Iceberg

.net Agents and .net runtime problem

Discussion created by Iceberg on Jan 25, 2012
Latest reply on Jan 26, 2012 by Iceberg
Today I disabled all our wily agents on our .net production hosts because the agent was loading the v2 .net profiler repeatedly, and causing a serious performance issue.

As to why this constant reloading of the profiler is happening, I have no leads. Not sure if it's an agent issue, or a windows issue. From the MoM, everything looks fine. In the event logs on each machine, there are hundreds of .net runtime errors from the agent.

What I do know is that the v2 profiler does a full restart of the worker processes each time it tries to load the profilier. This causes undue performance impact, since the worker process takes much longer to respond when first created. Subsequent request are met quickly, until such time as the wily agent once again tries to load the profilier and the worker process must reestablish everything in the cache.

I'm uncertain if a later .net runtime would do the same full restart of the worker process, or simply try to load the profiler. Would be nice though, if the agent used a more recent version of .net runtime.

If anyone knows why this might be happening, I'd love to hear from you.

Using...
- windows server 2008 R2 v6 (sp1) and IIS 7.5
-.net agent v9.0.7.0 (with APM9)
- native profiler (autoprobe not being used since the native profiler footprint was much smaller)
- all configurations match the CA specs (env. variables, wilypermissions.exe, registered dll's, etc...)

Thanks,
Stewart

Outcomes