[mrtg] data collection did not complete within interval

[Steve Shipway]

Steve Shipway
ITS Unix Services Design Lead
University of Auckland
Floor 2, 58 Symonds Street
09 3737599 ext 86487


> from 2 to 4 to 18, to no avail. Can someone comment on the different
> performance implications of increasing the Forks param vs. running
> multiple mrtg instances.

With multiple MRTG instances, you are able to control the sequence of checks (sometimes, you may want to have caching plugins) which you are not with forks; however in general forks is the best way to go.  In all cases, use daemon mode as this makes big savings on IO, particularly if you have RRDTool 1.4.x

> There are errors such as the following:
> 2010-08-18 15:23:54 -- 2010-08-18 14:54:54: ERROR:
> Target[sometarget][_IN_] ' $target->[671]{$mode}  * 1.8 + 32' (warn):
> Use of uninitialized value in multiplication (*) at (eval 1344) line 1.

This is a knock-on error caused by the previous SNMP query failing.

It looks like you are converting the degrees Farenheit -- you might like to take a look at the Routers2 extended option c2fi which can do this at display time for you, instead of converting at data storage time (which will allow you to store the data in SI units).  Of course this requires you to be using RRD and Routers2.

> Am I reading this correctly that this failing query took 29 minutes?
> What would prevent that from timing out? The mrtg snmp libraries use
> the default timeout/retry/backoff parameters.

The first timestamp is when the log entry was written by the loghandler; the second when it was generated.  The large discrepancy is a bit strange, but doesn't indicate that the SNMP query took this long to fail. Check any earlier SNMP failure log entries to find out for this.

For large swathes of 'unknown' in your RRD files, I'd suggest you also check the local system clock is correct and not jumping about (if you have NTP problems, or if it is running on a VM).

Also, try running with a greatly reduced set of Targets, and see what happens.  If you still get unknowns then the problem is going to be down to something else.  Verify that SNMP queries are still working form the MRTG server to the device (such as by using snmpwalk).

Steve