The problem

Two things are broken in the namespace shop. The pods of the Deployment api restart every few seconds. The last deploy of the Deployment web is stuck, and the 3 old pods still serve the traffic.

The pod api-8b74f5cbb-4zbdm has two containers: api prints api started, then db unreachable, and exits with code 1; log-shipper prints shipped every 10 seconds. The Deployment web has 3 revisions: 1 runs nginx:1.27, 2 runs nginx:1.28, and 3 asks for nginx:1.288, a tag that does not exist, so its new pod stays in ImagePullBackOff. The pod cache-0 runs Redis. Every command needs -n shop.