[Buildbot-devel] Trying to reproduce FileUpload timeout fail (ticket #1854)

[Dustin J. Mitchell]

On Mon, Mar 26, 2012 at 10:02 PM, Juan Lasheras <juan.lasheras at gmail.com> wrote:
> Hi Folks, I am taking a look at http://trac.buildbot.net/ticket/1854 and
> trying to reproduce the original issue.
>
> I am running on a CentOS 6.1 (64bit) machine (python 2.6).
>
> I have setup a virtualenv (as prescribed
> at http://trac.buildbot.net/wiki/RunningBuildbotWithVirtualEnv) and have
> installed v0.8.6p1.
>
> I am using a simple builder with a FileUpload step. I have placed a very
> large (~ 3 GB) file in the slave working dir. When I initiate the build, I
> run "ifdown lo" to simulate losing network connectivity. This essentially
> halts the transfer and in 15 minutes the Step terminates with Exception
> status. From the web status you can see it has an interrupt logged and the
> err.html and err.txt containers read something like "Failure:
> twisted.internet.error.ConnectionLost: Connection to the other side was lost
> in a non-clean fashion."
>
> Here is what the master.cfg looks like ( http://pastebin.com/tsx18MsZ ).
>
> Furthermore, when I restore the network with "ifup lo" the slave comes back
> online and can successfully complete subsequent FileUpload builds.
>
> Can anyone else confirm my these results?

That bug was reported against 0.8.2, and it's certainly possible that
things have changed since then.

JP, do you have some suggestions for reproduction here?

Juan, you might try reproducing with 0.8.2 - you can get to that
version pretty easily with 'git co v0.8.2', or, of course, install it
in a separate virtualenv with pip.

Dustin