fig under jenkins problems #199

Closed
ccverak opened this Issue Apr 25, 2014 · 29 comments

Projects

None yet
@ccverak
ccverak commented Apr 25, 2014

It does not flushes logs and has encoding issues when running on jenkins

@neoziro
neoziro commented Jul 15, 2014

I have no stdout on Jenkins too, I don't see what is the problem, I tried with an ssh -T (no tty) and it works. Help!

@ccverak
ccverak commented Jul 16, 2014

Is there a way, you can use fig concurrently ie: on 2 parallel jenkins job builds? I'm having containers collision

@neoziro
neoziro commented Jul 16, 2014

No, there is no collision in my case.

@ccverak
ccverak commented Jul 16, 2014

Can you please give me a hand?, i have a common fig.yml with pg and rails
with a pg link how do you start it on jenkins builds? do you use fig up or
fig run?

On Wed, Jul 16, 2014 at 9:14 AM, Greg Bergé notifications@github.com
wrote:

No, there is no collision in my case.

Reply to this email directly or view it on GitHub
#199 (comment).

@neoziro
neoziro commented Jul 16, 2014

I use fig run and there is no stdout even if I redirect the output in a file.

@littleowllabs

I am having a similar problem concerning lack of stdout from using fig run within Jenkins - did either of you find a solution to this?

@d11wtq
Contributor
d11wtq commented Aug 13, 2014

I'm unable to replicate this, but if one of you could try running with export PYTHONUNBUFFERED=true before you run fig, then let me know if that fixes the issue, I'll have a better idea. My suspicion is that the writes are being buffered, though I thought using os.write() wouldn't do that.

@dnephin
Member
dnephin commented Aug 13, 2014

I'm actually able to reproduce this as well. Other fig commands (up, stop, rm, build) all have output, but the output from run is only the "Creating mysqld_1" message (of the linked service), and then it skips ahead to the output of the stop command.

10:09:33 test-mysql-testing runtests: commands[1] | fig run tester bash /code/run_test.sh
10:09:53 Creating mysqltesting_mysql_1...
10:10:46 test-mysql-testing runtests: commands[2] | fig stop
10:10:50 Stopping mysqltesting_mysql_1...

Running on my local desktop I get a bunch of output from the fig run command.

I don't think it's buffered because there is lots of output after the run command finishes.

@dnephin
Member
dnephin commented Aug 13, 2014

The only difference I can see is that this is the only place dockerpty is used. I think d11wtq/dockerpty#4 is the issue.

@dnephin
Member
dnephin commented Aug 13, 2014

I think I've got a possible fix, let's continue the discussion in the dockerpty issue.

@littleowllabs

Great, I'll take a look over there.

@ccverak ccverak closed this Aug 20, 2014
@tf
tf commented Sep 10, 2014

Has this been resolved yet? I'm only using fig and jenkins (no dockerpty) and also see no stdout of run.

@dnephin
Member
dnephin commented Sep 10, 2014

fig uses dockerpty for fig run and fig up, it is not resolved yet

@pirelenito

Shouldn't this issue be open? As stated by @dnephin it is still not working.

@pirelenito

Hi @macropin, I understand that the root cause is d11wtq/dockerpty#6, but still, shouldn't we close this issue only when that one is resolved and the dependency updated in fig?

@macropin

@pirelenito I'm not the maintainer for this project so it's not my call. The decision to keep it open depends entirely on the maintainers and fashion in which this issue tracker is used. Eg Service centric vs project management vs SDLC. In any case, on Github tickets are usually only kept open if there is action to be taken in the specific project and in this case it's an upstream issue so no specific action for the fig project, unless they wish to identify it as 'fixed' in the future.

@mattupstate

Anybody have a fix for this? I'm still experiencing this with 1.0. Patching just seems like a major hack, and dockerpty is now vendored, making it even more annoying, and makes a case for re-opening this issue.

@dnephin
Member
dnephin commented Nov 4, 2014

For anyone following along , fig 1.0.1 should fix this issue by using fig run -T

@spiddy spiddy referenced this issue in spiddy/dind-jenkins-slave Dec 3, 2014
Merged

upgrade to Fig 1.0.1 #1

@tobowers

I'm still seeing this with docker-compose 1.1.0 (well something similar). I see the first line of stdout, but not anything after the first newline:

sudo docker-compose run web /bin/bash -c 'export PYTHONUNBUFFERED=1 && echo "hi" && echo "hi" && echo "hi"'
Creating rspec19_db_1...
hi
# nothing after here
@mattupstate

@tobowers as am I

@tobowers

I just used -T and it seemed like it helped, but didn't fix it. Seems like there are a couple of outstanding issues from january: #658 kind of sums it up @mattupstate

@mattupstate

@tobowers Yeah, I've been aware of this for a little while now. -T never worked perfectly for me under Jenkins even after it was thought to be fixed.

@zwily
zwily commented Apr 14, 2015

This is also still not working for me. Based on what everyone's saying, it seems like the issue should still be open.

@dnephin dnephin reopened this Apr 15, 2015
@dnephin
Member
dnephin commented Apr 15, 2015

It does appear like our earlier attempt did not completely fix this issue.

@suryagaddipati

I am seeing this as well. It seems to dropping the output when there is a pause in the output.

@arthurk
arthurk commented May 5, 2015

same here, the issue is still there. I'm running compose 1.2.0 and Jenkins 1.612

@aanand
Contributor
aanand commented May 5, 2015

If anyone's having truncation issues, you could try the binaries built by @zwily which contain the fix from d11wtq/dockerpty#36:

#658 (comment)

@dnephin
Member
dnephin commented May 24, 2015

This fix has been merged, and we've bumped the version of dockerpty used by compose. This fix is available in master, and will be inlcuded in the 1.3.0 release.

@dnephin dnephin closed this May 24, 2015
@dnephin dnephin added this to the 1.3.0 milestone May 24, 2015
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment