See here for solved Desktop, but unsolvable mobile error: https://help.nextcloud.com/t/best-way-to-edit-documents/226232/18
Hello everyone,
I’m seeking assistance with a stubborn Collabora CODE issue on my Nextcloud Docker Compose setup. My Nextcloud web interface works perfectly with Collabora, but the iOS mobile app consistently fails to open documents, showing “network connection lost.”
My Setup:
- Nextcloud (latest) and Collabora CODE (
collabora/code:24.04.11.1.1
) via Docker Compose. - Host OS: Ubuntu 22.04 LTS on a VPS.
Troubleshooting Steps & Findings (Chronological):
- Initial Errors:
coolmount failed
: Fixed by addingcap_add: [MKNOD, SYS_ADMIN]
to the Collabora service indocker-compose.yml
.ENOENT
on/var/cache/coolwsd/storage.xml
: Fixed by defining and mounting a Docker volume (coolwsd-data
) to/var/cache/coolwsd
, and then runningsudo chmod 777 /var/lib/docker/volumes/nextcloud_coolwsd-data/_data
on the host to ensure writability.touch testfile.txt
inside the container confirmed writability forcool:cool
(UID/GID 1001).
- Persistent
ENOENT
&linkOrCopy
for/opt/cool/child-roots/.../tmp/sharedpresets/template
:
- Despite
/var/cache/coolwsd
being writable, logs still showed:ERR Failed to get the realpath of [...]/tmp/sharedpresets/template (ENOENT: No such file or directory)
ERR linkOrCopy: nftw() failed for '/opt/cool/child-roots/.../tmp/sharedpresets/template'
- Confirmed the source
template
existed in/opt/cool/systemplate
. The issue was the copying/linking process failing. - Attempts that DID NOT fix this:
- Adding
extra_params=--o:mount_jail_tree=false
to environment. - Changing Collabora image to
collabora/code:24.04.11.1.1
(fromlatest
).
- Adding
- Current Situation (
privileged: true
for diagnosis):
- As a last resort diagnostic step, I set
privileged: true
for the Collabora container (removingcap_add
). - Result: The original
ENOENT
forsharedpresets/template
DISAPPEARED! This suggestsprivileged
mode provided the necessary permissions/bypasses for that specific issue. - NEW Critical Errors Appeared in Logs:
ERR Failed to stat or chown 65534:65534 /opt/cool/child-roots/.../etc/ld.so.conf: Invalid argument missing cap_chown?, disabling linkable
ERR link("/opt/cool/systemplate/etc/ld.so.conf", "/opt/cool/child-roots/.../etc/ld.so.conf") failed: Invalid argument. Very slow copying path triggered.
Summary of Current Problem:
Even with privileged: true
, the Collabora container is failing to perform fundamental file system operations (chown
to UID/GID 65534 - nobody:nogroup
, and link
) on its internal temporary files (ld.so.conf
), reporting “Invalid argument” errors and “missing cap_chown?”.
Hypothesis:
This strongly suggests a very deep-seated issue with the underlying VPS kernel or Docker environment itself, potentially due to highly restrictive host security policies (like AppArmor/SELinux profiles that interfere with low-level syscalls) or a non-standard virtualization layer preventing even privileged containers from performing these operations.
Question:
Has anyone encountered similar Invalid argument
errors for chown
or link
operations inside a Collabora Docker container, especially when running on a VPS and even with privileged: true
? Any insights or suggestions for diagnosing/resolving host-level kernel/Docker interactions would be greatly appreciated.