Hardware requirements

[mirddin]

Hi all!

Could anybody help with the minimum and recommended HW configuration for such installation:
Number of hosts 505
Number of items 21914
Number of triggers 3919

Now I have 4-core CPU and 4Gb RAM and it's dramatically slow

top - 08:45:53 up 4 days, 18:48, 1 user, load average: 60.02, 61.43, 65.22
Tasks: 643 total, 41 running, 600 sleeping, 0 stopped, 2 zombie
Cpu0 : 35.3%us, 16.8%sy, 0.0%ni, 43.9%id, 1.3%wa, 0.3%hi, 2.4%si, 0.0%st
Cpu1 : 34.0%us, 17.6%sy, 0.0%ni, 46.7%id, 1.4%wa, 0.0%hi, 0.2%si, 0.0%st
Cpu2 : 31.7%us, 17.5%sy, 0.0%ni, 49.4%id, 1.2%wa, 0.0%hi, 0.2%si, 0.0%st
Cpu3 : 29.5%us, 16.4%sy, 0.0%ni, 52.9%id, 1.0%wa, 0.0%hi, 0.2%si, 0.0%st
Mem: 3924444k total, 3746200k used, 178244k free, 122404k buffers
Swap: 2096120k total, 957908k used, 1138212k free, 251620k cached

All item intervals are increased and LLD-rules are also scheduled to run at 86400 sec interval.

thanks beforehand!

[dumi]

Your system should be ok for that number of hosts, please give some details about your database configuration, and zabbix config.
How you monitor your hosts, by agent, snmp or icmp?

P.S.: What is slow? Web interface?

[jan.garaj]

Provide details requested from Dumi.

CPU iowait time 5%
CPU softirq time 2%
CPU load 60
Zabbix busy pollers 100%
=> I don't like these values (maybe problems with HDD performance)
Please provide also more details about your zabbix server CPU utilization/load graph, network utilization graphs, zabbix server performance graph, outputs from the cmds:

iostat -xd 10 10
mpstat -P ALL 10 10
ps -fax
mysqltuner.pl (source http://mysqltuner.pl/)

[tchjts1]

I'm guessing he is running with the default number of pollers and cache settings in his zabbix_server.conf and the system can't even breathe because there are not enough of those allocated to handle the work.

[jan.garaj]

I'm guessing he is running with the default number of pollers and cache settings in his zabbix_server.conf and the system can't even breathe because there are not enough of those allocated to handle the work.

It can be 100% true if the symptom is only 100% Zabbix busy pollers, But it's more complicated now (lot of running proceses <- high CPU load, sw interruption time and high IOps activity <- high iowait CPU time)

Let's wait for more details.