I've Got 'Night Of The Living Dead' On My Homelab Server
=> There are 90 zombie processes.
On one of my Homelab servers running Ubuntu Jammy, I always seem to get zombie processes. A quick check with ps -eo pid,ppid,stat,cmd | grep -w Z shows them all . It just bugs me. I shut down the server in the most nicest of ways I know how with sudo shutdown -h now but I always get zombie processes shown on start up.
Am I missing something? Do these show up on your servers? How do you deal with them besides just ignoring them if they are ?
Awww man. Here I was thinking we were having a movie night.
Right? I’ve got the original and the 90s version in Jellyfin on my home lab server. 🧟♂️
LOL
Zombies that stick around for more than a few seconds indicate a signal problem in the parent process, where its init is stuck in the "wait" state, so the entry remains in the PID table.
It could be harmless, but it could become a problem if you need the resources. Curl shouldn't be doing this on its own.
That's the thing. None of them are consuming resources. I guess I should just ignore them, but it irritates me when I start my server, to see zombie processes. Makes me think something is askew.
Zombie processes do not use resources, well, a little, it's basically an entry describing how it exited.
The parent process is the thing keeping the zombie entry open. Killing it's parent should work if they bother you.
Agreed, but a very poorly-written program having a hanging memory or disk write, or a file lock could become a problem, especially if hundreds or thousands of zombies are waiting for something.
disturbingly pervasive shuffling noises intensify
that sounds like poor garbage collection in an application.
I've written software that had similar issues when writing to files and I failed to "close" the file after writing.
processes stay open, files stay open. 500 byte processes times 10000 orphaned process can make for a bad time.
No idea about your specific case, but in the past when I've had frequent zombies it's been due to some blocking network process; typically NFS mounts.
They don't use any resources, but I agree they're annoying and could indicate some underlying issue.
I would check for issues with I/O and network access, especially anything that happens at kernel level. Look for anything suspicious in dmesg.
I'll check that out.
Zero zombies here. I have a couple of Debian servers and one repeatedly upgraded Ubuntu on noble numbat that I'm too lazy to migrate to Debian. None have zombies.
Do you run a DE? Mine are headless.
I'm not familiar with the acronym. I'm going to assume Desktop Environment. My servers are all headless as well
Sorry. That is what I meant.
s'ok bro. I'm not on all cylinders today.
I got a few on my laptop but none on either of my long running homelab boxes (70-80 days uptime). On my laptop they all seem related to espeak, the tts program. Is there any pattern in what processes yours are from?
Ahh see I shut down my servers at night. I just couldn't justify having them run while I was sleeping, and since I am the only user.
Makes sense. I share my media library with 10-15 friends so there's usually a few streams late at night, and scrubs, container updates, and backups run early morning at like 2-4am.
I've never run into a situation where it's a problem. But I just checked one of my Ubuntu servers and don't have any. What were the processes doing? What do they belong to?
@[email protected] @[email protected] Most of them are from [curl] but one is from [health.sh]. I am assuming one of the docker containers uses
curlin it's processes as I have not initiated anycurlcommands.If you use containers with health checks (including with curl), you need to tell docker (or podman) to provide an init process to reap child processes. For docker that means providing
--initwhen running a container. It's a pretty common problem.What is their parent process?
Lots of these: Zombie PID: 230650 | Parent PID: 7791 | Parent Name: spawn-unnamed
Two of those: Zombie PID: 61072 | Parent PID: 7791 | Parent Name: bash
Lots of these: Zombie PID: 56798 | Parent PID: 7791 | Parent Name: health.sh
Lots of these: Zombie PID: 16761 | Parent PID: 7646 | Parent Name: curl
...and a box of naked lady tee's
What is 7791?
7791 7768 201 /usr/sbin/netdata -u netdat 8.8 0.5 01:49:50
Sounds like netdata doing health checks but not always reaping its children. If you can reproduce it, I'd file a bug report.
Seems like someone already did this:
https://github.com/netdata/netdata/issues/20565
Maybe upgrading will fix it?