Skip to end of metadata
Go to start of metadata

Remote agents periodically send a "heartbeat" signal to the Bamboo server. This is vital for tracking whether your remote agents are online or offline. The remote heartbeat is asynchronous, which means that if a remote agent goes offline and comes back online again it will reconnect instead of being shut down (as long as the same server is available).

However, you may wish to adjust the time parameters for the remote agent heartbeat, particularly if you have a lot of network activity already.

Icon

You need to be running Bamboo 2.0.6 or above to adjust the following remote agent heartbeat parameters.

There are three configurable parameters on the bamboo server for the remote agent heartbeat:

  • bamboo.agent.heartbeatInterval — This parameter governs the frequency of the heartbeat signal from the remote agents. This parameter is specified in seconds with the default being 5 seconds. (For Bamboo 5.x, the default value is 60 seconds.)
  • bamboo.agent.heartbeatTimeoutSeconds — This parameter governs how long the Bamboo server will wait before it times out an agent that it hasn't received a heartbeat signal from. A remote agent that has been timed out will be marked as 'Offline'. Any builds being run by agents which have timed out will be abandoned. This parameter is specified in seconds with the default being 600 seconds.
  • bamboo.agent.heartbeatCheckInterval — This parameter governs how often Bamboo checks for agents that have exceeded the heartbeat timeout specified in bamboo.agent.heartbeatTimeoutSeconds. This parameter is specified in seconds with the default being 30 seconds.

Please read the Configuring system properties page for instructions on how to change a remote agent heartbeat parameter for your Bamboo server.

2 Comments

  1. http://forums.atlassian.com/thread.jspa?messageID=257243188&#257243188

    I'd like to see this page updated with respect to adding more information regarding the wrapper timeout per agent vs the global settings here. 

    Here's my sequence of events:
    1. Remote JVM crashed and restarted with network message "NoRouteToHost". Is this a bug? The JVM should never crash?
    2. wrapper ping timed out.
    3. wrapper marked build offline because JVM process it was monitoring went away.
    4. server marked build offline for time out because wrapper marked JVM offline for time out.

    The end result was that the Bamboo server reporting time out. The was due to remote agent wrapper reporting time out and not JVM failure. Seems to me there should be some checking of the process still running by the wrapper and a crashed JVM message showing up on the Bamboo server rather than time out message.

    Feel free to move this to the forums or a bug if appropriate.

    The error message on the Bamboo server was "Agent 'onzo-1' went offline while building APTCOVERITY-HEADNIGHTLYX8664PCLINUX-JOB1-5. The build results did not finish."

    The agent didn't go offline, it restarted. ideally this message would state agent restarted due to JVM crashing.