NetWare 6.5 running out of memory

[n13975]

I have a SAN attached NetWare 6.5 SP5 IBM x365 server with 4Gb of RAM that continually runs out of memory and causes the server to multiple abend and crash (abends both through normal working day and when attempting to backup).

The server has dual Qlogic HBA FC's and accesses both NSS and traditional volumes. The traditional volume is 900Gb with approx. 3 million files that are currently compressed.

This server was upgraded from Netware 6.0 to 6.5 and the memory problems occured. The server was then rebuilt from scratch as NetWare 6.5 and then problems still occur.

When the server was NetWare 6.0 Sp5 all functions worked correctly.

I have upgraded the RAM upto 8Gb but this seems to make no difference in performance as the OS seems to really only recognise the first 4Gb.

The only function that this server performs is a data server.

The backup software being used is TSM Client version 5 Release Level 2.0. The version of TSAFS is 6.51.05 (although older versions still cause the same problem) .

Can TSA600 be used as opposed to TSAFS on NetWare 6.5?
Does anyone have any preffered memory setting e.g cache buffers etc.

Any ideas at all would be appreciated.

 

[marvhuffaker]

you can use TSA600-- that should work fine if it will work with your backup software. TSAFS seems to be the largest contributor to memory problems.

There's also a TID that talks about this.. You should go through each step and make the appropriate changes.

http://support.novell.com/cgi-bin/search/searchtid.cgi?10091980.htm

Marvin Huffaker, MCNE

http://www.redjuju.com

 

[terry712]

why are you using traditional and not nss
is it purely as the nss is will uncompressed ?

surely traditional is a big hit here
the load time must be slow for a volume that size

 

[n13975]

The reason for using a traditional volume is purely down to available disk space on the SAN. When tring to convert to NSS the volume becomes too large.

To mount the volume it takes approx. 15 mins.

Any sugggestions on how to handle memory over 4Gb?

 

[marvhuffaker]

Memory over 4gb is not your problem. Understand that with both volume types you are splitting the available resources between two different file systems. Also with 3mil files, you may be finding that you are running into certain limitations of the Traditional file system that you wouldn't have with NSS.

Marvin Huffaker, MCNE

http://www.redjuju.com

 

[Novelli]

,

nevertheless that it would be the best idea to go to NSS Volumes, mybe this helps you also

http://support.novell.com/cgi-bin/search/searchtid.cgi?10095039.htm

Andreas

Cheers Novelli
Novell Master CNE + CDE

http://www.eckertweb.de

 

[sstoppel]

I had a nearly identical issue. TSAFS would allocate more and more cache buffers and never release them. I had an interim solution in CRONning an unload and reload of the NLM. The actual answer was in this TID:

http://support.novell.com/cgi-bin/search/searchtid.cgi?/10091980.htm

 

[LawnBoy]

I also had a TSAFS related memory leak, but it was cured by SP4a. My other memory leak was caused by NSS and SP4b fixed that.

You seem to be assuming TSAFS is the culprit and you may be right, but you can tell for sure by using the Remote Manager's "List Modules" page. It'll show you the memory consumption of every module running listed top-down from fattest to leanest. Check this page periodically (depending on how fast the problem occurs) and track the top memory hogs to see just which module is the problem.

 

[marvhuffaker]

I'm also assuming that TSAFS IS the problem because I've seen so many cases of that. But put those tweaks in place as mentioned in that TID, and it will usually nip it in the butt. One problematic server of a customer has not had any problems post SP5.

But lawnboy is right, check and make sure there aren't other modules creeping up there in memory allocation. I recently had a server that was leaking, and found that Apache was sucking 400MB of RAM after only a couple days uptime. Found it in the Portal and was able to figure out that somehow it was running older modules and the Service Pack had completely skipped it, so it was running old code.

Marvin Huffaker, MCNE

http://www.redjuju.com