Zabbix 2.2.1 Pollers 100% busy

[madfrog]

Hi guys,

Recently I upgraded from zabbix 2.0.9 to zabbix 2.2.1 running on a more powerful server.

At first everything looked good with only several servers monitored. However, after I added more servers and more items, the zabbix pollers and trappers started getting busy, sometimes can easily go up to 100% and get many "hosts unreachable" alerts. Here is my current servers stats:

Number of hosts (monitored/not monitored/templates) 456
Number of items (monitored/disabled/not supported) 31842
Number of triggers (enabled/disabled) [problem/ok] 16825
Required server performance, new values per second 437.09

And here is setting I currently use:
### Option: StartPollers
# Number of pre-forked instances of pollers.
#
# Mandatory: no
# Range: 0-1000
# Default:
# StartPollers=5
StartPollers=300

### Option: StartIPMIPollers
# Number of pre-forked instances of IPMI pollers.
#
# Mandatory: no
# Range: 0-1000
# Default:
# StartIPMIPollers=0

### Option: StartPollersUnreachable
# Number of pre-forked instances of pollers for unreachable hosts (including IPMI).
#
# Mandatory: no
# Range: 0-1000
# Default:
# StartPollersUnreachable=1
StartPollersUnreachable=5


### Option: StartTrappers
# Number of pre-forked instances of trappers.
#
# Mandatory: no
# Range: 0-1000
# Default:
# StartTrappers=5
StartTrappers=25

### Option: StartPingers
# Number of pre-forked instances of ICMP pingers.
#
# Mandatory: no
# Range: 0-1000
# Default:
# StartPingers=1

### Option: StartDiscoverers
# Number of pre-forked instances of discoverers.
#
# Mandatory: no
# Range: 0-250
# Default:
# StartDiscoverers=1

### Option: StartHTTPPollers
# Number of pre-forked instances of HTTP pollers.
#
# Mandatory: no
# Range: 0-1000
# Default:
# StartHTTPPollers=1

Is there anything I can tune to make the pollers less busy? Do you guys have any idea how many pollers and trappers I should use based on the number of servers/item I am monitoring?

Another thing is those servers are being monitored by my old zabbix 2.0.9 as well and the old box does not have a single problem with the busy poller process even with less number of pollers it is used. Is it related to the upgrade to zabbix 2.2.1? Or the dual monitoring server together could cause the poller busy problem? BTW, all of my zabbix agents are using 2.0.9 still.

Thanks,
madfrog

[Jason]

I've just had the same problem on our server when upgrading from 2.0.11 to 2.2.2

Did you find a solution?

[Jason]

I increased number of pollers and problem went away.