Processes dying

[Jack C. Mahaffey]

I just started having some intermittent processes dying on one of my HP 11.0 servers. Is this likely a memory issue? Here's output from swapinfo -mat:

[1033]root@lukoptp2:/root # swapinfo -mat
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 1024 423 601 41% 0 - 1 /dev/vg00/lvol2
dev 1024 424 600 41% 0 - 1 /dev/vg00/lvol3
dev 1000 295 705 30% 0 - 1 /dev/vg00/lvswap3
reserve - 1894 -1894
memory 1501 589 912 39%
total 4549 3625 924 80% - 0 -
[1034]root@lukoptp2:/root #


Note: lvswap3 and lvol3 are temporary. I had a swap issue 3/4 months ago where I had to add swap on the fly. I will be combining these swap areas on next reboot.

Jack

[Paula J Frazer-Campbell]

Jack

Processes do not just die, they either:-

1. Get killed.
2. Error out.
3 Finish and exit gracefully.

What are the processes that apparently die?

Can you use Tusc and monitor them?


Paula

[RAC_1]

With swapinfo there does not seem to be memory issue? How much RAM you have?

Also are the proccesses creating any core files? Start the process with tusc and see where and why it's dying?

[Steven E. Protter · Accepted solution]

Could be memory,
could be kernel parameters like nproc and/or nfiles

You seem to have some swap available.

I'd run some data collection with the script I'm attaching(at this moment hp is going over my output in a software case) and look at the swap figures.

You might not want to combine them, you might want one swap area equal to half of ram as priority 0 and a secondary, larger one with a higher priority number which means its less likely to be used.

This approach maximizes throughput during lighter/medium system loads.

SEP

[Jack C. Mahaffey]

2gb memory 2 cpus
I'll need to get back on which specific processes. App administrator noticed that there are logins that don't need to be running. Some folks apparently are not logging out cleanly. Another sympton is there were some problems with logins hanging with no response from the server.. Once some of the really old logins were killed, this particular problem with the hanging logins went away.

Will come back later with more info.


jack...

[Jack C. Mahaffey]

Checking for core files...