How to address the Skipped Pure Path Issue? we are using Version 7, during our load test we are seeing lot of pure path is getting skipped. We tried the support ticket solution by setting Send & Receive TCP buffer setup & also changed from Medium size to Large size. Still we are seeing this issue. 1 sec transaction is not crossing beyond 1500. One thing we observed in the Collector Size Health dashboard, the Buffer Saturation chart is showing 100% and its staying 100%. Please advice if any one faced the same issue and a way to resolve this problem.
Buffer Saturation: The buffer utilization as a percentage per each connected Collector instance.
The heap recommendations for the Collector (1 GB for a maximum of 100 Agents, 4 GB for a maximum of 200 Agents and 8 GB for a maximum of 400 Agents) in the deployment scenarios described here are based on the following assumptions:
Did you try to follow the collector best practices?
Pl. answer below:
What is the Hardware like of the dynatrace server (CPU specs, Cores, Memory, Fast Storage) ?
Are you running collector and server on same host?
Are your load injectors kept separate from Dynatrace?
Is there significant latency between agents and collectors?
What is the average size of the purepath length (in server health you can plot this)?
Seems like your collector is being overloaded. It might be a good idea to do one (or two) of the following:
1. Increase the heap memory of the collector
2. Create a new collector instance on the same host (if the collector heap is already at 8GB and the host still has available memory)
3. Create a new collector instance on a different host
Since your transaction/sec is not more than 1.5k, I'm thinking that doing either 1 or 2 should suffice.
Thank you for your response. Server got 16 Core, 64 GB memory, and Collector got 4 core and 10 GB memory. Storage 1 TB on server. Server and Collector is in different data center. No sharing, just only server and collector process running. Total agent 48 (java and Webserver). Even we have deleted few plugins and BT, still having same issue. Where can I see the no of loaded classes and Instrumentation depth ?
Ping Results: between these server:
64 bytes from CollectoServerHost (xxx.xxx.xxx.xxx): icmp_seq=1 ttl=59 time=51.6 ms
64 bytes from CollectoServerHost (xxx.xxx.xxx.xxx): icmp_seq=2 ttl=59 time=58.8 ms
64 bytes from CollectoServerHost (xxx.xxx.xxx.xxx): icmp_seq=3 ttl=59 time=47.2 ms
64 bytes from CollectoServerHost (xxx.xxx.xxx.xxx): icmp_seq=4 ttl=59 time=62.9 ms
64 bytes from CollectoServerHost (xxx.xxx.xxx.xxx): icmp_seq=5 ttl=59 time=36.5 ms
64 bytes from ServerHost (yyy.yyy.yyy.yyy): icmp_seq=1 ttl=59 time=47.2 ms
64 bytes from ServerHost (yyy.yyy.yyy.yyy): icmp_seq=2 ttl=59 time=55.8 ms
64 bytes from ServerHost (yyy.yyy.yyy.yyy): icmp_seq=3 ttl=59 time=58.6 ms
64 bytes from ServerHost (yyy.yyy.yyy.yyy): icmp_seq=4 ttl=59 time=67.0 ms
64 bytes from ServerHost (yyy.yyy.yyy.yyy): icmp_seq=5 ttl=59 time=59.1 ms
64 bytes from ServerHost (yyy.yyy.yyy.yyy): icmp_seq=6 ttl=59 time=53.9 ms
Open the Agents Overview dashboard see the 'Class Load Count' and all other relevant columns.
What is the avg/max Instrumentor time? (Recommended should be < 100ms).
You can also see the 'Class Cache Size' of each collector from the Collecter Sizing dashboard.
You already have 10GB memory in the Collector Host, so increase the heap size of the collector from 2 to 8 in the dtcollector.ini in the first stage and later on you can add more memory in the host and then create one more collector instance.