Description
What would you like to be added?
The documentation of --quota-backend-bytes currently read as this:
Raise alarms when backend size exceeds the given quota (0 defaults to low space quota).
For me as a newbie, this means an alarm will be raised but etcd will continue to work fine. After running into this issue, I found that this is obviously not the case. If the quota is exceeded an alarm is raised and etcd is set to read-only mode, causing every Kubernetes cluster to fail hazardously.
The docs should clearly point out, that this sets the maximum storage that the database may ever consume. When exceeded, etcd is put to RO mode. The hint about "0 defaults to low space quota" is not helpful either. I need to dig into the source code to find that it's only 2Gi. This should be explicitly mentioned as well.
This blog entry has it correct. But a user normally doesn't search through blogs when trying to understand the commandline options.
Why is this needed?
Because the current doc is misleading and wrong.