DB2/UDB 8.1 HA Failover

[pthawley]

All,

I am doing some preliminary research on HA failover capabilities on
DB2/UDB 8.1. The docs are a little light on details so I thought I'd ask
this forum ...

Let say I have a two node cluster (NodeA and NodeB) with each node
running its own instance of DB2/UDB (InstanceA and InstanceB) servicing
completely different databases and applications. I understand one
supported configuration is a "mutual failover" where each node monitors
the other and fails over under the right fault conditions.

My question is ... When the NodeA fails, does the original instanceB on
NodeB simply "mount" InstanceA's databases and run rollforward recovery
on its logs. Or when failover occurs, must an entire startup of DB2
occur (including all the costs associated with it such as allocating
memory, setting up buffer pools, etc ...)?

Perhaps it depends on the HA software used (HACMP vs Veritas, etc.), but
I imagine DB2/UDB iteself needs some capabilities to support failover to
an already running instance.

Thanks in advance

Peter
--
Posted via http://dbforums.com

[Blair Adamache]

The disks must be shared in a twin tailed configuration. Typically, the
instance is started on NodeB using the IP address of the instance of
NodeA (so the applications that were connected to NodeA can reconnect to
the same IP address). When starting, the instance on NodeB performs
crash recovery, rolling forward through the log records generated by the
Instance on NodeA.

pthawley wrote:

All,

I am doing some preliminary research on HA failover capabilities on
DB2/UDB 8.1. The docs are a little light on details so I thought I'd ask
this forum ...

Let say I have a two node cluster (NodeA and NodeB) with each node
running its own instance of DB2/UDB (InstanceA and InstanceB) servicing
completely different databases and applications. I understand one
supported configuration is a "mutual failover" where each node monitors
the other and fails over under the right fault conditions.

My question is ... When the NodeA fails, does the original instanceB on
NodeB simply "mount" InstanceA's databases and run rollforward recovery
on its logs. Or when failover occurs, must an entire startup of DB2
occur (including all the costs associated with it such as allocating
memory, setting up buffer pools, etc ...)?

Perhaps it depends on the HA software used (HACMP vs Veritas, etc.), but
I imagine DB2/UDB iteself needs some capabilities to support failover to
an already running instance.

Thanks in advance

Peter
--
Posted via http://dbforums.com