Questions on MC/SG

[Deanna Tran_1]

I have 2 nodes cluster. Each node has its own package. These nodes are conneted to the storage. Performed error injection to the storage --- eventually take away one of the path ( each vg in the cluster set up via PVlinks). Node A panic " Lost of cluster lock diskes" causes the TOC to occur, node A rebooted. Maybe i am missing something about PVlinks, isn't it for multipathing...if I/O cannot sense on the primary path, it should alternate itself to alternate path without any problems. And since this is a 2 node cluster type, the cluster lock disk is need...so if the cluster lock disk is the disk that belong to one of the node resource...meaning this volume has pvlinks to it..shouldn't cause the node to do TOC...unless...there is another way to configure this that i didn't know..please help thank you

[David Bell_1 · Accepted solution]

Deanna,

You didn't mention what SPU and O/S you are using. Supposing HPUX 11.11, please look into patch PHKL_26743 (LVM Cumulative Patch). This patch supersedes patches that dealt with PVlinks, cluster lock, etc.

HTH,

Dave

[Jeff Schussele]

Hi Deanna,

Do a
vgdisplay -v /dev/vg_lock_name
Verify that there are indeed primary & alternate paths to the VG defined for the lock volumes.
Also verify that the cluster ascii file defines only the FIRST_CLUSTER_LOCK_VG & not just a single FIRST_CLUSTER_LOCK_PV for the obvious reason that if you only specify the one PV & it's link fails.......
I believe you only need one (VG) or the other (PV)...don't think both are required. But a quick scan thru the docs don't verify that.

HTH,
Jeff

[Deanna Tran_1]

hi it is running on o/s 11i
i did the swlist..and here is the result..and i didn't see the patches that you recommendated?

B3935DA A.11.13 MC / Service Guard
B8324BA A.01.03 HP Cluster Object Manager
BUNDLE11i B.11.11.0102.2 Required Patch Bundle for HP-UX 11i, February 2001
CDE-English B.11.11 English CDE Environment
FDDI-00 B.11.11.02 PCI FDDI;Supptd HW=A3739A/A3739B;SW=J3626AA
FibrChanl-00 B.11.11.09 PCI/HSC FibreChannel;Supptd HW=A6684A,A6685A,A5158
GOLDAPPS11i B.11.11.0112.6 Gold Applications Patches for HP-UX 11i, December
GOLDBASE11i B.11.11.0112.6 Gold Base Patches for HP-UX 11i, December 2001
GigEther-00 B.11.11.14 PCI/HSC GigEther;Supptd HW=A4926A/A4929A/A4924A/A4
HPUX11i-OE B.11.11.0112 HP-UX 11i Operating Environment Component
HPUXBase64 B.11.11 HP-UX 64-bit Base OS
HPUXBaseAux B.11.11.0112 HP-UX Base OS Auxiliary
HWEnable11i B.11.11.0112.5 Hardware Enablement Patches for HP-UX 11i, Decembe
OnlineDiag B.11.11.05.10 HPUX 11.11 Support Tools Bundle, Dec 2001
RAID-00 B.11.11.01 PCI RAID; Supptd HW=A5856A
perl B.5.6.1.C Perl Programming Language

Product(s) not contained in a Bundle:


PHCO_25723 1.0 q4 patch version B.11.20f
PHKL_25304 1.0 Preemption, VM allocation, Psets, MRG paging
PHSS_25685 1.0 L1000/L2000 41.38 firmware patch
PHSS_26138 1.0 OV EMANATE14.2 Agent Consolidated Patch
PHSS_26180 1.0 MC/ServiceGuard and SG-OPS Edition A.11.13

[David Bell_1]

Deanna,

You may also want to look at patch PHKL_26519 if you are indeed on HPUX 11.11. Also, it may be helpful to know what kind of storage you are using and giving a little information if this is a newly installed configuration, O/S level, SPU, and whether this test has been performed successfully in the past. If so, what changes have been made since the last test passed successfully.

HTH,

Dave

[Deanna Tran_1]

hi jeff, and David...
To Jeff question : yes, it is first_cluster_lock_vg and not disk...
and yes a vgdisplay show the alternative path...verification succesffully....
To David :
These tests have been performed in the past
without any problems...and i am running it against ESS storage....
thanks again for your help...

conclusion?

[Sanjay_6]

Hi Deanna,

To allow for the cluster lock disk path to fail, you should have a dual cluster lock for the configuration. Say the cluster lock disk is c0t0d0 and the path using Controller 0, "c0" fails, the system will not be able to find c0t0d0 and hence it will panic. But if you have a dual cluster lock configured, say c1t0d0, even if the path to c0t0d0 fails, the lock would still be working using the second cluster lock disk.

If both your nodes are working, the cluster lock disk is not used, but if one of the nodes fail the cluster lock disk is used to work as a tie-breaker between the nodes or it is used by the node that is active.

What it means is that when both the nodes are working, the loss of the path to cluster lock (in a single cluster lock disk configuration ) disk should not panic the system, but in case one of the nodes is down the loss of path to the cluster lock (in a single cluster lock disk configuration )disk can panic the system.

Hope this helps.

Regds

[Deanna Tran_1]

Hi Sanjay,
I am using cluster_lock_vg not a single lock diskes..and the cluster lock vg has multipathing.... therefore, i don't think the host will panick..but it did....
so perhaps as suggested the patches have not been updated completed??

[David Bell_1]

Deanna,

Take a look at the following document. I'm still looking for more information regarding the PVLINKS.

http://support1.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000058849467

HTH,

Dave