[lxc-users] /proc lost in some containers

[Tomasz Chmielewski]

LXD 3.6 from a snap on an up-to-date Ubuntu 18.04 server:

lxd 3.6 9510 stable canonical✓ -


Suddenly, some (but not all) containers lost their /proc filesystem:

# ps auxf
Error: /proc must be mounted
To mount /proc at boot you need an /etc/fstab line like:
proc /proc proc defaults
In the meantime, run "mount proc /proc -t proc"

#


I think I've seen something similar like this in the past.
Can it be attributed to some not-so-well automatic snap upgrades?



Tomasz Chmielewski
https://lxadm.com

[Stéphane Graber]

lxd 3.6 9510 stable canonical✓ -
# ps auxf
Error: /proc must be mounted
proc /proc proc defaults
In the meantime, run "mount proc /proc -t proc"
#
I think I've seen something similar like this in the past.
Can it be attributed to some not-so-well automatic snap upgrades?

That can either be a lxcfs crash or lxcfs bug of some kind.

Can you show "ps fauxww | grep lxcfs" on the host.

And then inside an affected container run "grep lxcfs /proc/mounts" and
for each of the /proc path listed, attempted to read them with "cat",
that should let you check if it's just one lxcfs file that's broken or
all of them.

[Tomasz Chmielewski]

Post by Stéphane Graber

Post by Tomasz Chmielewski
lxd 3.6 9510 stable canonical✓ -
# ps auxf
Error: /proc must be mounted
proc /proc proc defaults
In the meantime, run "mount proc /proc -t proc"
#
I think I've seen something similar like this in the past.
Can it be attributed to some not-so-well automatic snap upgrades?

That can either be a lxcfs crash or lxcfs bug of some kind.
Can you show "ps fauxww | grep lxcfs" on the host.

Hmm, running twice?

root 85202 0.0 0.0 382524 1236 ? S<l Sep24 9:37 lxcfs
/var/snap/lxd/common/var/lib/lxcfs -p /var/snap/lxd/common/lxcfs.pid
root 19414 0.0 0.0 530260 1716 ? S<l Oct05 3:03 lxcfs
/var/snap/lxd/common/var/lib/lxcfs -p /var/snap/lxd/common/lxcfs.pid

Post by Stéphane Graber
And then inside an affected container run "grep lxcfs /proc/mounts" and
for each of the /proc path listed, attempted to read them with "cat",
that should let you check if it's just one lxcfs file that's broken or
all of them.

Unfortunately I've already stopped / started all affected containers.


Tomasz Chmielewski
https://lxadm.com