Kubernetes: How do you prefer your multinode releases?

How are you deploying your releases when containers need to connect to each other?

  • Deployments
  • StatefulSets
  • Both (Depends on the release)

0 voters

Taking a poll because I’m curious how the community handles this. Currently at Codedge we use exclusively deployments, linking containers together with libcluster.

I’ve thought for a while about migrating to StatefulSets because

  1. The BEAM was designed for relatively static clusters.
  2. Cachex uses static node lists for distributed caching.
  3. HorizonalPodAutoscalers can’t (to my knowledge) ensure an odd number of replicas for a deployment.

Any input would be much appreciated. Deployments work well as is, but I’m always looking for better solutions.

1 Like