Mass testing of 10k+ triggers- can it be done?

[jonathan.vanderwatt]

Hi all,

I'm hoping that someone could possibly assist me with a rather bizarre query...

Because of the way our organisation works, I am required to have our Zabbix server send all traps to an external monitoring system. For now, it's only for one-time consumption to see how this external system will handle our traps and also to identify possible reconfigurations that may be required in our trigger expressions and names. The mandate is that the external system will run regexps over our traps, extracting highly specific CI data, and then passing the information on to our central command center in a format that their system and personnel would expect. I hope that makes sense. This step is working but...

What I'm looking for is a way to tell Zabbix to do a mass trigger of all 10675 triggers without having to unlink and re-link hosts to templates. Is this possible?

All triggers and items are either templated or in LLD rules. Also, we're currently running Zabbix 2.4.7 on CentOS 6.

[kloczek]

What I'm looking for is a way to tell Zabbix to do a mass trigger of all 10675 triggers without having to unlink and re-link hosts to templates. Is this possible?

Zabbix has been tested already on scale of millions triggers.

[jonathan.vanderwatt]

Zabbix has been tested already on scale of millions triggers.

Hi kloczek,

My question does not pertain to the scalability of Zabbix. I'd like to know if it's possible to manually do some sort of mass run of all my triggers- even if the status is "OK". The amount of triggers is irrelevant.

[kloczek]

Hi kloczek,

My question does not pertain to the scalability of Zabbix. I'd like to know if it's possible to manually do some sort of mass run of all my triggers- even if the status is "OK". The amount of triggers is irrelevant.

Every time when is received from agents or proxies point of any metric and if this metrics is used in some trigger trigger expression is evaluated and value of the trigger is updated.
So if you are asking about testing triggers on massive scale without monitoring traffic on top of standard zabbix answer on your question is: no, it is not possible to do this separately.
Such test done without even simulated/generated monitoring traffic will be worthless.
Such test will have only any meaning with handling whole monitoring data on calculating trends points and storing raw data in DB backend.

It really doesn't matter what is the value of the trigger. Only a changes triggers states matters as such change generates inserts to events table.
However handling stream of data going to events table usually is not a big problem.
Other aspect of working zabbix in context f triggers is what happens just after zabbix server start as one of the first thing which is doing server is reevaluating all triggers. It created kind of punch of read IOs on DB backend and every update query of the trigger as well is generating not only write IOs but read IOs as well.

If you really want to hammer zabbix you need to generate very specific monitoring traffic which will be generating change state of the triggers on every new point of raw monitoring data.