ZFS storage volume id mapping no longer works

Project Discussion LXD Support
1

ubuntu 26.04

zfs 2.4

lxd snap 5.21.4

After restarting a freshly recovered LXD installation, a ZFS storage volume containing a Postgresql 13 cluster is mounted with user/group 1000105/1000109 instead of 105/109.

Setting security.shifted=true did not change anything. All other LXD settings are at their default values.

Container raw.idmap is empty.

/etc/subuid and /etc/subgid both contain:
ladmin: 100000:65536

The recovery was made using lxd recover without a database.

The prior installation was running on ubuntu 24.04.

Thanks in advance for any clues

Best
Philippe

2

@sdeziel1 does this sound similar to `lxd recover` should better handle containers with `security.idmap.isolated=true` · Issue #13463 · canonical/lxd · GitHub ?

1 Like
3

Yes, that sounds in the same vein indeed.

1 Like
4

Should I try setting security.idmap.isolated=true? Right now it’s not set.

5

I went ahead and set it. Now the volume is mounted with nobody/nogroup ownership. Unsetting returns me to 1000105/1000109.

I’m not sure I’m in a better place now.

6

For what it’s worth, on the LXD host, the uid/gid numbers started at 2000000 instead of 1000000 as configured on the containers. So I changed owners on the containers affected using the correct values and the uid/gid’s on the host renumbered automatically starting at 1000000 as expected.

All is well.
Thanks

7

This topic was automatically closed 24 hours after the last reply. New replies are no longer allowed.