Issue running opencloud with persistent storage via VirtFS (9p virtio)

[caracola487]

Ahoj!

I am trying to use a persistent storage outside my VM by using viortioFS and ran into an issue.

Same configuration running fine within its own bubble, when host file system is not mounted.
When host file system is mounted to share the config and data folder for instance, the opencloud container restarts each 10 - 15 seconds.
I already tried to change the accessmode for viortioFS from passthrough to mapped.

Is it not possible to share an external filesystem or especially virtiofs? I think opencloud run into this issue due to any kind of permission failures. How to avoid them?

log.txt

[micbar]

Your search service is crashing. Are there multiple openCloud processes accessing the same storage?

[caracola487]

No, only the one from this VM.
I just checked the permissions again and compared the accessmodes. Well, passthrough is looking good, permission wise. But the behavior is the same as mentioned in my origin post.

HOST: kix -> uid 1000;
VM: nonroot uid:gid -> 1000:1000
mapped:

passthrough:

[micbar]

opencloud-1 | {"level":"error","service":"idm","db":"/var/lib/opencloud/idm/idm.boltdb","error":"invalid argument","time":"2025-11-13T13:27:26Z","line":"github.com/opencloud-eu/opencloud/pkg/log/logrus_wrapper.go:50","message":"Error opening database"}

[micbar]

What is visible inside the container in /var/lib/opencloud/idm/ ?

[caracola487]

After a dozent of tries the container gave me the info I looked for.. I looks like the app is collapsing while creating this folder. It is not existant.

Update:

And again: After umount the data and config directory everything works as expected.

[micbar]

Nice progress. Then we can close this ticket.

[caracola487]

Aehm the problem is not solved yet. Because the container still crashes with the following error in case the data directory is mounted via virtiofs. :)

opencloud-1 | {"level":"error","service":"idm","db":"/var/lib/opencloud/idm/idm.boltdb","error":"invalid argument","time":"2025-11-13T13:27:26Z","line":"github.com/opencloud-eu/opencloud/pkg/log/logrus_wrapper.go:50","message":"Error opening database"}

[micbar]

And again: After umount the data and config directory everything works as expected.

sorry, i got this wrong.

But that clearly indicates that there is "something wrong" with the storage integration.
I think we can rule out opencloud as an application, because it works on direct storage mounts. So what is the exact difference between those two scenarios storage wise?

[caracola487]

Ahh..... Storage scenarios is the point I guess. I am using two diffrent filesystems. xfs <-> ext4.
I will try to get them in line and retry.

[caracola487]

Opencloud is now also running on a xfs partition.
Now I am facing the error:
opencloud-1 | {"level":"error","service":"opencloud","event":"opencloud: Failed service 'service.SutureService{exec:(func(context.Context) error)(0x55859955da40)}' (2.997431 failures of 5.000000), restarting: true, error: invalid argument","service":"service.SutureService{exec:(func(context.Context) error)(0x55859955da40)}","supervisor":"opencloud","restarting":true,"failures":2.9974314849675414,"threshold":5,"error":22,"time":"2025-11-18T13:38:07Z","line":"github.com/opencloud-eu/opencloud/opencloud/pkg/runtime/service/service.go:400","message":"service terminated"}

logs3.txt

Update:
I just tried the following combinations, certainly also due to missing experience. But I wondered if Alpine is handling some things in other way then ubuntu due to its lightweight. Obviously not in this case because the behavior was the same as I reported before.
mounted -> crash
unmounted -> nope

HOST: Linux mint - fs ext4
GUEST: Alpine - fs ext4

HOST: Linux mint - fs ext4
GUEST: Ubuntu Server 24.04.3 - fs ext4

[caracola487]

In the last days I tried a lot and finally switched over to NFS.
By changing the protocol I could "solve" the service.SutureService issue. But then I came into the extended file attributes hell. Alpine and unraid supporting NFS 4.2 and I was able to setfattr and getfattr within both systems. But it seems that there is a general problem within the combination of alpine as GUEST / unraid as host. Because I was not able to get or set extended file attributes via NFS. details

I used different Kernel variants for booth systems, because I found out that some Kernels had xattr issues with NFS "recently". Confusing..

Anyway... Finally I tried Ubuntu Server 24.04 and everything worked directly via NFS. But the service.SutureService error when using 9p is still present.

What is the sutureservice and what could be the root cause of this error?

[micbar]

We need to see which service is not starting. The error message is too verbose. Fixed in opencloud-eu/opencloud#2052

[caracola487]

Unfortunately I am not able to get the docker image opencloudeu/opencloud-rolling:daily run. I am facing an endless "This is not recommended for public instances! [yes | no = default] Do you want to configure OpenCloud with certificate checking disabled?" warning/error/question/whatever. I tried to run the init by hand but that did not work.
I will wait for the official 4.2 release and test your new approach.