About JMeter Distributed configuration for load testing (not in cloud),
I can setup X JMeter masters in different machines and execute them with shared files using shared folder(s).
The benefits are:
Each master is oblivious to other and can be shutdown and start when needed with dynamic/different properties.
Each master have its own logs and results that can be explored separately.
I don't need network connection between JMeter masters' machines.
What are the benefits for using master-slave configuration in such case? It seems like an unnecessary overhead when focusing on load test.
The benefits are:
- centralization of results on 1 node (master), you can follow results in Summarizer from the master node, you have the CSV/XML file generated there and you can generate the web report at end of test using this
- centralization of jmx plan on 1 node (master)
- synchronization of the test from master, ie the master will start/stop the test from master
Besides the drawbacks you describe there are:
- network configuration complexity
- need to deploy csv on each node (although there are options with plugins (redis, simple table server)
- network traffic between nodes and master
It was created at time deployment automation was not available through things like vagrant, ansible, cloud ...
