Moving this bug over to lxcfs.
** Package changed: lxc (Ubuntu) => lxcfs (Ubuntu)
--
You received this bug notification because you are a member of Ubuntu
Touch seeded packages, which is subscribed to lxc in Ubuntu.
https://bugs.launchpad.net/bugs/1514690
Title:
rebooting container with
Many thanks Serge for figuring that out!
--
You received this bug notification because you are a member of Ubuntu
Touch seeded packages, which is subscribed to lxc in Ubuntu.
https://bugs.launchpad.net/bugs/1514690
Title:
rebooting container with systemd >= 226 fails to create /lxc/adt-
** Changed in: lxc (Ubuntu)
Status: In Progress => Fix Committed
--
You received this bug notification because you are a member of Ubuntu
Touch seeded packages, which is subscribed to lxc in Ubuntu.
https://bugs.launchpad.net/bugs/1514690
Title:
rebooting container with systemd >= 226
Yup, just making lxc sleep 2 seconds works. So this is purely a race.
I'll fix it by having lxc excplicitly wait until the cgroups are removed
before re-execing after reboot.
** Changed in: lxc (Ubuntu)
Status: Triaged => In Progress
--
You received this bug notification because you
Actually that might not help. If the cgroup really did still exist,
then lxc would have refused to re-use it. I.e .instead of /lxc/x1, it
would then use /lxc/x1-1.
--
You received this bug notification because you are a member of Ubuntu
Touch seeded packages, which is subscribed to lxc in
So the problem is indeed that the cgroups are deleted using the
cgroupfs, and fuse doesn't get that information. The fuse kernel module
is caching the information for one second before re-querying userspace.
I've pushed a patch to lxcfs to drop the caching to a half second. I'm
posting a patch
That was a red herring, actually. The cause of failure appears to be
the next line.
After fixing that so that the mkdir succeeds, it still fails on
Failed to allocate manager object: No such file or directory
--
You received this bug notification because you are a member of Ubuntu
Touch
** Changed in: lxc (Ubuntu)
Assignee: (unassigned) => Serge Hallyn (serge-hallyn)
** Changed in: lxc (Ubuntu)
Status: New => Triaged
--
You received this bug notification because you are a member of Ubuntu
Touch seeded packages, which is subscribed to lxc in Ubuntu.
With cgfs i actually wasn't able to start the container in the first
place. I've now fixed at least that.
it wasn't the mkdir which was failing, but the subsequent attempt to
attach itself to it. That's because while cgmanager used to chown the
child files after creating a directory for us,
Doh', it's because I had a total brainfart while writing that.
--
You received this bug notification because you are a member of Ubuntu
Touch seeded packages, which is subscribed to lxc in Ubuntu.
https://bugs.launchpad.net/bugs/1514690
Title:
rebooting container with systemd >= 226 fails to
When I test this using cgfs-backed lxcfs, the mkdir of init.cgroup fails after
setresuid(10, 10, 0).
This is odd since doing it manually using sudo -u \#10 -g \#10 mkdir
/sys/fs/cgroup/systemd/lxc/x1/x works fine.
--
You received this bug notification because you are a member
I believe what's happening is that systemd tries to mkdir init.scope
before the directory has been deleted and gets -EEXIST; then the kernel
finishes deleting it, then systemd tries to move itself to it but it is
gone.
Waiting for one more debugging build to verify.
--
You received this bug
I'm setting this to high as this breaks autopkgtesting on armhf quite
badly.
** Changed in: lxc (Ubuntu)
Importance: Undecided => High
--
You received this bug notification because you are a member of Ubuntu
Touch seeded packages, which is subscribed to lxc in Ubuntu.
FTR, same behaviour with Serge's lxcfs "testing" branch with cgfs.
--
You received this bug notification because you are a member of Ubuntu
Touch seeded packages, which is subscribed to lxc in Ubuntu.
https://bugs.launchpad.net/bugs/1514690
Title:
rebooting container with systemd >= 226 fails
14 matches
Mail list logo