Queue jumps up over 300+ and all monitors pause for a while

[aarons]

Hi all,

My zabbix server is generally working great. I'm monitoring about a dozen machines. The load average on my zabbix server fluctuates between .5 and 1.0.

My problem is that it seems like all items sit in the queue and are never fulfilled until the queue reaches around 300+ items, and then they all go through and everything is updated.

A symptom is that an item set to update every 60 seconds will update at, for example, 1:01:00, 1:02:00, 1:03:00, 1:06:10, 1:06:11 - because it froze for three minutes, then flushed two instances of the item through really quickly as it was catching up.

This leaves gaps all through my graphs, especially when the lockup exceeds five or ten minutes and the queue gets huge before it flushes through.

I've increased the number of zabbix_server threads and increased the number of concurrent mysql connections to 200, but that didn't change anything.

Is the root of this problem in mysql? An update lock, perhaps, keeping it from updating the queue and locking everything out for a while?

[Alexei]

I believe this is all about unreachable hosts. Every unreachable host slows pollers down. I would suggest tuning of timeout related parameters in ZABBIX server's configuration file or, if possible, switching to active checks.