[Linux-HA] 3 SERVERS in CLUSTER (db2)

Laurent LE PRIEUR laurent.le-prieur at ac-nancy-metz.fr
Tue Oct 4 01:13:12 MDT 2005


----- Original Message ----- 
From: "Alan Robertson" <alanr at unix.sh>
To: "General Linux-HA mailing list" <linux-ha at lists.linux-ha.org>
Sent: Tuesday, October 04, 2005 3:12 AM
Subject: Re: [Linux-HA] 3 SERVERS in CLUSTER (db2)


Le-Prieur Laurent wrote:
>
> ---- Messages d´origine ----
> De: Alan Robertson <alanr at unix.sh>
> Date: Lundi, Octobre 3, 2005 9:10 pm
> Objet: Re: [Linux-HA] 3 SERVERS in CLUSTER
>
>
>>Laurent LE PRIEUR wrote:
>>
>>>Hello,
>>>
>>>I have activate the CRM for 3 nodes in cluster.
>>>
>>>I would wish to set up the following design
>>>
>>>2 Nodes ACTIVE and 1 node PASSIVE connected on SAN Network.
>>>
>>>Each of the 2 active nodes launches DB2.
>>>
>>>With the structure and support IBM DB2, the directory for DB2 is
>>
>>/opt/IBM/db2>
>>>If the versions of DB2 are not the same ones I am constrained to
>>
>>authorize the fail over on the passive node only for the first node which 
>>would fall defective.
>>
>>>Can I put this operation places from there with the CRM and so
>>
>>yes how?
>>
>>
>>I assume you have two different DB2 instances on the two active machines?
>
>
> Yes Alan, two diffrent DB2 instances and potentially two different release 
> (fixpack) for DB2 (one node active node with fixpak 8 and other with 
> fixpak 10).

Is this going to work?  Can you actually have two copies of db2 on the
same machine at the same time?

> I have two partition per node on SAN :
>
> first partition mounted /opt/IBM/db2 (content /opt/IBM/db2, /var/db2 
> (symbolic link)  and homedir for dasusr1)
>
> second partition mounted /datainst1 (content homedir for db2inst1 et 
> db2fenc1 and the data for instance and databases).
>
> I have sync on all node /etc/passwd
> /etc/shadow
> /etc/group
> /etc/services
>
> And i have disable in /etc/inittab this line for HA environnement
> fmc:2345:respawn:/opt/IBM/db2/V8.1/bin/db2fmcd #DB2 Fault Monitor
> Coordinator
>
> I will use the passive node if one (and only one) of active is fault or 
> maintenance.

>I think each machine needs to have its own homedir installed - somewhere in 
>the homedir there hides the name of the machine it's installed on.

>You need to have ONLY the tables move from one machine to another - not the 
>homedir.

Ok Alan, but is not the response of my initial question...

I would wish to set up the following design
2 Nodes ACTIVE and 1 node PASSIVE
I am constrained to authorize the fail over on the passive node only for the 
first
node which would fall defective.


Laurent 




More information about the Linux-HA mailing list