Random gaps in the graphics of Host MIB - Hard Drive Space

[toshko3]

Whenever I create a graphic for 2 partitions (for example) and physical and virtual memory I get the following graphic. The log shows no errors but random "output: 0". I can't understand why.
From time to time it fails with no reason in the log. One time:

Code: Select all

 05/28/2010 04:50:05 PM - CMDPHP: Poller[0] Host[28] DS[193] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get total 6, output: 1072758784
05/28/2010 04:50:05 PM - CMDPHP: Poller[0] Host[28] DS[193] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get used 6, output: 1064435712
05/28/2010 04:50:05 PM - CMDPHP: Poller[0] Host[28] DS[192] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get total 5, output: 1776025600
05/28/2010 04:50:05 PM - CMDPHP: Poller[0] Host[28] DS[192] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get used 5, output: 1721630720
05/28/2010 04:50:05 PM - CMDPHP: Poller[0] Host[28] DS[191] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get used 3, output: 76311912448
05/28/2010 04:50:05 PM - CMDPHP: Poller[0] Host[28] DS[191] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get total 3, output: 229287903232
05/28/2010 04:50:04 PM - CMDPHP: Poller[0] Host[28] DS[190] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get used 2, output: 0 

The other time:

Code: Select all

05/28/2010 04:49:02 PM - CMDPHP: Poller[0] Host[28] DS[193] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get total 6, output: 1072758784
05/28/2010 04:49:02 PM - CMDPHP: Poller[0] Host[28] DS[193] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get used 6, output: 980287488
05/28/2010 04:49:02 PM - CMDPHP: Poller[0] Host[28] DS[192] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get total 5, output: 1541144576
05/28/2010 04:49:02 PM - CMDPHP: Poller[0] Host[28] DS[192] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get used 5, output: 1412366336
05/28/2010 04:49:02 PM - CMDPHP: Poller[0] Host[28] DS[191] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get used 3, output: 76311912448
05/28/2010 04:49:02 PM - CMDPHP: Poller[0] Host[28] DS[191] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get total 3, output: 229287903232
05/28/2010 04:49:02 PM - CMDPHP: Poller[0] Host[28] DS[190] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get used 2, output: 9329348608
05/28/2010 04:49:02 PM - CMDPHP: Poller[0] Host[28] DS[190] SERVER: /usr/share/cacti/site/scripts/ss_host_disk.php ss_host_disk 192.168.2.105 28 2:161:500:1:10:public:::::: get total 2, output: 21706477568 

Heeeelp, please!
It only happens on particular XP devices and random parts of the "Host MIB - Hard Drive Space" - one time on the hdd, the other on the physical ram, the other on virtual..., whenever I recreate the graphs.

[gandalf]

Windows SNMP is quite slow. So first bet is to increase the SNMP timeout for that host to e.g. 1000 ms
R.

[toshko3]

Windows SNMP is quite slow. So first bet is to increase the SNMP timeout for that host to e.g. 1000 ms
R.

OK but won't it be a problem, because i've set the poller to 1 minute?

[toshko3]

I've done the 1000ms timeout, but no change. Thanks for the suggestion anyway. Any other suggestions?

[beeker41]

Check your maximum database connections. I had this and had to increase them from the default in MySQL and it fixed it.

[BSOD2600]

I've done the 1000ms timeout, but no change. Thanks for the suggestion anyway. Any other suggestions?

Try something higher... 3000 or 5000ms

[Linegod]

I had this issue with Windows Servers for quite some time. Nothing seemed to help.

I finally just switched to gandalfs pure SNMP template: http://docs.cacti.net/usertemplate:data ... disk_usage - this seemed to clear up the disk issues.

(And I haven't mentioned lately how bad of an idea I think 1 minute polling is. It's a bad idea. There, I'm happy now).

[toshko3]

Check your maximum database connections. I had this and had to increase them from the default in MySQL and it fixed it.

Already done that: max_connections = 5000

[toshko3]

I've done now the 5000ms timeout and ... waiting...
I'll post the results soon

[toshko3]

Now the graphs are without gaps but the physical ram is all empty, although I recreated the graphs. The Linegod's graphs are OK but they don't show the drive labels at the graph label. If anyone knows how to add them, please tell!
This is all with 5000ms timeout of the device. (This is too much I think)

[toshko3]

OK, now the physical ram is shown but the virtual is messed...

[BSOD2600]

OK, now the physical ram is shown but the virtual is messed...

You've validated those numbers via another method to know this?

[toshko3]

Well, yes. The above template works just OK. Moreover - as you may see - the physical ram is missing some time in the graph. Just when it appears - the virtual messes up.
For now around half a day they are OK. I don't know. Will wait to see and post tomorrow.

[TheWitness]

Now the graphs are without gaps but the physical ram is all empty, although I recreated the graphs. The Linegod's graphs are OK but they don't show the drive labels at the graph label. If anyone knows how to add them, please tell!
This is all with 5000ms timeout of the device. (This is too much I think)

This timeout issue has absolutely nothing to do with Cacti, or whether or not it's a script only solution or snmp only solution. The bottom line is that the MS Agent it pitiful with regard to the polling of disk partitions.

The only good fix for this might be to move away from Microsoft's Agent entirely. I know that Net-snmp.org is now providing an agent. I don't know if it will suffer from the same performance issues or not yet.

Regards,

TheWitness

[TheWitness]

Well, yes. The above template works just OK. Moreover - as you may see - the physical ram is missing some time in the graph. Just when it appears - the virtual messes up.
For now around half a day they are OK. I don't know. Will wait to see and post tomorrow.

Here is the absurdity of the MS Agent. If you install a thumb drive into one of your USB ports, the Physical and Virtual graphs get messed up. Remove the USB stick, and all is better.

This goes back to a discussion that was had on the Net-snmp forums some time ago to NOT allow "temporary" mounts, like auto-mounted NFS volumes to be included in the Host Mib. I think it's a design flaw in the agent, but Cacti could handle it better too.

TheWitness