HSG80 + Netware cluster

[M. Gerritsen]

Hi,
We have a Netware 6.5 sp5 cluster running on DL580 G3's (PSP7.60) with 2 FCA2214.
These are attached to two Compaq SANSwitch 16's (1GB Broccade 28xx's?) with firmware v2.6.2d. These in turn are connected to 2 HSG80 controllers (V8.8F). Behind this we have 6 43xx vcabinets filled with 72.8GB disks and devided in RAID5 sets. Where the first raid set uses the first disk of each 43xx cabinet (6 disks in a raid set, effective disk size of 360GB).
We do not use Secure Path. our startup.ncf has: SET MULTI-PATH SUPPORT=OFF
LOAD QL2300.HAM SLOT=3 /LUNS /xretry=12 /xtimeout=120
LOAD CPQSHD.CDM
We have a lot of problems of server restarts, abends an other bad stuff in our cluster.
Googleing has found me this: http://www-1.ibm.com/support/docview.wss?uid=ssg1S1001783
I can't find any related HP dosuments. Can anyone help?

[BR839478]

Did you find out where the problem is?

I have a similar problem on a DL580 G3, Netware 6.5 and storage (EVA).
The DL580 G3 is in cluster with a DL380 G4.
There are no problems with the DL380 G4.
Only the DL580 G3 freezes or falls in a Netware-sort of a blue screen (I am not familiar with Netware too much).

I replaced all the HW, but the error repeats itself.
But there is sometimes also a problem with the HP shiped spare parts, who might come DOA or with an error.

[M. Gerritsen]

We have updated to the latest QL2x00.ham drivers (6.90d by HP), replaced the cpqshd.cdm with the scsihd.cdm driver and adjusted some of the parameters on the ql driver in our startup.ncf:
;Use netware multipath IO
SET MULTI-PATH SUPPORT = ON

; For servers with FC HBAâ s
LOAD QL2X00.HAM SLOT=3 /ALLPATHS /LUNS /CONSOLE
LOAD QL2X00.HAM SLOT=4 /ALLPATHS /LUNS /CONSOLE

We do not load dosfat.nss as stated by qlogic for enabling multipath failover. We could not see any relation with path failover and the dos volume.

We also have moved the volume causing the most I/O's (a GroupWise PO) to an EVA4000.
Next to the enormous performance increase on this PO (GWcheck runs 2 times faster, a backup even 10 times! 24h -> 2.5h), other PO's left on the MA8000 performed somewhat better too.

We now plan to move all our PO's to the EVA and leave some fileshares on the MA8000 for now.

We think that the cache on the controllers is just too little. 256MB per controller leaves maybe 100MB for read and/or write cache. Combined with RAID5 (treat a GroupWise PO as a database) this was killing performance.

We have submitted nss code to Novell but have not yet recieved an awnswer yet.
Hope this helps a bit ;)

Cheers!

[BR839478]

Tnx for your reply... it got us going on.

But we found out that we had a problem with PervasiveSQL and 4GB RAM. When we tried running the server with 2GB we had no problems, but not so with 4GB.
So after replaciang almost all of HW we found out this on a new instalation of PSQL on Netware.
We are OK after we configured PSQL not to use more than 700MB of RAM (which is enough).

-marko