100% CPU usage used with podman #157
Comments
|
With what workloads? |
|
No load at all. I guess a maximum of 1 site call a second.
What I forgot, there is a tor middle relay running which relays about 1MBit/s.
So no significant load I guess.
The CPU load dropped to normal levels after approximately half an hour. So it seems to be kind of sporadic?
|
|
Another thing I just noticed, it seems to limit my tor middle relay as I'm only having a throughput of 1MBit/s , but my network connection is capable of 200MBit/s so it seems it limits the throughput a lot. |
|
https://github.com/rootless-containers/slirp4netns/blob/master/slirp4netns.1.md#bugs this might be related |
|
changing |
|
@AkihiroSuda I remember we already discussed it but cannot find the comment now. Isn't #128 something that can be fixed with changing the receive buffer for the slirp4netns socket only instead of changing the default for the network namespace? |
|
True, but I couldn't dig out which buffer needs to be updated. I think setting sysctl was not needed when mtu=1500, I can't remember |
|
Is this still an issue? |
|
Closing, but feel free to reopen if this is still an issue |
Hi,
I'm using this project with podman. I'm not sure if this is related to podman or to slirp4netns. So I'm posting it here and at podman.
I have one container with networking mode host, all the others have no setting so I guess podman namespaces them I have four containers running, one with root permissions.
This is the command using all 100% of the CPU on one core.
/usr/bin/slirp4netns --api-socket /run/user/1000/libpod/tmp/18c14ae7aed050b0063c4b5f75af384721a280011fbf95a7e85042112af7458d.net --disable-host-loopback --mtu 65520 -c -e 3 -r 4 --netns-type=path /run/user/1000/netns/cni-7913c5a5-8763-5636-16ad-286fa2a0a376 tap0As far I can tell everything works as expected. So no problems on the functional site.
Any idea what is causing this?
The text was updated successfully, but these errors were encountered: