Nagios Support Forum

Are the 4.2.1 and the 3.5.1 hardware the same? Also, what are the RAM and processor information? I don't have a SPARC for testing, but I can try to replicate the system the best I can. I've got 4.2.2 installed on Solaris x86 in a VM.

Also, could you upload your nagios.cfg? Thanks!

Yes, the hardware are the same on 4.2.1 and 3.5.1, "System Configuration: Oracle Corporation sun4v SPARC T3-1". We currently have our production system running 3.5.1 on the same sunv4 SPARC T3-1 hardware in a different site and this is still performing like a charm.

Added the "nagios.cfg" file as requested. The main item I have adjusted was of course "check_workers=8", as the the SPARC T3-1 supports 128 CPU's/threads.

Let me know if I can assist with further testing, have the standby/backup site available to do some offline testing with real world environment.

Regards

Jaco Lesch
SAIX HLS

jacolza wrote: Let me know if I can assist with further testing, have the standby/backup site available to do some offline testing with real world environment.

My first thought would be to step through the various versions and see where performance issue was introduced. However, compile issues prevent that. 4.2.0 has documented issues with Solaris.

The next thing I would do is upgrade from 4.2.1 to 4.2.2. just to make sure the performance issue still exists (and also to make sure it isn't worse).

John may have more specific thoughts on how you can help test. Thanks so much!

Nagios 4.2.2 compiled and installed successfully, even the GUI install is working properly now without having to edit the Makefile. I have been running this on the standby platform for more than 5 days with a production workload and the CPU load has dropped form an average of 9 to below 1. Still seeing some <defunct> processes though, but Solaris seem to reap these <defunct> process in a timely manner, attached the output from 'ps -f -u nagios' for reference. Also attached truss output from one of the nagios worker processes.

Also set the check_workers to, "check_workers=24', without any strain on the CPU, or any run away processes like before in 4.2.1.

Thanks

Jaco Lesch
SAIX HLS

I sent the new info to one of our devs.

For clarity, is your issue resolved? If so, can we close this thread? Thanks!

Yes thanks, issue is resolved.

Regards

Thanks again for sending the info. I'm closing the thread.