Elastic Cloud on latest 7.3.x release and working with Ops to get APM agents installed in production. We have some questions we want to cover before flipping the switch regarding resource usage and performance impact of APM.
Similar to: APM Time Complexity Questions
Support case: #00389518
Questions:
-
For the Java and .NET agents, what is the memory overhead of adding APM (i.e. a VM running Tomcat has 2 CPU, 3GB RAM) - how much will the agent consume if added?
-
If for some reason Elastic Cloud is down, will APM agents buffer metrics in memory until cloud back up to push to Elasticsearch?
-
If APM agents do buffer metrics before shipping to Elasticsearch, how much memory will they consume or is there a cutoff?
-
On average what is the network impact or throughput for APM metrics being shipped to Elastic Cloud? (payload side and number of requests per page or servlet/endpoint interaction)
Anything else you can think of that would appease Operations team inquiries and get comfortable about installing APM in production is appreciated.