2

Why would I'd be getting gaps like this in graphs coming from Windows hosts?

Graph showing gaps

This problem seems to happen quite regularly on graphs coming from Windows hosts. The graph template for the graph above is "Host MIB - Available Disk Space", though this problem happens on all sorts of Windows graph templates.

SimonJGreen
  • 3,205
  • 5
  • 33
  • 55
  • Have you considered just firing up wirehark and filtering for snmp traffic destined for that host? What happens? – Zoredache Apr 24 '12 at 16:53

1 Answers1

1

If cacti gets no response from a query, it does not try again; the data for that time period is missing.

Check your cacti logs for clues or errors.

adaptr
  • 16,576
  • 23
  • 34
  • So this is likely caused by Windows SNMP agent not responding correctly? Can Cacti be made to try again? – SimonJGreen Apr 24 '12 at 12:20
  • SNMP retries are supported since 0.8.6i; in Console/Settings/General you can set the timeout and retry options. – adaptr Apr 24 '12 at 12:46
  • Oh retries is already set to 3. That's upsetting! I've increased the timeout from 0.5s to 1s, though that seems a little extreme for SNMP. – SimonJGreen Apr 24 '12 at 12:56
  • If retrieving this data takes more than the timeout every time, it will never get data... consider setting the timeout to 5 seconds ? As long as all your data queries *combined* don't exceed 5 minutes, it doesn't matter. Also increase the number of parallel queries to 10 or more, to make the best use of the time you have. – adaptr Apr 24 '12 at 12:58
  • Ah ok, so that timeout is the window for all queries going to that host, not just individual queries? SHOULD a windows host take this long to respond though? This happens with most of the windows hosts we monitor. – SimonJGreen Apr 24 '12 at 13:06
  • Where did I say any of that ? – adaptr Apr 24 '12 at 13:15
  • you didn't? they're questions not statements... – SimonJGreen Apr 24 '12 at 13:40
  • Increasing the timeout period on all hosts worked. Thanks. – SimonJGreen May 23 '12 at 10:18