Zabbix ( server & proxy ) how do I change interval for how often a host is checked ?

[[email protected]]

Zabbix ( server & proxy ) how do I change interval for how often a host is checked ?

I have a fairly large number of hosts ( about 1-thousand with ICMP Ping and SNMP checks ).

My Zabbix Server was getting close to 500 vps and very very sluggish on the web front-end , so I spun up a Proxy server. It seems that no matter how many hosts I move to the Proxy server ( some or all hosts ) , web access to the front-end Zabbix Server remains sluggish ( almost to the point of being broken ).
I noticed that if I move all hosts from the Server to the Proxy , the vps on both the Server and Proxy are still over 400 vps. If I disable all hosts on Server and Proxy , my vps drops down to around 19 vps.

On some of the other Network Monitor systems I work on ( trying to get out of other NMS systems and move everything to Zabbix ) , there are usually some check-host interval setting which will slow down how many hosts are checked in a 1-to-5 minute host-check cycle.
I have googled and googled and I am unable to find out how often Zabbix hosts are checked and where/how to change the host-polling-check cycle.
So , I would like to ask " How do I change how often Zabbix will perform host check cycles ( for both ICMP and SNMP ? "

What is the default host-check polling cycle time ( 5 seconds ?? ) ?
Is it possible to change the host-check polling cycle time to 15 or 30 or 60+ seconds ? ( I think I found some info on doing this but I got quickly lost in what I was reading ).

Am I correct in guessing that I can lower the vps values if I lengthen the host-check polling cycle times ?

Thank you for any replies.

North Idaho Tom Jones

[Atsushi]

The monitoring interval for monitoring is set by setting the update interval in the item settings. If you extend the update interval time, I think the VPS value will be smaller.

[[email protected]]

The monitoring interval for monitoring is set by setting the update interval in the item settings. If you extend the update interval time, I think the VPS value will be smaller.

Atsushi ,
Thank you for your prompt answer

It looks do-able to me .
I am thinking that I can edit a few Templates ( ICMP ping , my common SNMP macros ) and everything using those templates should get the new interval time -- I hope

Again - thank you

North Idaho Tom Jones

[[email protected]]

Any thoughts about vps ?

Some of the other non-Zabbix Network-Monitor-Systems I have worked on often have an issue with polling hosts. I suspect Zabbix might have the same issue as what I have encountered with Poll Timing.
- Most NMS systems have multiple schedules for when to check hosts. These schedules often hit every 60 seconds ( 1-minute ) and every hour.
- The potential issue is that every 60 seconds ( on the minute ) and every hour , the NMS system(s) get very busy checking hosts. Sometimes during these common/shared polling intervals where everything is being checked at the same time , the NMS can become quite sluggish.

As an example ; one of my other older NMS systems is WhatsUp Gold. I have 3,000+ hosts in 32+ maps. The default schedule to check hosts on all maps always would have a hit on every minute and every hour - which drove the NMS crazy with suddenly having to check all hosts at the same time.
What I did reduce the chances of all hosts being checked at the same moment is that I would modify the polling schedules to always be an odd number of seconds where my polling schedules would look something like this:
59 or 61 or 119 or 121 or 299 or 301 seconds. Other host would be polled at 57 or 63 or 117 or 123 or 297 or 303 seconds. These changes resulted in a much faster responding NMS system & much faster web interface because instead of checking all hosts at common time intervals , large groups of hosts were checked at regular schedules where the chances of everything being checked at the exact moment of time was greatly reduced -- which resulted in a much faster responding NMS system.

I am going to try this approach on my Zabbix update intervals , so that ICMP checks and SNMP checks and other checks all have a reduced hit period where everything is happening at the same time.

Does anybody have any thoughts on this ?

North Idaho Tom Jones

[Atsushi]

Zabbix automatically adjusts the timing to check each item. If you only set the update interval, Zabbix's process of retrieving item values will not be concentrated only at specific timings. In other words, only the time intervals for monitoring items are maintained. On the latest data screen, check the time when each item's value was retrieved.
If you really want to get the value at the same time as much as possible, you can set the item value to be checked at a specific time by using a custom interval. However, if the processing is too concentrated at that timing, the load will increase and the processing will be delayed, so be careful.

https://www.zabbix.com/documentation/6.0/en/manual/config/items/item/custom_intervals

[cyber]

As already said... Zabbix is not running your checks exactly at 00:01, 00:02 etc.. it tries to get values as close to specified interval, but it is interval, not a exact time. Queues can form, if you have too few pollers, maybe slower network and there just is no time to serve all requests. You can see that on process activity graphs... if they get close to 100% busy, you have an issue, you should increase the number of pollers...

What hardware you are running there? If you moved all hosts to proxy already and your server is still sluggish, then you may need to add some power there.. And also move frontend to separate host. And DB.. specially DB...