Installation is brittle
Not sure exactly where to put this, but I want to start tracking some of the ways our Kubernetes installation is brittle.
- Updating a GitLab installation can take a long time, with various pods coming up before others and then erring until dependent pods come up. Sometimes they time out though, and end up dead. e.g. I had a worker give up on connecting to the gitlab instance and I had to manually kill the runner pod.