We have Stash 2.10.0 linked with Bamboo 5.3 for builds, both running as Windows services.
Build Plans in Bamboo are set to poll Stash every 5 minutes for changes, and build when necessary. This works great most of the time, except when Stash is down for backup (which I do by stopping the service and tar'ing up Stash home before restarting).
During the Stash downtime, Bamboo reports "Unable to detect changes" errors for each plan - not surprising since Stash is not responding.
So in an attempt to avoid these, I tried shutting the Bamboo server service down for the duration of the Stash backup, but I've found that this then causes the remote Build Agents to go offline and I have to manually restart them.
The only solution I've come up with to avoid error conditions is to switch the repository polling from "Every 5 minutes" to cron based , excluding the the backup period, such as "0 0/5 06-22 * * ?" (ok the backup doesn't take 8 hours!). But this will mean every plan firing off at the same second every 5 minutes unless I choose different second/minute combination for each plan, which is a pain as I'm trying to standardise as much configuration as possible between plans.
Is there any other solution that you can suggest which avoids explicit scheduling? Getting the remote agents to cope with server downtime would be preferable I think.
