We're in the process of upgrading to version 8.2 from 7.2 (with hopes
that most of the issues we're currently facing will be fixed...). In
the interim, we have a new problem that has surfaced that is causing
us severe grief. A search of this newsgroup archive didn't turn up
anything. Hoping that someone else has seen this before.
Our db2agent pool is being exhausted for an unknown reason. We are not
running with INTRA_PARALLEL turned on. Generally we never see more
db2agent processes than connections to the database. We usually see an
agent attached to a database or in an idle state. What we end up
seeing is multiple db2agent processes appearing like this:
db2agent (instance)
All of these processes are owned by db2tcpcm or db2ipccm. Our
MAXAGENTS is 400. With 150 applications connected to the database
we'll see 250 of the "db2agent (instance)" processes. The rest of the
agents are linked with the database or idle.
I haven't figured out a graceful way to make these guys go away.
(Stopping and restarting the instance seems to be the only way so
far.) As long as they're there, no new connections can be obtained
into the database. (I can't even do a list applications at this
point.) It appears that anyone currently connected remains connected,
but all future connections are met with an error.
Anyone ever seen anything like this before?
The application connections come from Websphere's connection pool with
db2 client software 7.2 FP 8 on NT 4.0 service pack 6. The back end is
DB2 7.2 FP8 on AIX 4.3.
Thanks in advance,
Evan