Very Computer

by **Matilda Ang** » Wed, 03 Oct 2001 17:27:22

Hello Everyone,

I am running aix 4.3.3 in an SP environment and on some nodes the
snmpd CPU usage keeps growing. On one node, a 375 Mhz POWER3 wide
node, it has taken under 18 hours to reach 22.7% CPU usage.

COMMAND TIME ELAPSED %CPU
snmpd 08:03:27 17:43:23 22.7

It is also observed on a couple of S80's, however the rate of growth
is a lot slower but they will keep on going until they reach 100% CPU
usage.

The machines that display this are also running a dbsnmpd. I think
this may have something to do with it, but could also quite well be a
red herring.

Has anybody seen this before? Does anyone know how to cure it?

Thanks in Advance,

Matilda

by **koert5** » Thu, 04 Oct 2001 04:29:45

I have the same problem - dbsnmpd is an snmp subdaemon of Oracle (actually
it's the Oracle
intelligent Agent - ask your DBA about it).
At our site - once every two weeks this phenomenom occurs ... i just stop
the snmpd process
and start it again. When you stop snmpd, you'll notice that dbsnmpd goes to
100% cpu but when snmpd
is up & running again the situation stabilizes ... as far as I know the
Intelligent Agent keeps on running
correctly.

Cheers,
Kurt

Quote:> Hello Everyone,

> I am running aix 4.3.3 in an SP environment and on some nodes the
> snmpd CPU usage keeps growing. On one node, a 375 Mhz POWER3 wide
> node, it has taken under 18 hours to reach 22.7% CPU usage.

> COMMAND TIME ELAPSED %CPU
> snmpd 08:03:27 17:43:23 22.7

> It is also observed on a couple of S80's, however the rate of growth
> is a lot slower but they will keep on going until they reach 100% CPU
> usage.

> The machines that display this are also running a dbsnmpd. I think
> this may have something to do with it, but could also quite well be a
> red herring.

> Has anybody seen this before? Does anyone know how to cure it?

> Thanks in Advance,

> Matilda

by **koert5** » Thu, 04 Oct 2001 05:03:19

I checked on metalink - there are a number of things that can be wrong:
- there're quite some bugs in the early versions of intelligent agent
(8.0.5).
- if your DBA is checking events to frequently - this of course leads to
higher CPU usage
- if the intelligent agent is monitoring resources (DB's, listeners, nodes,
webservers) that do not exist
anymore the agent can spin on this - check the config files (snmp_rw*.ora in
$ORACLE_HOME/network/admin)
- if the DBA is running the CHUNK SMALL event and your DB has alot of
tablespaces/datafiles/segments then this
event is known to take up a lot of CPU - the DBA should tune SQL statement
in the CHUNK.tcl script (use the rule hint)
- but then again - I have intelligent agent 8.1.7.2 running on 24 databases
and 3 nodes, perfectly configured and every 2 weeks
i have to restart the AIX snmp subsystem....

Cheers
Koert

> I have the same problem - dbsnmpd is an snmp subdaemon of Oracle (actually
> it's the Oracle
> intelligent Agent - ask your DBA about it).
> At our site - once every two weeks this phenomenom occurs ... i just stop
> the snmpd process
> and start it again. When you stop snmpd, you'll notice that dbsnmpd goes
to
> 100% cpu but when snmpd
> is up & running again the situation stabilizes ... as far as I know the
> Intelligent Agent keeps on running
> correctly.

> Cheers,
> Kurt

> > Hello Everyone,

> > I am running aix 4.3.3 in an SP environment and on some nodes the
> > snmpd CPU usage keeps growing. On one node, a 375 Mhz POWER3 wide
> > node, it has taken under 18 hours to reach 22.7% CPU usage.

> > COMMAND TIME ELAPSED %CPU
> > snmpd 08:03:27 17:43:23 22.7

> > It is also observed on a couple of S80's, however the rate of growth
> > is a lot slower but they will keep on going until they reach 100% CPU
> > usage.

> > The machines that display this are also running a dbsnmpd. I think
> > this may have something to do with it, but could also quite well be a
> > red herring.

> > Has anybody seen this before? Does anyone know how to cure it?

> > Thanks in Advance,

> > Matilda

by **Matilda Ang** » Wed, 24 Oct 2001 20:47:31

I have found a permanent solution to this problem, apply the patch
associated with Oracle TAR 13402548.600. I found this when putting a
call into IBM. The problem stems from the Oracle dbsnmp daemon. The
patch has been applied by our Oracle support department and they said
it's not a big patch and they had no problem applying it. snmpd CPU
usage now stays at a nice 0.0 %. We are running with Oracle v8.1.6

I hopes this helps anyone who has observed the same thing, and thanks
for your comments Koert

Regards,

Matlida

> I checked on metalink - there are a number of things that can be wrong:
> - there're quite some bugs in the early versions of intelligent agent
> (8.0.5).
> - if your DBA is checking events to frequently - this of course leads to
> higher CPU usage
> - if the intelligent agent is monitoring resources (DB's, listeners, nodes,
> webservers) that do not exist
> anymore the agent can spin on this - check the config files (snmp_rw*.ora in
> $ORACLE_HOME/network/admin)
> - if the DBA is running the CHUNK SMALL event and your DB has alot of
> tablespaces/datafiles/segments then this
> event is known to take up a lot of CPU - the DBA should tune SQL statement
> in the CHUNK.tcl script (use the rule hint)
> - but then again - I have intelligent agent 8.1.7.2 running on 24 databases
> and 3 nodes, perfectly configured and every 2 weeks
> i have to restart the AIX snmp subsystem....

> Cheers
> Koert

> > I have the same problem - dbsnmpd is an snmp subdaemon of Oracle (actually
> > it's the Oracle
> > intelligent Agent - ask your DBA about it).
> > At our site - once every two weeks this phenomenom occurs ... i just stop
> > the snmpd process
> > and start it again. When you stop snmpd, you'll notice that dbsnmpd goes
> to
> > 100% cpu but when snmpd
> > is up & running again the situation stabilizes ... as far as I know the
> > Intelligent Agent keeps on running
> > correctly.

> > Cheers,
> > Kurt

> > > Hello Everyone,

> > > I am running aix 4.3.3 in an SP environment and on some nodes the
> > > snmpd CPU usage keeps growing. On one node, a 375 Mhz POWER3 wide
> > > node, it has taken under 18 hours to reach 22.7% CPU usage.

> > > COMMAND TIME ELAPSED %CPU
> > > snmpd 08:03:27 17:43:23 22.7

> > > It is also observed on a couple of S80's, however the rate of growth
> > > is a lot slower but they will keep on going until they reach 100% CPU
> > > usage.

> > > The machines that display this are also running a dbsnmpd. I think
> > > this may have something to do with it, but could also quite well be a
> > > red herring.

> > > Has anybody seen this before? Does anyone know how to cure it?

> > > Thanks in Advance,

> > > Matilda