We are using upsource 2.6 (soon planning to upgrade 3.0.1) as a code review tool. Frequently we experience issues where developers aren't able to load the review page. Generally we aren't able to reproduce such issue since the time instance already passes away and the issue certainly depend on various runtime parameters including the number of active users on the system.
Below are the server stats
CPU Cores & Clock Speed - 4, 2.4 GHz
RAM - 16 GB
Max Active users - 60-70
Upsource FrontEnd & Cassandra JVM is running with 3 GB heap size
Upsource Analyzer JVM is running with 2 GB heap size
Upsource Monitoring JVM is running with 1 GB heap size
We tried creating a load test script with JMeter to simulate and reproduce the issue but it was taking quite some time and effort. I would appreciate if you could share a parameterized load test script that could be executed at our end. If not,
1. What suggestions do we have in troubleshooting such performance issues?
2. What are the different ways of monitoring the upsource system (especially the jvm applications)? We could monitor the infrastructure metrics through open source monitoring systems like riemann/graphite but can upsource jvms send their metrics to these systems? If not, what are the alternatives?