It's already written in /etc/hosts file. That link seems OK. What do you think about of poller process on proxies?
-
-
Hi!
About Pollers can say, what 1000 it is too much, but as we can see on graphs it is ~80% busy.
When you send me performance graphs from proxy it was already 1000 or you set it because of my recommendations?
Actually, each Poller can processed data from couple of devices. If you have about 1500 devices, theoretically should work normally with 25-50 poller processes. But it depends from how fast devices respond.
If we set 1000 pollers on our test lab environment zabbix just dies because of war for configuration cash.
Tune number of Pollers according to Data gathering processes graph, so them should be about 40% busy.
As well can see how many Pollers are working and Idle running command ps aux | grep poller.
Regards,
AlexanderComment
-
I set 1000 before sending graphs. I don't know why is poller busy over 90% and how I can decrease it if it is necessary. Thank for your help.Comment
-
Regarding to missing items and big queue: I found I've never received any of "Alias of interface" and "Interface description" from some devices. When I do snmpwalk from zabbix-proxy host those devices are answering correctly. Other items from problematic devices are here.
One more question. I have four proxies. Three of them have autodiscovery rules and autoregistration actions. When host is discovered by one of proxies (for example proxy1) and registered by IP address (x.x.x.x) and after that I manually switch them to be monitored by some other proxy (for example proxy2) in next circle of discovering by proxy1 it is added as x.x.x.x_2. May I prevent it and how?Comment
-
Problem is still here.. Every day at about 2 pm history syncer goes to 100% and graphs and data are delayed by 15-30 minutes..
Comment
-
Since I changed partitioning resolution from one day to 6 hours I didn't notice any problem. ThanksComment
Comment