Is it possible track/log the response time for each connection within a process? We are working through spikes in execution times and it is difficult to determine with low latency enabled.
Difficulty in analysing Low-latency mode performance issues is very high. More details will be helpful to understand this case. What type of connections are we talking about in the process?. Are there database connectors inside the process or calls to other end-points?.
I guess you might have checked shared_http_server in the log folder of Boomi atomsphere if you have local atom/molecule.
Options that I know are as follows
1. Custom logging before and after each connection inside the process. Here adding custom logging (custom script) before and after each connector can help get the time taken by each connector. You can insert async sub process to get the metrics. It adds a bit of overhead but good insight can be achieved.
2. The problem with above approach is that there is a need to change the processes and if they are already in production going through change request become a long process to tackle this issue. However it can give step level details.
3. Other approach is to install monitoring on the Boomi server(s) where atoms are installed with VisualVM with JMX.Java VisualVM. It will provide performance details of each thread. I would suggest this as first step.
Good luck with your analysis!. If you find anything interesting or better then please post your findings.
Also, if possible when the server shows increased time for execution you can send request to your api in debug mode.
Retrieving data ...