Do not use Rolling Update strategy on deployments for stateful apps

[onedr0p]

This will bring up a new pod before the old one is destroyed which can lead to data corruption on the persistence data with stateful applications.

Use a deployment strategy of Recreate instead, this will destroy the old pod before the new one starts.

Another option is to use statefulsets instead of deployments for stateful workloads.

[KyleTryon]

Thank you for this! I have gotten lucky so far, one user mentioned that file locking on the local volume mount might be the reason this has been working so far.

Going to look into statefulsets and see if there is maybe anything we can do to bring back rolling updates at some point.

[onedr0p]

Deployments with the Recreate strategy is probably the most universal use for your users here who seems to be mostly docker-compose migrants. Feel free to pick my brain more if needed, I've been using containers for ~8 years and Kubernetes for ~5 years and my $dayJob is a Senior SRE.

[KyleTryon]

Thanks @onedr0p! I've checked out your repos and can see you've definitely got this all figured out 🙇. I can tell im going to be reading through your repos for a while.

[onedr0p]

definitely got this all figured out

At the same time I feel like I have no idea what I am doing 😆

[onedr0p]

If you're interested check out this discord, lots of people doing the home k8s goodness. It could be a good place to bounce ideas off of.

https://discord.gg/k8s-at-home