3 or more Zabbix Agents Unavailable in a Host Group

[BigDave]

Zabbix version 7.0.23

I'm tryng to set-up a trigger if 3 or more Zabbix agents go unavailable in a Host Group. I understand that one option is to have a calculated item and then trigger off of that. I'm struggling with the formula for both

Has anyone set-up something simliar so I can see how it's suppose to work?

[BigDave]

Update -
Using agent.ping was causing some false alarms so I replaced "agent.ping" and "is_online" in the items 'hosts_in_group' and 'hosts_online' with "zabbix[host,agent,available]"
Also deleted the 'Is host online' item from the template

So I have a host in each group with the below items -


And this trigger -

[guntis_liepins]

This is a bit tricky question and solution I offer may be not optimal, but since nobody suggest anything better here will my 5 cents.

1. we need some item common in host group which every host in this group has . I will use agent.ping in this example
Make calculated item

name	hosts_in_group
type	Calculated
key	hosts_in_group
Type of information	integer
formula	sum(exists_foreach(/*/agent.ping?[group="test_servers"]))

This work this way - exists_foreach Returns '1' for each enabled item , item selector - selects whole host group , then sum this value.
As each host must return 1 if agent.ping is enabled, for each host then sum of all items will be equalt to a count of hosts in host group.

2. we need item which returns 1 or 0 in each monitoring cycle, which would indicate if host are up or down.
agent.ping does not fit out purpose as it returns 1 - available, but does not return data if not available.
so we for every host create item is_online

name	Is host online
type	Calculated
key	is_online
type of information	Integer
formula	1-nodata(//agent.ping,60s)

Nodata will return 1 if host have not received successful agent.ping for one minute. So his item will return 0 if host is offline , but will return 1 if host is online.

3. Now we create function which will show us ho many hosts are online

name	hosts_online
type	Calculated
key	hosts_online
type of information	integer
formula	sum(last_foreach(/*/is_online?[group="test_servers"]))

This function will get last value for each is_online item in group test_servers then sum it together.
As online hosts count as 1 , offline hosts as 0 this item will return integer host many hosts are online

Now we can create trigger which will trigger when more than three hosts are offline

Many hosts down

last(/tt/hosts_in_group)-last(/tt/hosts_online)>=3

(I created virtual host tt which hold two items hosts_in_group and hosts_online and created problem there )

[BigDave]

Thank you for the response.

Would I be able to use zabbix[host,agent,available] instead of agent.ping?

From what I've found agent.ping returns no data in case of unavailability, whereas zabbix[host,agent,available] returns a 0

So I wouldn't need step 2 and can replace 'is_online?' in step 3 with 'zabbix[host,agent,available]?'

Correct?

Edit: Also to make this more difficult I have 3 host groups so this needs to work if 3 or more hosts are down in each group

2nd Edit: SOLVED: Created a host per group with the items - 'hosts_in_group' &' hosts_online' and the trigger. Each other host has the 'Is host online' item created via their template.

[guntis_liepins]

Sure. It was just example. Glad you made it!

[BigDave]

Update -
Using agent.ping was causing some false alarms so I replaced "agent.ping" and "is_online" in the items 'hosts_in_group' and 'hosts_online' with "zabbix[host,agent,available]"
Also deleted the 'Is host online' item from the template

So I have a host in each group with the below items -


And this trigger -