This message was deleted.
# troubleshootings
Slackbot
03/17/2023, 6:57 AMThis message was deleted.
b
Brandon Tan
03/17/2023, 7:08 AMYou could HARD RESET your supervisor and then it will start reading from the first available offset.
s
Sanjay
03/17/2023, 7:49 AMyeah, that is what we had to do currently but it is very difficult to hard reset all the time. 🙂
Sanjay
03/17/2023, 7:49 AMis this related — https://github.com/apache/druid/issues/11658 ?
g
Gian Merlino
03/21/2023, 5:40 AMin general i would recommend setting
resetOffsetAutomatically: falseGian Merlino
03/21/2023, 5:40 AMin normal situations (consumer keeps up with producer) the consumer should never fall out of range
Gian Merlino
03/21/2023, 5:40 AMvery strange to see this happen with so low load (7–10 messages/s) — generally when I've seen issues it's been to consumer being overloaded and falling behind
Gian Merlino
03/21/2023, 5:41 AMis your kafka retention period extremely short? like similar to or even less than your
taskDurationGian Merlino
03/21/2023, 5:42 AM(default is 1 hour duration; kafka retention should generally be minimum 4x
taskDurations
Sanjay
03/21/2023, 5:44 AMThe kafka retention is set to 1 day for the topic
Sanjay
03/21/2023, 5:45 AMThe task duration is set to an hour in supervisor task -
"taskDuration": "PT3600S",g
Gian Merlino
03/21/2023, 6:02 AMhmm, well, that was my big idea! — i'm not sure what could be going on here
2 Views