KB-2163 High heap usage of search server component

Symptoms

On an Appian 20.2 or 20.3 environment, functionality depending on the search server is negatively impacted. Symptoms are varied but include degraded performance, inability to upload documents, and/or issues deploying packages.

To validate whether the symptoms are a result of this issue, navigate to the search-server.log file in <APPIAN_HOME>/logs/search-server and confirm whether output similar to the following is observed:

[DEBUG][org.elasticsearch.action.admin.cluster.node.stats.TransportNodesStatsAction] [%<node_name>] failed to execute on node [<ID>]
org.elasticsearch.transport.RemoteTransportException: [Node <server>:<port>][<IP>:<port>][cluster:monitor/nodes/stats[n]]
Caused by: org.elasticsearch.common.breaker.CircuitBreakingException: [parent] Data too large, data for [cluster:monitor/nodes/stats[n]] would be [1036786612/988.7mb], which is larger than the limit of [986061209/940.3mb], real usage: [1036785320/988.7mb], new bytes reserved: [1292/1.2kb], usages [request=0/0b, fielddata=0/0b, in_flight_requests=1292/1.2kb, accounting=1887742/1.8mb]

Note: The values (e.g. [1036786612/988.7mb]) from the log output above are simply an example and may differ from the output observed in the logs for your environment. However, the values will likely be close to 1 GB (>900 MB) if using the default configurations.

Cause

This issue has been hotfixed via AN-170688 in Appian 20.3.205.0 and 20.2.235.0.

Action

Apply the latest hotfix to your Appian installation.

Workaround

If you are actively experiencing this issue and unable to take an outage to apply the hotfix at this time, the search server component may be restarted as a workaround. There is no need to shutdown the application server or take an outage to apply this workaround. On highly available environments that have the search server running on three nodes, the search server can be restarted in a rolling fashion. Start with the node that has experienced the problem. Then, restart each node one by one after ensuring that the cluster status is green. The cluster status can be monitored via the search_server_cluster.csv log in <APPIAN_HOME>/logs/data-metrics.

If this issue is being encountered frequently despite the search server being restarted, increase the search server heap allocation from the default 1 GB. Instructions to increase the search server heap can be found here. Note that increasing the search server heap allocation can result in higher overall system memory usage. Ensure there is enough system memory available before increasing heap. If you are unsure how much to increase the heap usage by, a safe option is increasing it to 2048 MB and monitoring the situation.

Affected Versions

This article applies to Appian 20.2 and 20.3.

Last Reviewed: November 2020