In flink, the currentEmitEventTimeLag is calculated by idleTime or currentTime, so this metric can keep grow for a completely blocked datastream.

Are you willing to submit a PR?
The text was updated successfully, but these errors were encountered: