[Pascal Quantin <pascal.quantin@xxxxxxxxx>]

2015-07-08 23:28 GMT+02:00 Gerald Combs <gerald@xxxxxxxxxxxxx>:

On 7/8/15 1:46 PM, Hadriel Kaplan wrote:
> Anyone seen this type of petri-dish failure before, on the Windows petri-dish, and know what it means?
>
> Resolving deltas:  71% (3551/4999)
> Resolving deltas:  72% (3600/4999)
> error: index-pack died of signal 11
> fatal: index-pack failed
>
> program finished with exit code 128
> elapsedTime=22.854000

It started showing up after the Gerrit 2.10 upgrade. As far as I can tell it's due to a combination of the following:

- Slow initial I/O speeds on AWS EC2[1].
- Git on Windows not being particularly fast.
- Gerrit 2.10 timing out HTTP connections after 30 seconds.
- The "retry" flag in Buildbot's Git step not doing much of anything.

Until one of these is fixed we have a build step named "prefetching git to work around Gerrit timeout and Buildbot retry bugs", which should trigger enough I/O reads to work around the timeout and retry issues. Unfortunately it's not guaranteed to work every time.

[1] http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ebs-prewarm.html

Which Git version is installed on the Petri Dish machine? Version 1.9.5?

If yes, would it be worth giving a try with the upcoming Git For Windows release found here and based on Msys2 and Git 2.4.5? https://github.com/git-for-windows/git/releases