hey, just started using flink 2.1.0 with kubernete...
# troubleshooting
m
hey, just started using flink 2.1.0 with kubernetes-operator on AWS EKS with session mode (our focus is on batch processing, not streaming), I've set up session cluster with example app that finishes processing, and generally it works (jobmanager is up), but my current problem is the fact that if we get an error in the processing and taskmanager pod is gone, then trying to get logs via jobmanager ends in getting http 500 because corresponding taskmanager instance is gone (pod shut down). Can someone point me how can I achieve somekind of integrated log durability for those? I know I can direct logs to other system but that's a bit inconvenient - I would prefer to somehow get a redirection in jobmanager so it is a seamless (as if in the jobmanager ui and so on, an not in external system such as loki+grafana or elastickstack). Thanks in advance.
historyserver? https://nightlies.apache.org/flink/flink-docs-release-2.1/docs/deployment/advanced/historyserver/ but seems like it's not integrated with a flink kubernetes operator?
r
do you run the 1.13 kubernetes flink operator version ?
m
Yeah
( also after today I know that historyserver is not enough, I need just another system for log processing, which I did before a few times with fluentd/fluentbit/elastic or promql/loki stacks)
ah sorry no, @raphaelauv actually its 1.12.1
Switched to 1.13.0 today
👍 1