Primary-Contingency Setup

[Anurag]

Hi,
DB2 UDB 8.1.5 on AIX 5.x (RS6K boxes).

Setup:
2 Primary DB2 servers running VCS (say, P1 and P2), Another Resilience
DB2 Server (say, R1) where just the DB2 software is installed.
Now, DB2 filesystems on P1 are mirrored continuously to the contingency
site. R1 can access these filesystems only when the mirror link is
broken. When this mirror link is broken, some startup scripts are
executed at R1 such that R1 is able to see all mirrored DB2
filesystems.
DB2 instance is not created on R1, only the DB2 users that exist on P1
have been created (user id, group id, group name, home directory etc on
R1 are exactly the same as that on P1).
DB2 instance specific entries have been added in "/etc/services" file.
Startup script edits the "db2nodes.cfg" to reflect the correct
hostname.

Question:
When the mirror links are broken, startup scripts run on R1 and DB2 is
started alright.
The question is simple - has anybody worked on this kind of setup and
can they share their experiences? Any additional settings that I am
missing out on? Any special test that I need to carry out?

I ask these questions because I am new to this kind of resilience - I
thought (and still do) that HACMP and HADR were better options anyday.
Please share your experiences and provide tips for the setup that I
described.

Regards,
Anurag

[Mark Yudkin]

We've been using such an arrangement on Windows and AIX for over 5 years.
For Windows, we mirror the entire hard disk content, so that when we boot
the system on the second computer centre, the machine IS the one from the
first computer centre. You only have to get the VLAN definitions right for
the IP address movement. For AIX, the system disk isn't mirrored, so there
are some additional steps in the takeover scripts, such as the host name.
I'm from the Windows side, so I don't have as much knowledge of the AIX
design.

Obviously you need to test it. Plan takeover tests, analyse what happens,
what worked, what didn't, fix what didn't, loop back to the beginning of
this sentence. Perform regular (say every 6 months) takeoever tests once you
have it working. Consider this as part of regular maintenance.

"Anurag" <an**********@gmail.com> wrote in message
news:11*********************@o13g2000cwo.googlegro ups.com...

Hi,
DB2 UDB 8.1.5 on AIX 5.x (RS6K boxes).

Setup:
2 Primary DB2 servers running VCS (say, P1 and P2), Another Resilience
DB2 Server (say, R1) where just the DB2 software is installed.
Now, DB2 filesystems on P1 are mirrored continuously to the contingency
site. R1 can access these filesystems only when the mirror link is
broken. When this mirror link is broken, some startup scripts are
executed at R1 such that R1 is able to see all mirrored DB2
filesystems.
DB2 instance is not created on R1, only the DB2 users that exist on P1
have been created (user id, group id, group name, home directory etc on
R1 are exactly the same as that on P1).
DB2 instance specific entries have been added in "/etc/services" file.
Startup script edits the "db2nodes.cfg" to reflect the correct
hostname.

Question:
When the mirror links are broken, startup scripts run on R1 and DB2 is
started alright.
The question is simple - has anybody worked on this kind of setup and
can they share their experiences? Any additional settings that I am
missing out on? Any special test that I need to carry out?

I ask these questions because I am new to this kind of resilience - I
thought (and still do) that HACMP and HADR were better options anyday.
Please share your experiences and provide tips for the setup that I
described.

Regards,
Anurag