0.7.0-rc5: lxc-start: No such file or directory - stat(/proc/<pid>/root/dev//console)

[dudebout]

I updated the Arch Linux package to the latest 0.7 rc and running containers does not work anymore.

I can import images but build or run do not work and give the following error:

$ docker run -i -t debian /bin/bash
lxc-start: No such file or directory - stat(/proc/12296/root/dev//console)
2013/11/17 19:26:00 Unable to locate /bin/bash

The error is the same one, no matter what the command is.

[shykes]

Nicolas, are you trying to use containers and images from previous RCs? These will not work as the storage format in rc1-4 was not yet stable and will not be auto-migrated.

​

@solomonstre
@docker

On Sun, Nov 17, 2013 at 11:29 AM, Nicolas Dudebout
notifications@github.com wrote:

I updated the Arch Linux package to the latest 0.7 rc and running containers does not work anymore.
I can import images but build or run do not work and give the following error:

$ docker run -i -t debian /bin/bash
lxc-start: No such file or directory - stat(/proc/12296/root/dev//console)
2013/11/17 19:26:00 Unable to locate /bin/bash

The error is the same one, no matter what the command is.

Reply to this email directly or view it on GitHub:
#2739

[dudebout]

I had that problem and figured out as much.

However, I think the problem I experienced had to do with the way I created my debian image. I'll try and figure it out.

[dudebout]

I regenerated my image and the error vanished. Thanks @shykes for the prompt answer.

[codeaholics]

We're still seeing this issue on a completely clean 0.7.0 install.

[cressie176]

The issue resolved after rebooting the host

[cressie176]

And it's back again

[cressie176]

I set a repeating CI build off yesterday with the following sequence...

1. build on CI
2. push to private registry
3. ssh to production server
4. pull from private registry
5. run

It got through 58 builds, but then started failing during the pull (step 4).

[Step 4/7] Step 10 : RUN mkdir -p /var/log/myapp
[18:48:06][Step 4/7]  ---> Running in 31e2eb08c7e8
[18:48:11][Step 4/7] lxc-start: No such file or directory - stat(/proc/2695/root/dev//console)
[18:48:11][Step 4/7] Error build: The container failed to start. Unkown error

I think root cause is this bug in lxc, which causes the docker build to fail. However I suspect docker doesn't return a non zero exit code, so the build continued. Somehow docker push still pushes the image, but presumably doesn't work properly, so the pull ends up failing with

Tag 59 not found in repository myregistry:5000/dummy/myapplication

Apparently the lxc bug has been fixed (requires a kernel upgrade too), but it would be good if docker build returned a non zero status code...

lxc-start: No such file or directory - stat(/proc/31745/root/dev//console)
Error build: The container failed to start due to timed out.
$ echo $?
0

[rawouter]

Can you tell more about the kernel upgrade needed? We still face this issue in 0.7 as well, and I believe we run latest lxc version that has the lxc bug fix.

[cressie176]

We tried the kernel upgrade (to 3.8.0-33-generic #48-Ubuntu) + lxc 1.0.0 alpha we still got the issue so reverted back to 0.9.0 :(

[prologic]

https://gist.github.com/therealprologic/8909525

^^^ I've been experiencing the same issue as well. Trying to read this Issue carefully to identify any "quick" work-arounds. Hopefully the Docker 0.9+ Release will fix this? :)

[prologic]

To summarize (correct me if I'm wrong):

• Docker currently has a HARD requirement on lxc 0.8
• lxc >= 0.8 has some bugs/issues that can cause issues like this one
• A temporary work-around would be to downgrade lxc to 0.8

[tianon]

Yes @therealprologic, the best solution currently is to use LXC 0.8.

[prologic]

@tianon I'll do some testing here locally on my desktop which is CRUX 3.0 +
Docker 0.8 + lxc 0.9.0

I'm basically uncertain as to whether I should provide a lxc 0.8 port for
CRUX or not :) (I'll do some testing on this issue).

James Mills / prologic

E: prologic@shortcircuit.net.au
W: prologic.shortcircuit.net.au

On Tue, Feb 11, 2014 at 1:17 AM, Tianon Gravi notifications@github.comwrote:

Yes @therealprologic https://github.com/therealprologic, the best
solution currently is to use LXC 0.8.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/2739#issuecomment-34642633
.

[prologic]

@tianon I performed the exact same steps on my CRUX 3.0 + Docker 0.8 + lxc 0.9 on a btrfs graph/storage just now and it worked perfectly. This only sems to fail on Ubuntu 13.04 + Docker 0.7.6 + lxc 0.9 on an AUFS graph/storage.

Does this make sense to anyone?

[prologic]

@tianon FYI: I was not able to reproduce this on CRUX 3.0 + lxc 0.9 + docker 0.8 on a btrfs backend/storage.