Polling problem since upgrading to .8.6h

[nuinco]

Ever since upgrading to Cacti 0.8.6h the polling cycle does not complete correctly. I have 40 devicesand 616 datasources on routers, switches, firewalls and Windows servers.

If I disable all devices except one then the polling cycle completes for that device. I have tried this on all 40 devices and 1 at a time always completes. But as I add more devices the polling hangs when it takes longer than 45 seconds. It can be any different combination of devices and datasources. It is not always exactly 45 seconds when it hangs, it has gone as long as 90 seconds but if it is over 45 seconds it always hangs at some point. Then I have to kill the php-win.exe processes or the system will not release the memory and the system becomes unusable after several failed polling cycles.

I have tried using Cactid. Sometimes it would be successful but other times it would hang. When it hangs the CPU is maxed at 100% and only rebooting the machine makes it usable again.

This sounds odd and I am close to trying a complete reload of Cacti. Everything worked well before updating, the polling cycle was at appr. 2.5 minutes for all devices and datasources. I only updated to get the Mactrack plugin functioning and it works fine.

Any help is appreciated.
Thanks

[BSOD2600]

1) you installed the patches for cacti yet?
2) tried cactid 0.8.6g?
3) tried using more polling processes (setting in cacti -> pollers) ?
4) tried increasing the php timeout values (in php.ini) ?

[nuinco]

Yes
Yes
Yes
Yes

All good questions but I did them.

Thanks

[BSOD2600]

When you turn your cacti log to debug for a cycle, do you see what script/query it hangs on? In the past, I've read posts where it hangs on WMI scripts.

[nuinco]

Yes, I can see what script it hangs on, but unfortunately it is different depending on the device that it stopped on. Sometimes it is the drive info, sometimes CPU and sometimes interface. All are the SNMP datasources.

[nuinco]

Something I do see in the log multiple times when the poller runs is this,
"01/16/2006 05:21:39 PM - CMDPHP: Poller[0] ERROR: Either there are no items in the cache or polling is disabled"

I did clear the poller cache but don't know why this continues to repeat? Polling is not disabled.

[BSOD2600]

If your poller cache keeps emptying itself, then it sounds like there is something wrong with your mysql database. Try running: mysqlcheck --auto-repair --databases cacti