Resource intensive processes

[Paul Rafferty]

Hi,
I have noticed that several, about 14, processes that monitor time clocks are hogging the CPUs of late.
I have reniced the processes which has eased the situation/load on the system somewhat but these reniced processes are still taking more than their fair share of the CPUs, see attached. If I run the same app/process on my backup server same version of hpux 11.00 and Universe 9.6 the process cpu percent drops to 2.5% instead of the 9% for each process on the main server. The majority, 95%, of the wait state on the backup server for the process is sleep whereas on the main its reported as other.
A trace of the process produces the following.
****
sigprocmask(0x400c87e0, NULL, 0x400c880c) . [running]
sigprocmask(0x400c87e0, NULL, 0x400c880c) . = 0
select(0, NULL, NULL, NULL, 0x7f7e2a28) . = 0
sigsetstatemask(NULL, NULL, 119675188) . = 0
time(0x7f7e23a4) .. = 1006724853
time(NULL) .. = 1006724853
sigprocmask(0x400c87e0, NULL, 0x400c880c) . = 0
sigprocmask(0x40064100, NULL, 0x4006412c) . = 0
read(30, 0x40128df8, 8192) .. = 0
sigsetstatemask(NULL, NULL, 119675188) . = 0
time(0x7f7e23a4) .. = 1006724853
time(NULL) .. = 1006724853
time(0x7f7e23a4) .. = 1006724853
time(NULL) .. = 1006724853
sigprocmask(0x400d0370, NULL, 0x400d039c) . = 0
select(0, NULL, NULL, NULL, 0x7f7e2a28) . = 0
sigsetstatemask(NULL, NULL, 119675188) . = 0

****
The only changes made recently to the main server which hasn't been made to the backup server was the installation of the patches that come with the bundle QPK1100 sept 2001.

A lot of patches were installed through neccessity.
Any suggestions assistance would be appreciated
Thanks in advance.
Paul.

[Tom Geudens]

Hi Paul,
I don't know if your processes qualify for the following, but apparently there have been some problems with QPK1100 sept 2001 that are solved by PHSS_25787.
You might to take a look at http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xf3d8f715edc6d5118ff10090279cd0f9,00.html

Again, I don't know if your processes qualify ... but it might be worth a try.

Regards,
Tom Geudens

[Paula J Frazer-Campbell]

Hi Paul

These are universe phantoms written and instigated by your devel team, I would have a word with them to :-

1. Give the phantom a better name ie. phantom-clean-up.
2. Find out how many phantoms should be running.
3. Are they doing large selects?
4. Find out what each phantom does and why.
5. Query as to why the are running so many phantoms.

Do you have tusc installed if not it is worth getting as it can be run against a PID to find out what it is doing.

Universe phantoms unless written and tested very carefully can be processor/system gready, they can either loop around and continue doing a job or have an inbuild delay.

PS I am running Universe 9.4.1.1.I on three servers.