DB2 Clustering

Hi,
>
I am currently in the process of gathering info/experiences for an
incoming Linux DB2 clustering phase we actually know nothing about
(since we are doing it for the first time ever), so I would really
appreciate if you could share some of your experience here or recommend
some additional websites/literature aside from IBM Developerworks -
specifically, if there are some caveats or things to look out for during
the process, etc....

It would help if you gave some more details. Clustering is ambiguous,
and could refer to high-availability clusters (using a tool like Veritas
Cluster or SteelEye, or even using HADR), or it could refer to a
database running with the database parititioning feature (DPF).

Ian wrote:>
Clustering can also mean defining a clustering index, or using
Multi-dimensional Clustering (MDC).
>
Even if you are referring to using multiple nodes, you need to specify
if you are want to cluster for high availability or for scalability.
>

chmmr wrote:>
It would help if you gave some more details. Clustering is ambiguous,
and could refer to high-availability clusters (using a tool like Veritas
Cluster or SteelEye, or even using HADR), or it could refer to a
database running with the database parititioning feature (DPF).
>
>

Even if you are referring to using multiple nodes, you need to specify
if you are want to cluster for high availability or for scalability.

Sorry for being imprecise - I was referring to the DPF feature of the DB2.

DPF is primarily designed for massive parallel processing of large data
warehouse databases. It can be used for HA, but is not really designed
for that purpose.

Ian wrote:>
Sorry for being imprecise - I was referring to the DPF feature of the DB2.

chmmr wrote:
>>
>
>
This is much too broad of a topic. DPF is fairly straightforward and
easy to use if you have a solid understanding what you're dealing with.
>
However, this understanding is critical, and DBAs make lots of critical
(but simple) errors when they don't really understand.
>
>

Ian wrote:>
Clustering can also mean defining a clustering index, or using
Multi-dimensional Clustering (MDC).
>
Even if you are referring to using multiple nodes, you need to specify
if you are want to cluster for high availability or for scalability.
>

Mark A wrote:>
Thank you very much for all your replys, currently we're investigating a
problem with instance-owning users on one of the machines that (will)
participate in the cluster.
>
Perhaps what I forgot to mention we're trying to build a Linux DB2 Cluster.
>
In particular, after stopping all the related services (be it via
db2stop or pkill) and deinstalling the DB2 product (./deinstall
scripta), I cannot find where does DB2 store its information about
instance-owners.
>
What happens is that, with each consecutive installation, I can't get
rid of the db2inst(X) user that was valid in the previous installtion.
So, now we have users db2inst1, db2inst2.....db2inst4, even though we've
thouroughly checked everything db2-related that we know of, deleted the
users via userdel command and deleted the groups with groupdel plus
editing/confirming with the /etc/groups and gshadow files to make sure
all groups are really gone.
>
So, what we do know now is that, OS-wise, those users and groups are gone.
>
And the reason we're trying to get rid of these users is that, during
GUI installation, when I try to use the existing user, during the final
phase of installation, ./db2setup fails to switch to that user and I am
forced to abort and reinstall with a different instance-owning username.
All other users perform "normally" and are deleted after deinstallation.
>
Right now, being left with no options, I've removed the DB2 and all
traces of it and am doing a disk-wide string search for "db2inst4"
string (which, I remind you, is one of the users that are still reported
as 'used' by the db2setup script).
>
Since this will take a long time, any help is appreciated.