I started the day with several devices that I defined years ago working, their graphs recording, but everything new I added simply stayed in the status "Unknown".
I attempted to follow the advice here:
http://docs.cacti.net/manual:087:4_help ... #debugging
I did not read far enough down - I ran /usr/bin/spine --verbosity=5 from the command line 3 times within 10 minutes - making all my working graphs stop working. Rather than attempt to fix all the rrd's i'd just broken, I decided to start over. When working properly, it only takes me a few hours to put in the devices I need, and creating graphs from there is pretty simple
At this point, I have removed all the deivces and attempted to re-create them. They stay "unknown" unless I run spine manually. I have verifiied that /etc/cron.d/cacti is set to call poller.php every 5 minutes just as in the example given in the link above. If I run spine as root, the devices all show up as Up, cacti creates a few graphs, but all stay empty. I get errors if I run spine as cacti.
To summarize:
Check Cacti Log File
Setting the log level to debug, I never see the poller run unless I run it. I see no snmp timeouts.
Check Basic Data Gathering
I can do SNMP get's to any device I try.
I have to confess I don't know what .pl to run to check perl processing, I need more guidance here.
Check Cacti's Poller
I have not tried cmd.php, because we've been using spine for quite some time.
Setting to debug, I never see spine process, and I see no errors as to why in the cacti log.
If I run spine manually (as root), it appears to process fine, my devices chang eto Up, but the RRD's fill with nan.
The cacti user (cacti) has the correct crontab settings.
Running spine as the cati user form the command prompt results in errors about not running as root.
Check MySQL Update
No idea how to do this - if it helps, I see no SQL errors in the cacti log.
Check RRD File Update
Files are updating every time I run spine. Updating with nan's, but hey are updating.
Check RRD File Ownership
All the rrd's are owned by cacti.
Check RRD File Numbers
when I ran this, there are no ds[loss]. entries - there are other .min and .max entires, and they are correct now. I think running spine 3 times in 10 minutes caused the problem described in this section of the troubleshooting instructions, but at this point I've deleted and re-created all these files, all appear fine now. Other than filling with nan's
Check RRDTool Graph Statement
Graph Statements are fine.
I checked for multiple crontab entries, the only crontab that had the poller.php in it was the cacti crontab.
Oh, I deleted all the old rrd's as well, not just the devices and graphs, there were some abandoned RRD's from several years ago that really needed to go anyway.
I have not modified the poller.php in any way that I am aware of - and the fact that before I started messing with things my old graphs were working indicates that at least SOMETHING was happening.
Other things to note - I shut off three core switches about three weeks ago, and cacti said they were Up before I deleted their objects ths morning - indicating that spine has not been running for some time.
We just forklifted our entire LAN, I thought perhaps I was having snmp issues, but in fact when I run a verbose quesry of any device, I get full detail.
Any help would be appreciated - Cisco sold us Ciscoworks along with our forklift upgrade, assuring us that it would do what Cacti does and more. It does not. I need to get things going in here soon.
