Multiple polling instances?

billn

23-03-2005 13:08:18

I've got an enterprise network with two distinctly segregated topologies, both attached to a third. I'd like to run multiple polling instances, on in each of the segregated areas, feeding to the main web server/database instance in a third.

Is there a way to build a standalone netmrg-gatherer poller, configured to only poll devices in a specific group? Something as simple as
[code1e5b004fd09]
Zone 1 poller:/etc/netrmg.xml
<polling>
<interval>300</interval>
<group>Core Devices</group>
<group>Border Routers</group>
</polling>

Zone 2 poller:/etc/netmrg.xml
<polling>
<interval>300</interval>
<group>CPE Devices</group>
<group>Customer Aggregation Devices</group>
</polling>


Zone 3 poller:/etc/netmrg.xml
<polling>
<interval>300</interval>
<group>Firewalls</group>
<group>Office Devices</group>
</polling>
[/code1e5b004fd09]

Great work on the poller, btw.

balleman

23-03-2005 19:13:41

The biggest drawback to this is that all RRD files need to be accessible on the web server host. Having the gatherer be configurable to only poll specific groups would involve an NFS mount or similar arrangement, which would probably negate the benefit of the distributed polling. Do you agree with my assessment? I'd love to have distributed polling, but unfortunately I think that most times the biggest bottleneck in the process is RRDTOOL disk I/O.

billn

24-03-2005 12:17:48

[quoted69975b951="balleman"]The biggest drawback to this is that all RRD files need to be accessible on the web server host. Having the gatherer be configurable to only poll specific groups would involve an NFS mount or similar arrangement, which would probably negate the benefit of the distributed polling. Do you agree with my assessment? I'd love to have distributed polling, but unfortunately I think that most times the biggest bottleneck in the process is RRDTOOL disk I/O.[/quoted69975b951]

Yeah, I realized this after I posted yesterday. I'm working on a solution. }=)