Node watcher process under high load

[simon123]

I recently added some monitoring on performance and performance tuning to my Zabbix (DM) setup.

I saw that the node watcher process (key: zabbix[process,node watcher,avg,busy]) kept being really busy.
a 50% average with peaks to 100% every hour (peaks of about 15 minutes).

Is there any way to tune my setup further to give the node watcher process some more breathing space.
Alter some poller/setting maybe?

FYI: Excep the timer process, the rest of my poller and internal processes run fine (under 40% avg).

i have them set up as following:

startpollers -> 10
startpingers -> 10
startdbsyncers -> 6
startpollersunreachable -> 3

my nodes have a load of about 40 VPS (on average).

[ataylo13]

I am seeing similar usage on my "node watcher" metric. Not sure how to increase the number of "watchers" though. So tagging along.

[rodom77]

Did you ever get an answer to this question? I'm seeing the same issue on the latest version of Zabbix.

What can be done to give more resources to the node watchers?

[shanghan]

i modify CONFIG_NODEWATCHER_FORKS=2,then start zabbix_server
the problem again.
Zabbix node watcher processes more than 75% busy.still 100%
how can i fix that..
thanks

[shanghan]

now fix the problem.

at config zabbix DM,have to problem:
1.Node watcher process under high load?
2.Master node can't recive child node history data?

fix:
compare master and child node hardware,and compare master and node my.cnf infomation.
example master node use innodb,and the my.cnf some infomation:
innodb_log_file_size = 32M
innodb_flush_log_at_trx_commit = 1
innodb_file_per_table = 1
innodb_buffer_pool_size = 1G

that need fllow ur hardware .
if someone meet same problem,email me:[email protected]...
..