[1.3.8] - how to refresh triggers?

[limo]

We have some virtual hosts (data projectors). That projectors are like normal hosts, but have only one item (trapper item). We use simple trigger like host:errstate.last(OK)=0. Time by time, we send information using zabbix_sender which says if there is some problem with projector.

But when I upgraded zabbix, triggers are unknown. There are still old latest data, when I look into overview/data, I see OK messages sent by zabbix_sender. But triggers are unknown until I send all trapper items again.

I think zabbix should look into history when starting and update triggers to their right state, not to wait for next data to send. Or did I miss something? Maybe there is another way how to update unknown triggers without sending OK by zabbix_sender to all projectors?

Thanx!

[Alexei]

ZABBIX follows the following logic:

If ZABBIX server was down, it has no idea was information has been sent by sender, ZABBIX agent, whatever. That's why it sets status of triggers which fell into the downtime period to UNKNOWN.

[limo]

No, this is not my problem. Values was sent BEFORE zabbix server went down and they are already in DB. I just restarted zabbix_server and triggers are unknown. But they should not be, because {projector:errstate.str(OK)}=0 should be expanded as false. Maybe I do not understand logic of updating triggers. They are updated just after getting data (asynchronous) ? This could be source of problem because there are no data in history for errstate from time zabbix_server was started again. But there are data in history for errstate for time of previous zabbix_server run.

Maybe housekeeper could try to update this unknown triggers by looking into history. Or maybe I still missimg something. But as I said, result of this expression is OFF, value in triggers is UNKNOWN.

thanx!

[Alexei]

The logic of ZABBIX server is very simple:

1. It assumes that it knows nothing about the downtime period.

How can it know that you didn't send new values for the errstate while it was down?!

2. It updates trigger statuses to 'unknown' after restart

[limo]

OK. So we probably use bad scenario because of bad understanding how zabbix works. Only to be sure that I understand it, maybe good to know for another zabbix users:

- We have some device which is not real host (does not have IP, like dataprojectors)
- Our admin of this devices uses zabbix to set this host faulty or OK (using zabbix_sender) when device state changes.
- Item errstate is string and has history set to 351 days
- We have map of this, any user can look if dataprojector is OK (see attachement)
- After zabbix server restart, triggers will be unknown and our admin has to manualy send all informations again by zabbix_sender (even if they are already in history).

Am I right ? If so, I have to find another way how to achieve similar map.

[Alexei]

Nice map!

You may set a refresh period to a very large number for these items in order to fool ZABBIX server. It should not set status of these triggers to UNKNOWN if the last timestamp of the items is not older than the refresh period.

[limo]

OK, latest thing in this thread

Thanx for help, I changed delay to very big number. But small suggestion: I had to do it directly in db because I cannot set update interval for item type zabbix trapper.

Thanx again!

[dash]

I have the same request: Please make the interval configurable.

[Alexei]

Instead of making the interval configurable, I changed logic. Now ZABBIX won't update status to UNKNOWN if a trigger consists of trapper items only.