[closed] poller does not collect all hosts

[std]

hi all,

I've problem with poller. I created 32 hosts with snmp-support. Poller collect data for the first 12 hosts. The cactilog says

SYSTEM STATS: Time:298.5772 Method:cmd.php Processes:1 Threads:N/A Hosts:32 HostsPerProcess:32 DataSources:145 RRDsProcessed:61

What could by wrong?

openSUSE 11.4
Cacti Version: 0.8.7h
Plugin Architecture Version: 3.0

[tscoltrane]

I don't mean this to sound flippant, but many things could be wrong. You're taking nearly the entire polling interval (300s) for your run. Something's wrong. You're not polling that many hosts. So, while you can (and probably should) tune the poller settings (and/or use the alternative poller), I wonder if you've borked your SNMP implementation.

You should get on the host running Cacti (hoping that it's a Linux/BSD machine). From there, run snmpwalk against each of the target machines. If each run doesn't complete (cleanly) within a second or two, you have a problem. Don't open the entire tree. Just open the parts that you need (more secure, sure -- but it makes it run much faster). if this is "all good" (probably isn't), then look at the poller settings/tuning/etc.

[TheWitness]

There is a pretty serious bug in 0.8.7h. Please replace the file lib/utilities.php from SVN to correct the underlying issue. Then, repopulate your poller cache.

TheWitness

[std]

There is a pretty serious bug in 0.8.7h. Please replace the file lib/utilities.php from SVN to correct the underlying issue. Then, repopulate your poller cache.

TheWitness

I've replaced the lib/utilities.php with the build 6886. But no change.

Stefan

[TheWitness]

Well, I assumed anyway. Follow the debug steps in the second link on my signature. Let us know what you find.

[std]

Hello,

I changed to spine as poller and found this message

11/15/2011 04:00:02 PM - PHPSVR: Poller[0] ERROR: Input Expected, Script Server Terminating
11/15/2011 04:00:02 PM - SPINE: Poller[0] FATAL: Connection Failed, Error:'1040', Message:'Too many connections' (Spine thread)

Stefan

[TheWitness]

Edit 'my.cnf' and increase 'max_connections'. Spine takes quite a few, and then you have to leave overhead for users on the website.

connections = processes * ((threads + 1) + (script servers))

[std]

And now there is this warning

11/15/2011 05:05:01 PM - POLLER: Poller[0] WARNING: There are '1' detected as overrunning a polling process, please investigate
11/15/2011 05:05:01 PM - POLLER: Poller[0] NOTE: Poller Int: '300', Cron Int: '300', Time Since Last: '300', Max Runtime '298', Poller Runs: '1'

Stefan

[TheWitness]

You have a host that is misbehaving. Pleas follow the debug steps that include how to run spine from the command line.

[std]

There was a problem with a script startet by snmpd.

Regards