ceph smoketest jobs do not honor timeout

[Jean-Baptiste Lallement]

The ceph smoketest jobs do not time out and ran for more than 2 days despite timeout is defined in ts_control and tc_control files

for example:
http://10.98.0.1:8080/job/raring-server-i386-smoke-ceph/12/
http://10.98.0.1:8080/job/raring-server-amd64-smoke-ceph/14/

[Max Brustkern (nuclearbob)]

I'm looking into this. On the test run I ran to recreate it, the UTAH state file indicated all tests were done executing within a few hours, but the process didn't return for some reason. I'll continue dissecting the process to try to ascertain why.

[Max Brustkern (nuclearbob)]

When I logged into the machine to run the testsuite, it returned within several hours. I'm getting a more exact time so I can determine if the SSH session is timing out.

[Joe Talbott (joetalbott)]

http://10.98.0.1:8080/view/Raring/view/Smoke%20Testing/job/raring-server-amd64-smoke-ceph/31/console is an example of this happening too. Paul Larson logged into the console and found that utah was still running and strace showed:

write(1, "ved\nPackage 'libapache2-mod-"..., 4096^C<unfinished...>

I think this is related to the truncated logs issue we have. It looks like we might be filling up the stdout buffer space or something.

[Joe Talbott (joetalbott)]

http://thraxil.org/users/anders/posts/2008/03/13/Subprocess-Hanging-PIPE-is-your-enemy/ seems to be the issue we're encountering.

[Max Brustkern (nuclearbob)]

I think we've fixed this. Is it still occurring?