Dear all,

I got a problem on db2 for aix running a high workload messaging
system with more than 5,000,000 of deliveries per day. During high
peak hours it happens frequently that the application,
the application threads, are all waiting for a commit to complete.
This seems a typical case of a
missing or locked shared resource. The lock anyway is not at row or
table level during an SQL operation, but it is at commit time. In
fact, different applications at the same time are affected by
this problem. The db2 is v8.1 fixpak 14 and aix is 5.3 fixpak 05 with
5 processor (power5+ 1.6Ghz)
and 64 GB of memory. While the commit hang the db2 machine has no I/O
wait (no disk usage),
plenty of memory and processing power and plenty of network resources
(that's why I suspect
a lock).

To complete the scenario, the databases (same hw, same sw) are two in
continuos replication. The apply programs are running at the target
database. MIN_COMMIT is 5

The question is which monitoring parameter should I check to
understand what happen to
the db. Is there any operating-system resource that I can check?

As last thing, why the rootvg disks are working a lot? iostat says
that they write 8/10 MB per
second... I suspect something related to authentication due to
continuos connect/disconnect of replication, but I don't understand
what it matters with commit.

Thanks in advance
Matteo Gelosa