Disappearing /proc endpoints.

[Ben Green]

Hi all,

First off, thanks LXC. We have a very efficient set of servers running.

Trouble today though, /proc has gone away for all containers.

root@mycontainer:~# ls -lah /proc/
ls: cannot access '/proc/stat': Transport endpoint is not connected
ls: cannot access '/proc/swaps': Transport endpoint is not connected
ls: cannot access '/proc/uptime': Transport endpoint is not connected
ls: cannot access '/proc/cpuinfo': Transport endpoint is not connected
ls: cannot access '/proc/meminfo': Transport endpoint is not connected
ls: cannot access '/proc/diskstats': Transport endpoint is not connected

Is there a way I can reconnect the endpoint?

When I restart containers lxcfs is re-attached, so it lxcfs does appear
to be running correctly? Could this have been from a restart of lxcfs? I
can't see one having happened.

Is there a fix I can apply to remount lxcfs on the containers?

--
Cheers,
Ben Green

[Narcis Garcia]

Host OS? Architecture?
LXC version? Privileged? Unprivileged? LXD involved?
Guest OS? Architecture?



Narcis Garcia

__________
I'm using this dedicated address because personal addresses aren't
masked enough at this mail public archive. Public archive administrator
should fix this against automated addresses collectors.
El 27/3/22 a les 13:47, Ben Green ha escrit:

[Andreas Laut]

Hi, we recently have the same happening.
The reason was a update für lxcfs on Debian 10. (don't know if your
system is Debian 10?)
After the update lxcfs restarted and all the proc endpoints got
disconnected.

So after that we set lxcfs on hold with 'apt-mark hold lxcfs' so this
didn't occour again for us.
I don't know why this update was released, the timestamp for the update
is from last year:
https://metadata.ftp-master.debian.org/changelogs//main/l/lxcfs/lxcfs_3.0.3-2+deb10u1_changelog

Do you have a debian 10 host or some other distri where this happend?
Kind regards


Am 27.03.22 um 13:47 schrieb Ben Green:

[Ben Green]

Yes, this was exactly the reason it turns out. We too will hold lxcfs.

What I'd like to be able to do though is reconnect those /proc entries
to lxcfs rather than having to restart all our containers.

Cheers,
Ben

--
Cheers,
Ben Green

[Andreas Laut]

https://github.com/lxc/lxcfs/issues/193#issuecomment-309370327
The lxcfs maintainer didn't get this memo.

If you have docker running:
https://github.com/xigang/lxcfs-admission-webhook/blob/dev/script/container_remount_lxcfs.sh
But if you have bare lxc running and mount is forbidden (like the
default in debian) than no, I don't think you can remount them without
restarting lxc. You need the permission to mount inside the lxc and
normally it's good if you can't do it.

Regards,
Andreas

Am 30.03.22 um 10:08 schrieb Ben Green:

[Ben Green]

Hi Andreas,

I have all possible capabilities dropped inside the containers, so it
looks like a reboot will be the only way.

Thaks for all the info, greatly appreaciated.

Cheers,
Ben