DB2 Install in HA environment

Thread view:

Enable EMail Alerts

Start New Thread

Thread rating:

Bruno LIVERNAIS - 02 May 2007 12:01 GMT

Hi,

Is there someone on earth (or elsewhere ;-)) who knows how to install a
DB2 V9.1 database engine in a High Availability cluster environment?

I need help on that particular point.

Best regards,
Bruno LIVERNAIS

Reply to this Message

Mark A - 03 May 2007 02:20 GMT

> Hi,
>
[quoted text clipped - 5 lines]
> Best regards,
> Bruno LIVERNAIS

You are going to have to clarify exactly what you mean by "High
Availability cluster environment"

This could vary depending on if you are using HACMP, HADR, VCS, etc,
etc.

Reply to this Message

Bruno LIVERNAIS - 03 May 2007 07:24 GMT

Mark A a écrit :

> You are going to have to clarify exactly what you mean by "High
> Availability cluster environment"
>
> This could vary depending on if you are using HACMP, HADR, VCS, etc,
> etc.

My environment is hp MC/ServiceGuard. This HA cluster is a shared
nothing config with one package (a.k.a. resource). So it's an
active/passive config.

I've installed DB2 ESE V9.1 on the first node, with db2setup but without
creating an instance. Then I've copied config files (passwd, group,
services, etc...) on the other node. After that I've installed DB2 ESE
V9.1 on the other node, just like I've done on previous node.

Then I've created an instance on the first node and check that
db2start/db2stop is correctly working. But when I stop DB2 (db2stop) on
the first node, switch the package on the other. I can't start DB2 (db2)
with that message :

[root@node2 ~]# su db2r01 -c /db2/R01/sqllib/adm/db2start
SQL1092N "DB2R01 " does not have the authority to perform the
requested command.

That's the trick... Please help!!

Thanks

Reply to this Message

Ian - 03 May 2007 18:33 GMT

> Then I've created an instance on the first node and check that
> db2start/db2stop is correctly working. But when I stop DB2 (db2stop) on
[quoted text clipped - 4 lines]
> SQL1092N "DB2R01 " does not have the authority to perform the
> requested command.

I'm not familiar with HP/UX or ServiceGuard. But, I'm assuming that it
takes care of failing over all of the necessary file systems (i.e.
"switching the package").

Given this, can you start DB2 manually by first logging in as db2r01
(su - db2r01) and then running 'db2start' ?

Reply to this Message

Bruno LIVERNAIS - 03 May 2007 19:22 GMT

Ian a écrit :

>> Then I've created an instance on the first node and check that
>> db2start/db2stop is correctly working. But when I stop DB2 (db2stop)
[quoted text clipped - 11 lines]
> Given this, can you start DB2 manually by first logging in as db2r01
> (su - db2r01) and then running 'db2start' ?

On node1 :
1. Install DB2 with db2setup with instance creation (db2r01)
2. db2start and db2stop are successfull
3. Create the database (connection test is successfull)
4. Switch LVM volumes on node2

On node2:
1. Install DB2 with db2setup with instance creation (db2r01)
2. db2start and db2stop are successfull
3. ????

On step 3, I can not create the database because it reports an existing
one (hopefully :-))... But I can not connect either because the database
alias name or database name "R01" could not be found.

So what do I have to do to have access to my database on node2???

Reply to this Message

Ian - 03 May 2007 22:24 GMT

> On node1 :
> 1. Install DB2 with db2setup with instance creation (db2r01)
[quoted text clipped - 12 lines]
>
> So what do I have to do to have access to my database on node2???

Typically in an HA environment you don't create an instance on the
standby node. The instance (and all of its associated resources)
all get failed over to the standby.

My guess is that you are missing a file system (or multiple file
systems) that your database is using.

Since you are creating the database as part of this exercise, I assume
that you know where the database is storing its data. So, you should
verify that the database path and all storage paths / tablespace
containers are included in your "package".

If you need additional help let's take this offline.

