Over time, we're seeing longer and longer pauses between Bamboo plan runs, to the point that they're not keeping up with demand. In our scenario, a small plan runs on a remote agent once for each Perforce check-in, acting like a pre-processor for other build plans, at a rate of a few check-ins per minute, occasionally up to 20 check-ins per minute. The duration of our jobs in the plan is about 8 to 10 seconds, so they do queue up when there are bursts of check-ins. But the problem is over time, thousands of builds, there's a growing pause between plan runs, and even within jobs of the plan. Eventually it's only running once per couple minutes, and at that pace, it's not keeping up with an ever-growing queue of check-ins. Restarting the server (5.4.1) seems to help, but the pauses eventually come back. Any ideas?
With debug logging on the server and remote agent during the slowed builds, the agent is just sitting idle waiting for the server, and the server isn't working on anything either. We see a number of Quartz jobs starting and expiring, and message consumers polling for messages:
2015-02-10 11:28:46,919 DEBUG [bambooAgentMessageListenerConnector-1] [FingerprintMatchingMessageListenerContainer] Consumer [PooledMessageConsumer { ActiveMQMessageConsumer { value=ID:remoteagent.ourcompany.com-39415-1423368652369-3:1:1:1, started=true } }] of session [PooledSession { ActiveMQSession {id=ID:remoteagent.ourcompany.com-39415-1423368652369-3:1:1,started=true} }] did not receive a message
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.