You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Running a shell and allowing it to idle makes shell unresponsive #561
Running a shell like bash using fig run test sh and then allowing it to idle for a few minutes will make the shell unresponsive. Typing any command will produce no results.
I recently updated to Fig 1.0.0 and I'm running on and up-to-date Arch Linux.
Here is my fig.yml
test:
image: busybox
command: echo "Hello"
And the command I ran is fig run test sh then I waited for a few minutes and tried to run ls in the container. The shell initially works, so you have to idle for some amount of time.
My particular use case is running Guard within a box setup as a Rails application container. The box stops responding and is left running even after I stopped the other ones. The only way to get rid of the stray box(es) is to restart the docker service.
I don't recall seeing it before 1.0 (and docker 1.3), but I rarely run interactive, so I might have just missed it.
I was able to Ctrl-C, and then docker attach to the container again. I believe it's just the tty interaction that is having issues, so it could be dockerpty related.
@fredjean the issue you're describing sounds a lot more severe then what I ran into.
I experimented with this a bit. The interactive shell goes away at 60s. I'm pretty confident this is related to this change: d11wtq/dockerpty@6ded926
Someone else has reported problems with it as well. When I use fig 1.0 without the vendored dockerpty, and a dockerpty==0.2.3 (before that change), it works fine.
I think the change to which streams are flushed needs to be reverted.
I guess I have the same problem. Using docker 1.3 and fig 1.0. In my case I'm using bash or zsh and after n minutes the shell becomes unresponsive. I have to stop that container with ctrl + d or ctrl + c and start a new one. However If I type fig ps I see a lot of containers running with different sequence numbers:
mycont_run_319 up
mycont_run_320 up
mycont_run_321 up
Activity
dnephin commentedon Oct 20, 2014
I ran into this yesterday as well
aanand commentedon Oct 20, 2014
Did you experience this error before 1.0? I wonder if it's TLS-specific.
fredjean commentedon Oct 20, 2014
+1 using 1.0.
My particular use case is running Guard within a box setup as a Rails application container. The box stops responding and is left running even after I stopped the other ones. The only way to get rid of the stray box(es) is to restart the docker service.
dnephin commentedon Oct 20, 2014
I don't recall seeing it before 1.0 (and docker 1.3), but I rarely run interactive, so I might have just missed it.
I was able to Ctrl-C, and then docker attach to the container again. I believe it's just the tty interaction that is having issues, so it could be dockerpty related.
@fredjean the issue you're describing sounds a lot more severe then what I ran into.
dmp1ce commentedon Oct 20, 2014
I didn't experience with 0.5.2 and I couldn't reproduce with docker alone.
dnephin commentedon Oct 21, 2014
I experimented with this a bit. The interactive shell goes away at 60s. I'm pretty confident this is related to this change: d11wtq/dockerpty@6ded926
Someone else has reported problems with it as well. When I use fig 1.0 without the vendored dockerpty, and a dockerpty==0.2.3 (before that change), it works fine.
I think the change to which streams are flushed needs to be reverted.
fredjean commentedon Oct 21, 2014
I was also able to use docker attach to connect to a stray fig container and shut it down cleanly. Many thanks to @dnephin for the tip!
andres-root commentedon Oct 21, 2014
I guess I have the same problem. Using docker 1.3 and fig 1.0. In my case I'm using bash or zsh and after n minutes the shell becomes unresponsive. I have to stop that container with ctrl + d or ctrl + c and start a new one. However If I type fig ps I see a lot of containers running with different sequence numbers:
So now I have up to 350 containers running.
dnephin commentedon Oct 26, 2014
I guess we need to update the vendored copy of dockerpty, or release it to pickup this change.
bfirsh commentedon Oct 27, 2014
#588 :)
thomasboyt commentedon Oct 29, 2014
Can this issue be reopened, since the bug hasn't been fixed yet? Almost opened a duplicate before I thought to look in the "closed" tab...
djmaze commentedon Nov 3, 2014
+1 for reopening. Fig is currently almost unusable for me because of this bug.
andres-root commentedon Nov 4, 2014
+1 fo reopening. I'm not using the new version because of this bug.
aanand commentedon Nov 4, 2014
I've just merged #588, which fixes it for real.
andres-root commentedon Nov 4, 2014
Thank you very much @aanand