Ian

Reply to this Message

Bruno LIVERNAIS - 04 May 2007 14:51 GMT

Ian a écrit :

>> On node1 :
>> 1. Install DB2 with db2setup with instance creation (db2r01)
[quoted text clipped - 16 lines]
> standby node. The instance (and all of its associated resources)
> all get failed over to the standby.

When I failover my "package", all related filesystems are unmounted from
node1 and mounted on node2. Then when I want to start DB2, it fails
because it does not know my database (and instance)... Be sure that all
datafiles are stored on the shared volume.

Reply to this Message

Ian - 04 May 2007 18:11 GMT

> Ian a écrit :
>
> When I failover my "package", all related filesystems are unmounted from
> node1 and mounted on node2. Then when I want to start DB2, it fails
> because it does not know my database (and instance)... Be sure that all
> datafiles are stored on the shared volume.

Does this include the instance owner's home directory?

Reply to this Message

Bruno LIVERNAIS - 09 May 2007 15:32 GMT

Ian a écrit :

>> Ian a écrit :
>>
[quoted text clipped - 4 lines]
>
> Does this include the instance owner's home directory?

Should it do so ?

Reply to this Message

Bruno LIVERNAIS - 10 May 2007 15:20 GMT

Well, as a conlusion, it appears that DB2 is really not cluster
friendly... And no one is able to demonstrate the opposite!!!!

Bruno LIVERNAIS a écrit :

> Hi,
>
[quoted text clipped - 5 lines]
> Best regards,
> Bruno LIVERNAIS

Reply to this Message

Serge Rielau - 10 May 2007 16:17 GMT

> Well, as a conlusion, it appears that DB2 is really not cluster
> friendly... And no one is able to demonstrate the opposite!!!!

No one being Ian and Mark? Or have you done some more research into
this? Like looking for Whitepapers, Redbooks, contacting IBM ...

Cheers
Serge

Signature

Serge Rielau
DB2 Solutions Development
IBM Toronto Lab

Reply to this Message

Keith - 10 May 2007 18:53 GMT

> > Well, as a conlusion, it appears that DB2 is really not cluster
> > friendly... And no one is able to demonstrate the opposite!!!!
[quoted text clipped - 9 lines]
> DB2 Solutions Development
> IBM Toronto Lab

On both nodes: Install db2 separately.

Node 1:

Create instance
db2start/db2stop
switch LVM (All necessary file systems -- including instance home)

Node 2:

rm -rf ~<InstanceHome>/sqllib (since <InstanceHome> directory was
also failed over in #3 above)
Re-create the instance with same exact command used on Node 1
db2start/db2stop

Test failover and also db2start/db2stop both manually and via package
(don't forget to check/clean shm segments, if any hanging)
Create db on primary.
Test failover and also db2start/db2stop both manually and via package

Keith Ponnapalli
IBM Certified Advanced Database Administrator - DB2 UDB V8.1 LUW
IBM Certified Solutions Expert - DB2 UDB V7.1 Database Administration
INFORMIX Certified Database Administrator

Reply to this Message

Serge Rielau - 10 May 2007 19:14 GMT

Fifth hit on google:
http://h71028.www7.hp.com/enterprise/downloads/IBM%20DB2%20ESE%20V8.1%20with%20H
P%20MC-ServiceGuard%20HA%20Software.pdf

There also is a newer version which I have sent you by email assuming
your email works...

Cheers
Serge

Signature

Serge Rielau
DB2 Solutions Development
IBM Toronto Lab

Reply to this Message

Bruno LIVERNAIS - 10 May 2007 19:19 GMT

Thanks for that doc. I've already read that one but there is nothing
concerning instance/database creation... and the DB2.sh script does not
work in Linux context.

Serge Rielau a écrit :

> Fifth hit on google:
> http://h71028.www7.hp.com/enterprise/downloads/IBM%20DB2%20ESE%20V8.1%20with%20H
P%20MC-ServiceGuard%20HA%20Software.pdf
[quoted text clipped - 4 lines]
> Cheers
> Serge

Reply to this Message