I was wondering, Do you know of a limit on how many rootless conrainers can one run on a linux host?
Running fedora server, I have resources but once I pass about 15 containers podman starts to hang and crash.
I then need to manually delete the storage folder under ~./local/share/... for podman to work again.
It might be related to user ns keep-id flag.
You must log in or # to comment.
I’m not convinced what you run into is a specific podman issue. It’s a resource issue and configuration issue likely. “vanila” podman with proper rootless containers will run as much workload as machine can handle from my experience. My company costomers seem to be running production workloads with it just fine.
Oh wait, by rootless container you really meant running podman rootless? still don’t see an issue though. What specifically are you doing? I mean, what’s the configuration and what’s the workload?
My config is nothing special. Running all containers as user 1000 via qadlets. Sporadically, I get:
Dec 05 13:40:27 home-lab systemd-coredump[200811]: [🡕] Process 200795 (podman) of user 1000 dumped core. Module libbz2.so.1 from rpm bzip2-1.0.8-18.fc40.x86_64 Module libsepol.so.2 from rpm libsepol-3.7-2.fc40.x86_64 Module libpcre2-8.so.0 from rpm pcre2-10.44-1.fc40.x86_64 Module libcap-ng.so.0 from rpm libcap-ng-0.8.4-4.fc40.x86_64 Module libgpg-error.so.0 from rpm libgpg-error-1.49-1.fc40.x86_64 Module libpam_misc.so.0 from rpm pam-1.6.1-4.fc40.x86_64 Module libpam.so.0 from rpm pam-1.6.1-4.fc40.x86_64 Module libattr.so.1 from rpm attr-2.5.2-3.fc40.x86_64 Module libacl.so.1 from rpm acl-2.3.2-1.fc40.x86_64 Module libcrypt.so.2 from rpm libxcrypt-4.4.36-10.fc40.x86_64 Module libeconf.so.0 from rpm libeconf-0.6.2-2.fc40.x86_64 Module libsemanage.so.2 from rpm libsemanage-3.7-2.fc40.x86_64 Module libselinux.so.1 from rpm libselinux-3.7-5.fc40.x86_64 Module libaudit.so.1 from rpm audit-4.0.2-1.fc40.x86_64 Module libseccomp.so.2 from rpm libseccomp-2.5.5-1.fc40.x86_64 Module podman from rpm podman-5.3.1-1.fc40.x86_64 Stack trace of thread 200805: #0 0x0000558789bfa4a1 runtime.raise.abi0 (podman + 0x934a1) #1 0x0000558789bd6cc8 runtime.sigfwdgo (podman + 0x6fcc8) #2 0x0000558789bd51a5 runtime.sigtrampgo (podman + 0x6e1a5) #3 0x0000558789bfa7a9 runtime.sigtramp.abi0 (podman + 0x937a9) #4 0x00007efdbc0cad00 __restore_rt (libc.so.6 + 0x40d00) #5 0x0000558789bfa4a1 runtime.raise.abi0 (podman + 0x934a1) #6 0x0000558789bbda26 runtime.fatalpanic (podman + 0x56a26) #7 0x0000558789bbc998 runtime.gopanic (podman + 0x55998) #8 0x0000558789bd64d8 runtime.sigpanic (podman + 0x6f4d8) #9 0x000055878a5a7842 github.com/containers/storage.(*layerStore).load (podman + 0xa40842) #10 0x000055878a5a9608 github.com/containers/storage.(*store).newLayerStore (podman + 0xa42608) #11 0x000055878a5bc7dd github.com/containers/storage.(*store).getLayerStoreLocked (podman + 0xa557dd) #12 0x000055878a5bc935 github.com/containers/storage.(*store).getLayerStore (podman + 0xa55935) #13 0x000055878a5cc451 github.com/containers/storage.(*store).Mounted (podman + 0xa65451) #14 0x000055878ac99b88 github.com/containers/podman/v5/libpod.(*storageService).UnmountContainerImage (podman + 0x1132b88) #15 0x000055878abec81a github.com/containers/podman/v5/libpod.(*Container).unmount (podman + 0x108581a) #16 0x000055878abe8865 github.com/containers/podman/v5/libpod.(*Container).cleanupStorage (podman + 0x1081865) #17 0x000055878abe965b github.com/containers/podman/v5/libpod.(*Container).cleanup (podman + 0x108265b) #18 0x000055878ac6c2ce github.com/containers/podman/v5/libpod.(*Runtime).removeContainer (podman + 0x11052ce) #19 0x000055878ac6aad0 github.com/containers/podman/v5/libpod.(*Runtime).RemoveContainer (podman + 0x1103ad0) #20 0x000055878ad05948 github.com/containers/podman/v5/pkg/domain/infra/abi.(*ContainerEngine).removeContainer (podman + 0x119e948) #21 0x000055878ad06745 github.com/containers/podman/v5/pkg/domain/infra/abi.(*ContainerEngine).ContainerRm.func1 (podman + 0x119f745) #22 0x000055878ace297b github.com/containers/podman/v5/pkg/parallel/ctr.ContainerOp.func1 (podman + 0x117b97b) #23 0x000055878aade678 github.com/containers/podman/v5/pkg/parallel.Enqueue.func1 (podman + 0xf77678) #24 0x0000558789bf8c41 runtime.goexit.abi0 (podman + 0x91c41) ELF object binary architecture: AMD x86-64I have enugh RAM and CPU and Disk to spare…
when this error happens, I cant run any
podmancommands without a core dump. i.e I cantpodman imagespodman psand so on… The only solution is to delete the storage folder manually, pull all the images again. and it’s back to normal.
The practical limit to the number of containers you can run on one system is in the high hundreds or more thousands, depending on how you configure some things, and your available hardware. It’s certainly more than you’ll even use unless you get into some auto-scaling swarm config stuff.
The issue is more about resource limits, and access to shared resources. I’d start by trying to figure out if there are certain specific containers that don’t play well together. Bring your setup online slowly, one container at a time, and take note of when things start to get funky. Then start testing combinations of those specific containers. See if there’s one you can remove from the mix that suddenly makes things more stable.
Small update.
It seems to be caused by
UserNS=keep-id. when adding it to an image with a lot of files, podman hangs for a while, then crashes when doind itschown. This causes some layers to be invalid.