implementing ability to reset the state file used by etcd server

kopeio / etcd-manager

operator for etcd: moved to https://github.com/kubernetes-sigs/etcdadm

Apache License 2.0

164 stars 45 forks source link

implementing ability to reset the state file used by etcd server #220

Open mmerrill3 opened 5 years ago

mmerrill3 commented 5 years ago

This is a first pass at https://github.com/kopeio/etcd-manager/issues/218, let me know if this is acceptable approach from a high level

justinsb commented 5 years ago

Thanks for this PR ... I really need to think about this one as this shouldn't have mattered in my mental model (the per node info is the node's view of the world), but from the issue it sounds like my mental model is causing problems. Hopefully this will drive me to better document my model and then hopefully I can figure out where things are going wrong.

I'm trying to triage this against releasing the latest kops though ... how high consequence is this? Is it just log messages, or does this actually cause a failure?

mmerrill3 commented 5 years ago

@justinsb this doesn't cause any failures, its benign. It's log messages currently. Actually, doing this work was an exercise to understand etcd-manager a bit more in case I need to put out a fire in production at some point in time.

Do we need this info? From what I saw, the presence of that file means "don't do the legacy way of discovering the cluster", which was manifest files in the legacy directory. Maybe just remove that legacy behavior, and keep another indicator to show that the new etcd server has been initialized and joined a cluster?