NaN in graphs

[wbierman]

Greetings. Many thanks to the developers for such a wonderful product!

I am experiencing the dreaded NaN graphs when querying interface errors from my 3com switch. Traffic graphs work normally from the same device and interface.

I have followed the steps laid out in this article: http://docs.cacti.net/manual:087:4_help ... #debugging but have not yet found my problem. I am hoping someone with more experience with Cacti can show me what I've missed.

When I execute snmpget manually, I receive the following information:

IF-MIB::ifInErrors.67108993 = Counter32: 0

However, when I use rrdtool to show me what it is receiving, I see the following:

aeo:/usr/share/cacti/site/rra# rrdtool fetch core_switch_7700_errors_in_89.rrd AVERAGE
errors_in discards_in discards_out errors_out

1308602400: nan nan nan nan
1308602700: nan nan nan nan
1308603000: nan nan nan nan
1308603300: nan nan nan nan

I'm assuming that since the rrd file is being updated with nans that it is not a permissions issue. However the file is owned by the same user the poller is running as in crontab.

It seems that my manual SNMP query is receiving a sane value (though I would expect the value to be greater than zero) but rrdtool is not receiving a sane value.

Can anyone offer any insight on what I should check next?

Thanks!

Edit: Upon closer examination, when I run the rrdtool update command as root, the zeros are added to the rrd file. When run as www-data (the user the poller runs as), nans are added. Could it be that the www-data user does not have access to query SNMP data or some such thing? And if so, why do the graphs of traffic work perfectly? Note also that when I use rrdtool updatev instead of rrdtool update, both as root and as www-data I receive a return_value of 0.

[gandalf]

NaN is a perfectly valid value for rrdtool. It simply tells you, that there has been no data. But that is "valid".
You surely are expecting data from your target. To debug, please see 2nd link of my sig
R.

[wbierman]

The second link in your page is precisely the link I posted as having gone through already.

Even the zeros don't really make sense. There should be some errors happening all of the time.

[gandalf]

Edit: Upon closer examination, when I run the rrdtool update command as root, the zeros are added to the rrd file. When run as www-data (the user the poller runs as), nans are added. Could it be that the www-data user does not have access to query SNMP data or some such thing? And if so, why do the graphs of traffic work perfectly? Note also that when I use rrdtool updatev instead of rrdtool update, both as root and as www-data I receive a return_value of 0.

A return value of 0 for rrdtool update says nothing. When feeding 0 or NaN to rrd file, both times the return value would be 0 (== everything is ok).
So we need to know, whether the pooling user does receive a 0 or "U" (== NaN). And it is not encouraged to switch between root and any other user. It may happen, that a rrd file belongs to root and can't be updated through the cacti polling user
R.

[wbierman]

I'm not sure what you're asking me for. I already stated that the polling user receives a return value of zero. And it wasn't the return value that I thought was noteworthy, necessarily, but the fact that that is ALL I received.

[gandalf]

Yep, I know.
But next, it would be necessary to follow the steps that describe verifying the "rrdtool update" process.
At least, when receiving "0", the data source max should not be an issue.
But e.g. file permissions may be an issue or a mismatch between data template and rrd file
R.