Re: [nv-l] Ping is falling behind

[Stephen Hochstetler <shochste@us.ibm.com>]

There is a bit of an art to this since there are so many factors that could be involved.

1) What hardware are you using? How many processors does it have? A great answer is 4 for a very busy NetView server.

2) In a similar situation I found that installing a DNS on the NetView server greatly reduced overall CPU requirements. You should have your external DNS resolve anything your internal DNS cannot resolve. They key to reduced CPU requriements was for us to build the DNS files from NetView information. For example, if you do a lookup by name, it should resolve the address that is the 'management' inferface for that device. For the reverse lookup, ALL interfaces should resolve to the same name. Most DNS only have one interface for both name and reverse lookup.

3) Are you using MLMs? If so and are distributing polling I have seen this affect the polling cycles as netmon discovers new nodes.


Stephen Hochstetler shochste@us.ibm.com
International Technical Support Organization at IBM
Office - 512-838-6198 (t/l 678) FAX - 512-838-6931
http://www.redbooks.ibm.com
"Bursik, Scott {PBSG}" <Scott.Bursik@pbsg.com>

"Bursik, Scott {PBSG}" <Scott.Bursik@pbsg.com>
Sent by: owner-nv-l@lists.us.ibm.com

03/15/2006 09:13 AM

Please respond to
nv-l@lists.us.ibm.com

To	<nv-l@lists.us.ibm.com>
cc
Subject	[nv-l] Ping is falling behind

NetView 7.1.4 AIX 5.2

* I am using the script below to check the ping status of NetView and I
am seeing numbers around 30,000 at times. I am having problems with
nodes going down for reboots and coming back up within about 5 minutes
and I am getting node down events. I have a timer rule in place that
says if I get a node down event hold it for 14 minutes. If I get a node
up event drop the node down. If the 14 minutes expires and no node up
event came in send the node down event.

* Well the nodes are coming back well before the 14 minutes and we are
still getting the node down. I believe that NetView being behind in
polling is causing this. Are there any suggestions as to fixing the
polling falling behind issue?

* Within the last year SNMP was enabled on our workstation machines
resulting in about 7000 new objects being discovered by NetView. Since
they have the same OID as the Windows servers they are brought into
discovery. I want to restrict them from being discovered by name. They
all have a common set of characters in the name. The naming convention
is ***wu******  or  ***WU******. The same number of characters are used.
If I read the documentation in the seed file correctly I could use this
login correct?

!???wu??????
!???WU?????

Am I interpreting the question marks correctly from the documentation?

 #      Negative Entries
 #          e.g.        !10.1.1.2                  Specific entry
 #                      !10.*.1.1-100              Ranges using * or -
 #                      !router*.tivoli.com        Wildcards using * and
?(single char)
 #                      !@oid 1.3.6.1.4.1.9.*      Wildcards as final
char using *
 #                                                 (Note space after the
prefix "@oid ")
 #                      !@oid 0                    This entry will
filter out all
 #                                                 non-SNMP supported
devices

* I have unmanaged as many devices as I can that aren't being monitored
at the moment and have restricted discovery to several OID's. My DNS in
on another machine but on the same local subnet. Has anyone here written
a script to test the speed of resolution from DNS?


Here is the result of the ovobjprint -S command:

ovobjprint -S
Number of objects defined in the database: 93939

Total number of fields defined in the database is: 306.

Total number of field values in the database: 1958450
         Number of Integer fields: 511561.
         Number of Boolean fields: 754239.
         Number of String fields: 521397.
         Number of Enum fields: 171253.


##############################script####################################
##########
#!/bin/ksh
#set -x
cat /dev/null > /usr/OV/log/netmon.trace
/usr/OV/bin/netmon -a 12
sleep 3
if [ -f /usr/OV/log/netmon.trace ]; then
  echo "Netmon is " `grep [-].*[:] /usr/OV/log/netmon.trace | wc -l `
"behind in status pinging";
else
  echo "Netmon is too busy to report now. Try later."
fi
exit
##############################script####################################
##########

I know this is a lot of information and I appreciate any and all
feedback I get.

Thank you,

Scott Bursik
PepsiCo