https://flink.apache.org/ logo
Join Slack
Powered by
Hi, I'd like to "reboot" an existing k8s FlinkDepl...
# troubleshooting
u
Hi, I'd like to "reboot" an existing k8s FlinkDeployment with 
upgradeMode: last-state
once , discarding the current state, but once it has rebooted it should still be in last-state. Basically restart from 0. One way to do so would be to delete the 
FlinkDeployment
and recreate it - but is there a way to do it just editing the existing resources, not deleting the FlinkDeployment?
a
The docs states "the operator will redeploy the job to the savepoint defined in the `initialSavepointPath`" but if I call correctly checkpoints also should work.
u
Thank you - yep, checkpoints work for that, but can I somehow start from an "empty checkpoint"?
a
Ah. Good question. Not sure, but I imagine it would blow up complaining it couldn't find the save-/checkpoint file.
For my jobs I usually have a bootstrapped savepoint that only contains staged enrichment data (imagine geo reference data), so I've never had that particular use-case.
r
@Urs Schoenenberger - what are you trying to achieve by not deleting of the flinkdeployment? I had some experience with deletion and creation of a flinkdeployment and while it is slower to recover than simply editing it - it works well. As for the above suggestion, have you tried pointing 
initialSavepointPath
to a non-existing save-point path? it might work
3 Views
Open in Slack
PreviousNext
Powered by