Does Rally 2.7.0 support Telemetry Devices? If so is the expectation that all tracks are supported? Meaning you can run any telemetry devices listed under 'esrally list telemetry'. (the special conditions around jit , gc , jfr stats notwithstanding)
The example in the rally documentation calls out specifically "esrally race --track=pmc --telemetry jfr". Given there are more than just the 'jfr' telemetry devices available and there are other tracks beyond 'pmc' I figured it would work with others. However I have not been able to get any results generated with say "esrally race --track=nyc_taxis --telemetry node-stats" for instance.
Before I start devoting time digging into what might be going on, I figured I would check first.
All the *-stats telemetry devices and node-stats specifically records metrics and sends them to metric store. So if you haven't already, please make sure you set up a dedicated metrics store (docs link). You will then find captured node stats in the rally-metrics-* indices in documents where "name" is "node-stats".
Does Rally 2.7.0 support Telemetry Devices? If so is the expectation that all tracks are supported? Meaning you can run any telemetry devices listed under 'esrally list telemetry'.
For the most part, yes. Some telemetry devices might not work with older versions of Elasticsearch if an API wasn't available (e.g. the disk-usage-stats telemetry device won't work on clusters prior to <7.15 because the API doesn't exist).
The example in the rally documentation calls out specifically "esrally race --track=pmc --telemetry jfr". Given there are more than just the 'jfr' telemetry devices available and there are other tracks beyond 'pmc' I figured it would work with others. However I have not been able to get any results generated with say "esrally race --track=nyc_taxis --telemetry node-stats" for instance.
By default Rally uses an 'in-memory' metrics store to collect and store race metrics. The results summary presented at the end of a race is calculated off of these metrics stored in-memory. We don't currently summarise the metrics returned from node-stats, which is a roundabout way of saying that I don't think the docs make it clear that we assume you have a remote metrics store setup under [reporting] in ~/.rally/rally.ini in order to analyse the output of some telemetry devices (including node-stats).
Ah! Thank you for the clarification regarding the in-memory and remote metric store configurations. I will have a look at the documentation referenced regarding remote metric store set up.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.
