topic VCS 3.5 with CFS in Operating System - HP-UX

VCS 3.5 with CFS

Innopath Support — Thu, 14 Jun 2007 12:17:30 GMT

I have a HA cluster with two nodes and one Cluster File system (CFS). One of the node fails and now the surviving node is not able to restart.

carkdsa2#root# hastart
carkdsa2#root# hastatus
attempting to connect....connected
VCS:11307:Node has not received cluster membership yet, cannot process HA command
VCS:11307:Node has not received cluster membership yet, cannot process HA command
VCS:11307:Node has not received cluster membership yet, cannot process HA command

I already try "hasart -onenode" and it looks like it will start but cvm resource will never be offline, VCS will try and eventually declare the resource faulted.
carkdsa2#root# hastatus -sum

-- SYSTEM STATE
-- System State Frozen

A carkdsa1 UNKNOWN 0
A carkdsa2 RUNNING 0

-- GROUP STATE
-- Group System Probed AutoDisabled State

B ClusterService carkdsa1 Y Y OFFLINE
B ClusterService carkdsa2 Y N OFFLINE
B cfs carkdsa1 Y Y OFFLINE
B cfs carkdsa2 Y N OFFLINE
B cvm carkdsa1 Y Y OFFLINE
B cvm carkdsa2 Y N STARTING|PARTIAL

Any idea if I need to reconfigure before I can restar the surviving node?

thanks,

Re: VCS 3.5 with CFS

Sameer_Nirmal — Thu, 14 Jun 2007 16:09:43 GMT

I think you haven't modified the main.cf file before starting VCS. If you want to make a single node VCS cluster, you need to modify the main.cf file to remove the failed node. You don't need LLT and GAB running and should be stopped. Then you can start the VCS with "hastart -onenode"

Refer this document for steps to make single node VCS cluster.

http://ftp.support.veritas.com/pub/support/products/ClusterServer_UNIX/252159.pdf

Re: VCS 3.5 with CFS

Suraj Singh_1 — Fri, 15 Jun 2007 02:40:08 GMT

What was the cluster status before one of the nodes failed? (O/p of hastatus -sum).

If your cluster was configured properly, then both the nodes should have been online before the node went down, and even after the node went down, the other node would have remained in ONLINE state, you are not required to execute 'hastart' command on the surviving node.

Looking at the o/p of hastatus -sum command, it looks that the CVM service group was in OFFLINE state on node 'carkdsa2' when the node 'carkdsa1' went down.

Can you paste the o/p of "gabconfig -a" command, and the relevent entries from "/var/VRTSvcs/log/engine_A.log".

Re: VCS 3.5 with CFS

Innopath Support — Fri, 15 Jun 2007 10:25:00 GMT

Hi Suraj,

You are right, the cluster was up and all resource were online. After carkdsa1 went down, carkdsa2 node remain up. After that I had to reboot carkdsa2 for other reasons and now the cluster fails to start because under the current configuration it waits for the other node see.

carkdsa2#root# gabconfig -a
GAB Port Memberships
===============================================================

(notice this output if the results from command "hastart -onenode".

Also I reconfigure main.cf with only one node.
############
TAG_E 2007/06/15 08:18:17 (carkdsa2) VCS:13001:Resource(cvmvoldg1): Output of the completed operation (monitor)
state: joining nodeId=0 masterId=-1 neighborId=-1 members=0x0 joiners=0x0 leavers=0x0 reconfig_seqnum=0x0 reconfig: initialized
state: joining nodeId=0 masterId=-1 neighborId=-1 members=0x0 joiners=0x0 leavers=0x0 reconfig_seqnum=0x0 reconfig: initialized
######################

I think there should be a way to restart only one node right?

Re: VCS 3.5 with CFS

Innopath Support — Tue, 19 Jun 2007 15:06:01 GMT

I found that i need it to reconfigure gab with:

gabconfig -cx

then issue
hastart