Skip to main content
Feedback

Rolling restart of runtime clusters and clouds

You can modify the rolling restart behavior on runtime clusters or runtime clouds.

If you restart all the nodes in a runtime cluster or runtime cloud at the same time, the runtime cluster or runtime cloud becomes temporarily unavailable. By default, attempts to restart the nodes in a runtime cluster or runtime cloud on a rolling basis. During a rolling restart, the following occurs:

  • The nodes within a runtime cluster or runtime cloud take turns restarting. While restarting, a node is unresponsive to requests.

  • Each node waits for any running processes to complete before restarting.

A rolling restart of runtime clusters and runtime clouds is performed in the following situations:

  • When a new release of is available (after June 2016).

  • When release updates to certain connectors force a runtime restart.

  • When you set or change a property on the Properties panel, click Save, and choose the restart now option.

  • When you click Restart Runtime on the Runtime Information panel of Runtime Management (Manage > Runtime Management).

  • When you use a Java monitoring tool such as Java VisualVM to manually restart a runtime cluster or runtime cloud and you select the rollingRestartCluster operation.

You can modify the rolling restart behavior by changing the following properties on the Advanced tab of the Properties panel:

  • Cluster Rolling Restart Percentage (com.boomi.container.cluster.rollingRestart.clusterRestartPercentage) controls what percentage of the nodes are restarted simultaneously. The default is 50%.

  • Rolling Restart Force Timeout (com.boomi.container.cluster.rollingRestart.forceRestartTimeout) controls how long a node will wait for restart to complete. After this time, the node begins restarting regardless of the status of any other nodes. The default is 30 minutes.

  • Rolling Restart Next in Line Timeout (com.boomi.container.cluster.rollingRestart.nextInLineRestartTimeout) controls how long a node will wait for the "next in line" node to complete the restart process and return to RUNNING status. The default is 5 minutes.

note

For some types of release updates, the update process is not compatible with earlier releases of . For those updates, all nodes must be restarted simultaneously; a rolling restart of runtime clusters and runtime clouds cannot be performed. Release updates that are not backwards-compatible are infrequent, and provides advance notice of such releases.