Monitoring current traffic per VIP

I think it's a global setting only but not 100% sure.

Sorry I know this is a dumb question, but does the little "m" in the graph indicate milli??

 

 

So if I've got 250m connections/sec then that equates to 15 connections/min???

 

 

Thanks.

 

 

Andy

 

 

 

 

 

That's correct.

Hi Jason,

 

 

I've got the cacti template for the F5 working great! Will there be any future plans to add monitoring for the nodes so we can find out the traffic, connection rate and current connections per node too?

 

 

Thanks.

 

 

Andy

I'm always looking for suggestions to expand the template. The nodes are probably pretty easy, pool members will be more challenging.

I have some perl code that does this... Pulling the data is easy... (Although my code uses iControl, not SNMP).

 

 

I think what's needed most is a way to have the cacti server poll the F5 every X seconds (e.g. 86400, 1 day) and automatically discover the VS's, Pools and poolmembers, nodes etc and automatically start gathering stats (And stop) based on pool membership, etc...

 

 

H

That would be incredibly useful. Does cacti have an API for actions like this? I haven't looked that deeply. Without, would be some serious DB work to make it happen.

I'd have to look. I suspect it doesn't and would require direct manipulation of the cacti database...

I'm getting strange results for the number of current connections to our dns virtuals. We have two dns virtuals and the first one shows no current connections where as the second virtual always display this weird graph shape for the same time period.

 

 

Any ideas??

 

 

 

Is your cacti server overloaded? The polling on cacti isn't very intelligent, and breaks easily under load. If your poll times go much above 30 seconds then you're heading into territory that will result in dropouts (Manifested usually as missing data).

 

 

BTW.. That looks like a lot of connections... I'd drop the idle connection timeout for udp/53 down to about 10 seconds... (Especially if they're not doing recursive queries, the first timeout for DNS is 5 seconds so even going down to 10 should ideally be too long to cause any problems).