linux docker image not starts on ZFS storage at Proxmox LXC

Question

linux docker image not starts on ZFS storage at Proxmox LXC

Closed this issue 2 months ago · 8 comments

Docker runs on ZFS backing filesystem type

Storage Driver: overlay2
  Backing Filesystem: zfs
  Supports d_type: true
  Using metacopy: false
  Native Overlay Diff: false
  userxattr: true

when trying to run image we are getting error docker: failed to register layer: ApplyLayer exit status 1 stdout: stderr: unlinkat /usr/share/build-essential: invalid argument.

when storage type

Storage Driver: overlay2
  Backing Filesystem: extfs
  Supports d_type: true
  Using metacopy: false
  Native Overlay Diff: false
  userxattr: true

image starts without issues

how to run it on zfs ?

Answer 1 · 2023-04-26T10:18:45.000Z

@Kramtoske could you provide the command line or compose file you use to start Orthanc ?

Answer 2 · 2023-04-26T10:20:16.000Z

@amazy simple docker run osimis/orthanc:23.4.0 works on extfs but not on zfs

Answer 3 · 2023-04-26T11:12:25.000Z

Thanks.

Note that we currently have no experience and no system with ZFS and this might take quite some time before we can reproduce, analyze and fix the issue.

@Kramtoske , can you try to run the base image ?

docker run -it osimis/orthanc-runner-base:bullseye-20230320-slim-stable bash

You should have the same error since the /usr/share/build-essential folder should be there as well.

From that point, with your experience, you might be able to suggest a patch to this dockerfile

Thanks in advance for your help

Alain.

Answer 4 · 2023-04-26T12:06:51.000Z

@amazy docker run -it osimis/orthanc-runner-base:bullseye-20230320-slim-stable bash started up without any error

same issue found there nextcloud/all-in-one#1490 (reply in thread)
when image created not on lxc it created UIDs out of allowed ranges for lxc, when docker created in lxc it defines uid/gid in scope. https://kcore.org/2022/02/05/lxc-subuid-subgid/

Answer 5 · 2023-04-28T09:07:28.000Z

As I understand it (but honestly, I don't understand much about it), there's nothing to do on our side but wait for a Docker patch or ask the user to remap ids. Is that correct ?

Answer 6 · 2023-04-28T09:24:00.000Z

there's theoretical chance that setting user would help for example what we used on meddream and not met issues on any platform where we ran docker image

ENV SERVICE_USER="meddream"
ENV SERVICE_USER_ID="10001"
ENV SERVICE_GROUP="meddream"
ENV SERVICE_GROUP_ID="10001"

RUN groupadd -r -g $SERVICE_GROUP_ID $SERVICE_GROUP \ 
    && useradd -r $SERVICE_USER -u $SERVICE_USER_ID -g $SERVICE_GROUP_ID \
    && chown -R $SERVICE_USER:$SERVICE_GROUP /opt/meddream \

WORKDIR /opt/meddream/
USER $SERVICE_USER

Answer 7 · 2023-04-28T09:37:51.000Z

@Kramtoske Could you try to do that in an image that you build and that derives from osimis/orthanc:23.4.0 ?

Answer 8 · 2024-07-03T14:45:50.000Z

Closing issue due to inactivity