Published: October 27, 2017
In Platform version 8.0.0, an out-of-memory error on a Linux node could cause the Apprenda Logstash service to be repeatedly killed by the node’s out-of-memory killer. The node will then fail the Platform’s Logstash health check, because the Logstash service is not running. After failing the heath check, your Platform will transition the node into the Reserved State, preventing future deployments to the node and possibly causing resourcing problems for your Linux workloads on the Platform.
Until the out-of-memory error is resolved, the Platform will continue to transition the node into Reserved as the node fails health checks, even if you manually move the server out of the Reserved state.
To prevent this issue, lower the out-of-memory score of the Logstash service on your Platform’s Linux nodes so the out-of-memory killer will kill processes with higher scores first if a node starts to run out of memory. This will only help if the root cause of your Linux nodes being placed into Reserved is an out-of-memory error.
Contact your support representative for more information about lowering the out-of-memory score of the Logstash service or if you are having trouble diagnosing cause of a Linux node issue, contact your support representative.