mid server performance peaking on CPU with jdk in madrid version

[jb2291]

Many of our mid servers are peaking with 100% CPU utilization even though the service is set to 25 threads during certain phases of discovery.  The OpenJDK Platform Binary appears to be consuming all the CPU when I see the process pegged in the task manager.    Are other people experiencing performance bottlenecks when discovery is running on the mid server?   

Are there particular versions of the JDK that have improved performance management of the CPU loading?

[tompowe]

I seem to be having this same problem on one of my MID Servers.  I got this error in the log:

WARNING *** WARNING *** Unexpectedly high SSHClientEngine select polling period: 20.0s

[sr_surendra]

Hello @jb2291 

I am facing same issue.

Could you please help me if you were able to rectify similar issue you were facing.

Regards,

Surendra

[jb2291]

This issue is not resolved.  We are on the New York instance.  We still see 99-100% CPU loads on mid servers during certain phases of discovery schedule. Fortunately, most servers do recover depending on what phase discovery is in.  Unfortunately, these high CPU loads cause ServiceNow to think the mid server service is down when in fact the load on the server is blocking simple status connection for monitoring.

[sr_surendra]

Ok,

Thanks for replying.

Mid server which I'm using is used for specific set of IPs only so I was thinking about dividing IP ranges mentioned to sub ranges to different discovery schedules and run these discovery schedules on different times.

If that doesn't solves issue; I'm thinking about increasing server capability (RAM ,, some other things).

Will let you know if that solves our issue.

Regards,

Surendra