Bamboo sometimes deletes git repository cache, causing builds to queue

[choover-kayak]

We have a pretty big git repository (12gb). A couple of times per week, our Bamboo server encounters a transient error while fetching from this repository, and deletes it. For example:

2014-10-10 15:45:21,370 WARN [Connection monitoring thread - f43779d7-35ef-4a01-bbd3-d46fb736b884] [SshProxyCommand] While connecting to [git@stash-server/10.82.0.208:7999]: Can't connect session, connectFuture is not connected...
2014-10-10 15:45:21,474 WARN [4-BAM::PlanExec:pool-8-thread-3] [GitRepository] Deleting directory /bamboo/home/xml-data/build-dir/_git-repositories-cache/9360a00282e07c71c10adb9b42b674a2f2237b1b

(The error is just as often "connection refused" or "closed unexpectedly")

After deleting the repo, Bamboo starts a fresh clone. Since our repo is large, a clone can take a long time - sometimes as long as an hour. While the re-cloning happens, no builds trigger, and any manual builds stay queued until the clone is complete.

Are there any configuration changes we can make to mitigate this problem? Why does Bamboo delete the git repository so aggressively? We don't see any corresponding errors in our Stash logs.

[Timothy Harris]

I know this doesn't help your specific issue now

But a 12Gb repository is insane and I can't help making a comment . 

Have you thought about splitting your repository?

Trying to work Agile with CI/CD principles will never work with repo that size...