Web monitoring not working properly

[sikander]

Hi.
I am testing zabbix, and has been for a couple of months now. I like it a lot, but compared to Nagios, the availability report gives 1-3 % better numbers.
Yesterday i set up a new test to a website i can turn on and off as i like.
When i turned it off, something strange happens.
It reports the site as down, as it should, but once in a while it reports the site as up again without me doing anything to it.
Does anyone know what the problem is?
More info and images coming in the comments.

[sikander]

Trigger:

{Zabbix-test:web.test.fail[Zabbix-test; EOS].count(180,0,"ne")}>1 | {Zabbix-test:web.test.fail[Zabbix-test; EOS Adm].count(180,0,"ne")}>1

[sikander]

Latest data for "Failed step" for the time period the site was down:
image 1

[sikander]

Events for the same time period:
image 2

[Heilig]

You use incorrect key in trigger expression, check the documentation.
Try this:
{Zabbix-test:web.test.fail[Zabbix-test].count(180,0,"ne")}>1

[sikander]

If you mean the [; EOS]-part, then its unfortunately not it.
I put semicolon and the name for its group in the name of the web-test so that it would be easyer to know exactly what test that failed (we have several services with the same name in different groups).
Or is the semicolon not supported in names in zabbix? Are there any symbols that is safe to use?

[Heilig]

Sorry, was inattentive.
Comes to mind only update zabbix to latest stable version and/or switch debug mode 4 and tracking...