gaps

[btomjack]

Well, im not sure if this is directly covered somewhere else i did some looking and didn't see anything, so here goes.

Cacti is setup on a Linux server that is on an special internal network, and one of the hosts it polls is the firewall box that is up one level. this firewall box has 3 Ethernet interfaces eth0 the internet, eth1 is the intranet, and eth2 is a special protected intranet. eith 1 and 2 are making prefect graphs. (as well as all other 100 hosts the cacti is monitoring) but for some reason on the eth0 interface graph 3 days ago it started getting weird gaps in the inbound traffic. the outbound traffic still displays without gaps on the same graph. none of the other interface graphs on that box have any problems, and they have a much heavier load on them as compared to the one that is gapping. attached is a pic of the gaps on the daily, any insight suggestions, or debugging tips is greatly appreciated


note: the gaps are false, there is traffic over the interface at all times during the gaps, i have verified that much, and the box doing the polling is not under heavy load.

[TheWitness]

I would suspect that you are execeeding your 125mbps limit on 32 bit counters. You should verify if you device has 64 bit counters. It it does, you should switch to those. If not, you should poll the interface every minute instead of five.

In the case of Cacti 0.8.7a, you would have to implement 1 minute polling and setup your data template to do so as well. Then you would have to trash your old RRDfiles.

TheWitness

[btomjack]

I hadn't even thought about that, they are very close to that level. that is likely the problem. ill do a little changing around with the polling and see what happens. thanks much!