My understanding is that, if a pod is configured to have one container, that container will run its "main" process as PID 1 and that's pretty much it. My pods have only one container each, and they very often have multiple processes running (always copies of the same process) - why does this happen?
On one cluster I have:
USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND
root 1 0.3 0.4 276668 76076 ? Ssl 16:50 0:48 python manage.py drive_consumer_worker
root 19 0.0 0.0 34432 2756 ? Rs 20:28 0:00 ps auxOn another cluster (running the same Deployment), I have:
USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND
root 1 1.2 0.3 1269420 104388 ? Ssl Mar16 240:12 python manage.py drive_consumer_worker
root 26 0.0 0.2 1044312 84160 ? S Mar16 0:01 python manage.py drive_consumer_worker
root 30 0.0 0.0 34440 2872 ? Rs 20:30 0:00 ps auxAs you can see, the memory size is significant enough to indicate that it's a "real" process, but I don't know what I can do to continue debugging. I don't see any pattern with the number of pod replicas defined and the process count.
Snippet from the deployment definition:
containers:
- args:
- newrelic-admin run-program python manage.py drive_consumer_worker
command:
- /bin/bash
- -cWhat is going on here?
It really depends on the parent process, if it doesn't spawn any children the process 1 is all you'll have in the container. In this case, it looks like python manage.py drive_consumer_worker is spawning children processes, so it will be more up to the application to control if it spawns children as more processes in the container.