HPUX 11.0: very high PRI queues?

[UX-Linux CC]

Looking for experience, root cause of extremely high pri queues on a low utilized hpux 11.0 system (L3000/4{550}/4G mem). We have ruled out measurement errors (MIDAEMON), and application interference. Pri queues running >1000 for days at a time, no apparent reason. Other queues less than 1.
Anyone recognize this problem?

P.S. This thread has been moved from HP-UX Technical Documentation to HP-UX > sysadmin. -HP Forum Moderator

[RAC_1]

What is your kernel setting for timeslice?? Should be 10. Also overall, what kinf od a bottleneck do you experience??

Check with glance

Anil

[Henk Geurts]

This thread might help:
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=43928
regards.

[UX-Linux CC]

KMTUNE returns:
timeslice (100/10)

THere are no bottlenecks...the system performs well...we see these huge pri (not RUN!) queues with no apparent reason.

[RAC_1]

You sure you do not see any kind of bottlnect?? How is disk utilization, lan and memory utilization??
Is sar -d 2 2 OK?? glance -B, glance -l

Anil

[Steven E. Protter]

glance gpm is the best way to detect the bottleneck.

I'm attaching some sar scripts for those like myself who no longer have access to gpm.

SEP

[UX-Linux CC]

gbl_cpu_util <5
gpl_disk_util_peak < 20
gbl_mem_pageout_rate <2
gbl_mem_util < 70
GBL_net_packet rate <400 (with occasional peaks >1000)
(these numbers for 10 day time window)
No evidence of any bottlenecks.

[UX-Linux CC]

Still no solution to this. If anyone comes across useable information, please contact:
chuck_smart@agilent.com

thanks!