Hi,

Yeah, max rep makes a *lot* of difference when you can use it. We can't enable it as default for a lot of devices since it causes issues with some software revisions and/or snmp agent versions.

It's especially useful with higher latency polling, where it's beneficial to get more data with each request.

adam.

Sent from Mailbird

On 01/09/2015 01:54:35, Aaron Finney <aaron.finney@openx.com> wrote:

Actually, this has improved polling times significantly for all of the Juniper devices. E.g. a QFX 5100 48s has gone from 95s average to 20s, an MX480 from 80s to 25s.

On Mon, Aug 31, 2015 at 5:30 PM, Aaron Finney <aaron.finney@openx.com> wrote:
Update: the polling time on the QFX stacks has gone from ~450 seconds to ~250 seconds by adding an snmpbulkwalk max-rep value for junos to config.php:

$config['snmp']['max-rep'] = TRUE;
$config['os']['junos']['snmp']['max-rep'] = 70;

This value was not scientifically/precisely derived; I simply ran snmpbulkwalk against ifEntry with increasing values until I started seeing a significantly diminishing rate of improvement.



On Sat, Aug 29, 2015 at 9:54 AM, Aaron Finney <aaron.finney@openx.com> wrote:
Hello,

We've just started a POC project to determine if Observium is a good fit for our environment, and so far have been very pleased with the feature-richness and general usability of the entire system. In fact, we've discovered several things in our environment that were immediately addressable. Cheers!

One thing that's been a head-scratcher for us so far is the length of the polling time for our Juniper switches...anywhere from ~2 mins for a 48 port EX ToR switch, to 8 minutes for a two-member QFX aggregate switch. It's the port polling that's so costly according to the output of the poller script, around 450 seconds on the QFXs. I've tried some basic troubleshooting...disabling MAC accounting, disabling all ports for polling/alerting, etc...but it hasn't helped. Is there something simple I'm missing in the config that's causing it to bog down on the port module? Aside from the time, it's not hurting the QFXs, but our poor EX4500s sometimes spike to 90% CPU during polling. 

For context, the Observium POC system is located in the same datacenter as the devices we're polling against, and the QFXs are 96 port 5ks, with uplinks to our EX ToR switches. Any ideas/suggestions you have would be much appreciated!

Also, I did add support for Baytech PDUs; if there's interest, I'd be happy to provide the code.

Aaron




--
Aaron Finney | Network Engineer
888 East Walnut Street, 2nd Floor | Pasadena, CA 91101

Watch how we make online advertising simple: http://bit.ly/Ent_vid
www.openx.com   |   follow us on:   Twitter   Facebook    LinkedIn




--
Aaron Finney | Network Engineer
888 East Walnut Street, 2nd Floor | Pasadena, CA 91101
office: +1 (626) 466-1141 x6035

Watch how we make online advertising simple: http://bit.ly/Ent_vid
www.openx.com   |   follow us on:   Twitter   Facebook    LinkedIn

_______________________________________________ observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium