User`s guide

on
dimes.unx.sas.com
because
dimes
is not a router. The
snmpCollect
daemon ignores this metric for
dimes and proceeds normally with collecting the rest of the metrics for dimes.
Another kind of message you may see is like this:
Tue Jun 06 11:41:40 1995 : swimmer.unx.sas.com doesn’t reply to a 100 object
but responds to sysUpTime. Be sure SNMP timeouts are not
set too small (SNMP interval: 0.80s retry: 3).
In this example, the message indicates that the swimmer host could not respond within the timeout period.
The snmpCollect daemon bundles multiple requests in a single Protocol Data Unit (PDU). Depending on
your network, the agent on the host may not be able to respond within the timeout and retry parameters. There are
two ways around this problem:
The better fix is to reduce the number of requests that snmpCollect makes per PDU. The file
/usr/OV/lrf/snmpCol.lrf specifies the number of requests that snmpCollect makes per PDU.
The following steps describe how to edit the file and then replace the existing file with the edited file:
i. Make a backup copy of /usr/OV/lrf/snmpCol.lrf by issuing this command:
cp -p /usr/OV/lrf/snmpCol.lrf /usr/OV/lrf/snmpCol.lrf.orig
ii.
Change the number of objects per PDU by editing the file
/usr/OV/lrf/snmpCol.lrf
.
Specify the number nn of objects by adding
-m
nn
to the third field on the second line of snmpCol.lrf. The value of nn to use depends on your
network load. A larger number means snmpCollect can "batch" more requests and improve
performance of snmp get requests.
The updated file should look something like this:
snmpCollect:/usr/OV/bin/snmpCollect:
OVs_YES_START:trapd,ovwdb,ovtopmd:-m 20:OVs_WELL_BEHAVED:120:
iii.
Stop the
snmpCollect
daemon with this command:
/usr/OV/bin/ovstop snmpCollect
iv. Re-register snmpCollect with this command:
/usr/OV/bin/ovaddobj /usr/OV/lrf/snmpCol.lrf
v. Re-start the snmpCollect daemon with this command:
/usr/OV/bin/ovstart snmpCollect
vi. Verify that snmpCollect was properly started:
ps -ef | grep snmpCollect
You should see an entry containing, for this example,
snmpCollect -m 20
vii. Check the trace file, in the same way as described earlier in this action, for messages.
Another way to fix the timeout problem is to increase the global default timeout by following this path