Hi Team, I have savepointing and HA enabled with f...
# troubleshooting
n
Hi Team, I have savepointing and HA enabled with flink 1.18. now as part of code change one operator was deleted. so during deployment, task manager was not coming up so I tried with --allowNonRestoredState. But still not Flinkdeployment is going in RECONCILING state. https://nightlies.apache.org/flink/flink-docs-release-1.19/docs/ops/state/savepoints/#allowing-non-restored-state. Any help would be really appreciated. CC: @Dheeraj Panangat @Abhishek Joshi
p
Do you have
pipeline.auto-generate-uids
set? If you don't, Flink will struggle to match the state with the operators
n
No @Pedro Mázala we have hard coded uids
p
Hum... So you got the right uid but it's always stuck on reconciling even with non restored state set. What is your state size? Is the machine struggling on resources?
And what logs do you see on your JM?
n
State size is in few KBs. JM I am seeing application completed and behaving like batch job. Also health check url is failing
p
JM I am seeing application completed and behaving like batch job.
I saw it in the past when there was an issue while connecting to kafka. Killing everything and redeploying made the trick. Perhaps if you enable debug logs you may find some deeper knowledge on why it's behaving like a bounded source
n
Sure. Thanks @Pedro Mázala
🙌 1