Feature: finalizers #136
Feature: finalizers #136
Conversation
|
|
||
| To achieve this, the server containers in the `Pods` have | ||
| a `preStop` hook configured and finalizers are added to the `Pods` | ||
| ands `PersistentVolumeClaims`. |
| if err := os.Chmod(targetPath, 0755); err != nil { | ||
| cliLog.Fatal().Err(err).Msg("Failed to chmod") | ||
| } | ||
| } |
There was a problem hiding this comment.
Just for the record (one more time): LOL about this trick.
| } | ||
| // Ensure the cleanout is triggered | ||
| log.Debug().Msg("Server is not yet clean out. Triggering a clean out now") | ||
| if err := cluster.CleanOutServer(ctx, memberStatus.ID); err != nil { |
There was a problem hiding this comment.
I cannot immediately see how often this is called. It seems that as long as the cleanout is ongoing, a new cleanout is ordered every time this routine is called, which happens every time runPodFinalizers is run. I found a loop doing this but I could not immediately determine how fast this loop could turn. I do not think it is a real problem to call CleanOutServer again, but I have a bad feeling if many of these calls would be done in rapid succession, whilst it is trying to clean itself out.
There was a problem hiding this comment.
The interval depends on external events.
So while it is unlikely that these calls are made in rapid succession, it is possible.
fixes #143
This PR adds a finalizer to dbserver Pods and a preStop hook to prevent it from being deleted before all data is safe.
It also adds a finalizer to all PVCs to prevent them from being removed while their member still exists.
TODO: