Help needed for clustering with serviceguard

[Jessica P]

Hi all

I have four servers which are to be configured into 2 clusters of 2 nodes each. All these nodes hav just the OS and are a part of SAN which will hav all the data.
could you help me a deatiled procedure to go ahead with clustering.
also can i hav cross cable b/w the nodes for heart beat .
how do i go ahead with quoram disk

plz help...
dont worry about points...


thanks
vivek

[melvyn burnard]

Well you should start by reading the Managing Serviceguard manual available at
http://docs.hp.com/en/ha

But if you have not done this before then it may pay you to get some outside consultancy to assist.

[Steve Steel]

Hi

Indeed

http://docs.hp.com/en/B3936-90073/index.html

http://docs.hp.com/en/B3936-90073/ch05s06.html#d0e10820


Steve Steel

[BPatrick]

Hi
You can find step by step procedure on setting up serviceguard in the "Managing serviceguard" documentation available in this location
http://docs.hp.com/en/ha
you will find this documentation for different version of serviceguard on different os.
Regarding your question about heartbeat, Yes you can have cross cable for your heartbeat and you can have you quorum disk setup on san.
before setup of serviceguard check whether you have appropriate hardware which can take advantage of redundancy provided by serviceguard.

Regards
Patrick

[Geoff Wild]

Can't give a detailed procedure without detailed info on your servers - IE network, disks, OS / SG versions, etc.

Best to read those docs on docs.hp.com and/or go on a ServiceGuard course.

Basically, just follow the steps in the http://docs.hp.com/en/ha.html#Serviceguard for the ServiceGuard version you have.

Rgds...Geoff

[A. Clay Stephenson]

Your question is almost at the level of "How to I drive a car?". Answer: Carefully.

Installing MC/ServiceGuard is easy; it's all of the thing you do to make your systems robust enough that MC/SG will very seldom be needed that is hard. You need to study all the guides and think about attending a class before tackling your first production cluster.

I would not use a crossover cable for heartbeat; just buy an inexpensive hub. The indicators on the hub can be very helpful when trying to diagnose a problem and since you will always have more than one heartbeat LAN (you will, wont't you?), the failure of a hub adds almost zero additional risk and the risks are far outweighed by the diagnostic benefits. The only requirement for a lock disk is that it be an LVM disk that is accessible by all the nodes; however, these days, even for 2-node clusters a Quorum Server is the preferred method. A small PC running Linux makes an excellent QS.

[freddy_21]

about MCSG procedur:\
1. please export vg at server 1 and import vg ( please using map file )

server 1:
vgexport -v -s -p -m mapvg01 vg01

copy to mapvg01 at server 1 to server2

server2:
mkdir /dev/vg01
mknod /dev/vg01/group c 64 0x010000
vgimport -s -v -m mapvg01 vg01


2. set your IP for heartbeat. Cross cable possible to configure it. NOTE heart beat IP must have diferent subnet with your production IP


3. test connection using linkloop between your production IP with heartbeat IP ( must isolated)

4. create file cmclnodelist at /etc/cmcluster

the contain is:
server1 root
server2 root

5. run command at server 1 where the shared disk active

cd /etc/cmcluster
cmquerycl -V -C cluster.ascii -n server1 -n server2


the output is cluster.ascii. you can check about the configuration. usually cluster lock disk already created. Check and do some changes if neededfor your ascii file.
ex :
CLUSTER_NAME
NETWORK_POLLING_INTERVAL 9000000
MAX_CONFIGURED_PACKAGES


run cmcheckconf and cmapplyconf.

If your configuration right. i think you not have problem about that.



thanks
freddy