Assistance required

[Systems Department]

Have an A500 running 64 bit HPUX 11.10 and every three weeks the system crashes with : /usr/lib/dld.sl : call to mmap() failed - BSS /usr/lib/libnsl.1
followed by : /usr/lib/dldsl : not enough space
The system is not live yet and no one is using it except for the occasional testing.
Last occured one week ago and the memory usage since then has increased from 23% to 34% and the secondary swap has gone from not being used to 3%, the primary swap still stands at 0%. No one is on the system except myself and I'm sat at the shell below root.

[Charles Harris]

Just a thought, the dld.sl is a dynamic library and library loader ( fresh from the man page ). The error you are getting implies that there is issufficient file system space to load the library although this may not be the case. The fact that is dies regularly would suggest that either there is a memory leaky application or maybe a bad process being run by the cron. It could also be a patch level problem that may have been fixed in the latest bundle or about a million other things.....

Hope this gives you some pointers......

-ChaZ-

[Magdi KAMAL]

Hi,

I've got this problem to. You should consider a swap space from 3 to 4 times the amount of memory.

You gave a precision on the swap percentage used but you must also consider what the hp-ux intend to reserve space on the swap !

This value of RESERVING space can exceed what the available swap area have. So, in this case mmap appear !!!

Good luck

[John Palmer · Accepted solution]

Hi,

It sounds as though a process has some sort of memory leak and is eating up swap space.

'mib2agt' used to do this on HP-UX 11.00 if patch PHSS_21046 wasn't installed. You'd have to check the equivalent patch for 11.10.

You can list processes in Virtual Set Size order with the following simple script:-

UNIX95= ps -e -o ruser,vsz,pid,args | sort -rnk2 | more

(Attributable to Bill Hassell). Note the space after = is required. The second field output is the process virtual set size in Mb and the output is sorted in descending order of process size.

Hope this helps.

Regards,
John

[Stefan Farrelly]

Weve had a similar problem.

Some process/application is definitely gobbling up all your swapspace. Slowly, but surely. You need to monitor you server constantnly day by day to find the culprit. Eventually we did, it was an Oracle IFS deamon which has a huge memory leak - still not resolved except by retsarting the application every night!

[Shahul]

Hi

I think U can solve this problem by loading dld.sl patch. There are latest patch available for dld.sl. And U told that swap size is increasing. I am not sure, But it may be because of some NFS related problem. Once I faced this, I could solve this by stoping and starting NFS client and server services. U can also try the same. For that

# cd /sbin/init.d
#./nfs.client stop
#./nfs.server stop
#swapinfo
#./nfs.clent start
#./nfs.server start

Best of luck
Shahul

[Systems Department]

Thank you all for your assistance. The patch PHSS_21046 has been replaced with PHSS_23670. I had found that the 'mib2agt' process was taking 451Mb of memory. The patch applied is keeping it stable at 600kb.

Thanks again to all of you.