With autoscaler, if I stop a flink deployment with...
# troubleshooting
m
With autoscaler, if I stop a flink deployment with a savepoint (
upgradeMode: savepoint, allowNonRestoredState: false, takeSavepointOnUpgrade: true
), can I start it again from savepoint (
initialSavepointPath
in config) with parallelism of 1 (1 TM pod) and let autoscaler scale it up again? I've been experimenting and it seems like the job dies, not being able to fully (or correctly) restore state from original parallelism of 10 (10 TM pods). Is it correct to expect that the JM will be able to resume the job from savepoint if number of pods had changed after restart?