01-Jun-2005 - MRTG Errors

Announcements concerning Networking & Related News, Planned Outages, Anything which may affect your services.

Moderator: Admins

Post Reply
porcupine
Site Admin
Posts: 703
Joined: Wed Jun 12, 2002 5:57 pm
Location: Toronto, Ontario
Contact:

01-Jun-2005 - MRTG Errors

Post by porcupine »

Hi Guys,

As many customers may have noted, some of our MRTG graphs have been "broken" to an extent for the past several days. While all of the graphs will have displayed accurate information for the majority of their polls (90-95% of the stats on the graph), some graphs were experiencing a few "glitches" due to SNMP problems with sw01, and msfc01.

Users with affected graphs will notice drops in their graphs (to 0 bits/sec in/out), that did not actually occur in any capacity, and/or spikes on their graphs (often in excess of the interface threashold, eg. 325mbps on a 100mbps fast ethernet connection).

Generally speaking, customers who utilized a fair amount of bandwidth (1mbps+) saw drops on their graphs, whereas customers who utilized very little bandwidth would have seen potentially massive spikes.

We have been working on the issue, and it's simply a matter of the SNMP polls timing out before a given cycle can complete. We've been tweaking several of the settings on our mrtg.prioritycolo.com server, removing duplicate checks on several of the switch/router interfaces, and lowering the number of individual statistics we collect (what can I say, we also love our graphs, especially CPU usage, temperature, and other highly relavent graphs).

Customers who have seen spikes on their MRTG's that they did not contribute to, need not worry about being billed for said spikes (anyone due for overages will have their graphs checked out by hand, the spikes removed, and will then have their billing processed). Any customers who have seen repeated drops to 0 bits/sec on their graphs also need not worry, as this is simply a graphing issue, and has never been service impacting. Hopefully we've got the issue resolved now, though we're still implamenting a few tweaks to try and prevent it from occurring in the future.

Thanks for your patience,
Myles Loosley-Millman
Priority Colo Inc.
myles@prioritycolo.com
http://www.prioritycolo.com
Post Reply