Sematext Scalable Performance Monitoring is brand new and we are
working on gathering frequently asked questions - let us know if you
Q: What should I do if I can't find the answer to my question in the FAQ?
Q: Do I need to enable JMX in ElasticSearch?
A: No, SPM for ElasticSearch gets ElasticSearch statistics via its HTTP API, not JMX.
Q: Do I need to add the SPM agent to all ElasticSearch nodes?
A: Yes, the SPM agent needs to be added to
the JVM command like for all ElasticSearch nodes. This is required
because each SPM agent instance collects only local node's
ElasticSearch and JVM metrics. While the SPM agent could gather
statistics for all ElasticSearch nodes and their JVMs from a
single ElasticSearch node, such a setup would suffer from having a
Single Point of Failure (SPoF) - if the node with SPM agent were
to die, all ElasticSearch and JVM metrics for all cluster nodes
would immediately stop being collected. Having SPM agent running
on all ElasticSearch nodes thus provides for a more resilient
monitoring system - an important property.
Q: Do I need to add a separate System for each ElasticSearch server/node I want to monitor?
A: No, one System is enough. Think of an SPM
"System" as a "ElasticSearch Cluster". Thus, to monitor N ElasticSearch servers/nodes that
belong to the same cluster you would create just a single SPM
System and use its Token in SPM configuration file on all ElasticSearch
servers that are a part of this cluster.
Q: Do I need to assign fixed names to ElasticSearch nodes? How?
A: Yes. By default ElasticSearch assigns
random names to nodes, but when using SPM it is best to assign
each node a fixed name so that metrics can be permanently
associated with it and so that metrics filtering can be performed.
Simply assign each node a name in elasticsearch.yml