Deploy by 11/16?

Is there any chance of getting Update paws-data-pipeline.toml by c-simpson · Pull Request #46 · CodeForPhilly/cfp-sandbox-cluster · GitHub merged/deployed before our meeting Tuesday night?

Thanks,
Cris

Done! Let me know if all looks good :slight_smile:

1 Like

Thank you, will check.
What’s the right process to use in the future for this kind of thing?

You got the process right, I was just a little bit asleep at the wheel :sweat:

This is the ideal spot for these kinds of requests to come in though, and feel free to bump us through any other channels when we’re a little slow on the uptake

1 Like

Thanks!

It looks like something is not going well:
0/3 ContainerCreating 0 76m

‘Describe’ doesn’t show anything else and I don’t have the rights to see events. Any ideas?

@Cris I can see that the new pod is stuck waiting for its persistent volume to become available, and it’s not becoming available because it is mounted to the previous pod instance still:

I’m going to manually delete the previous pod and replicaset instance to allow the new pod to get ahold of the persistent volumes and come online

To fix this issue, you should set .spec.strategy.type==Recreate in the Deployment. The default strategy, RollingUpdate, tries to bring the new version online before shutting down the old version, but this is not compatible with using persistent volumes that can only be attached to once instance at a time. You must use the Recreate strategy with exclusive persistent volumes are in play, as that strategy will terminate the previous version before creating the new one.

Will do Recreate next time. Thanks!

This actually brings up an interesting point-- It’s best practice to define any workload that uses ReadWriteOnce volumes as a StatefulSet. If little problems like this continue to pop up, it may be worth considering redefining this workload