Trigger dependencies notifications problem

[quantim]

Hi,
I configured triggers dependencies and this work well, but in next step I add actions with delay as is describe here. If "master" host goes unreachable, in webinterface I see only this trigger, but actions is fired for both triggers. Is it bug or any misconfiguration?

I use Zabbix 1.8.2 from debian squeeze repository

[quantim]

ok, I make some test and problem is simple. If zabbix first check "master" device, everything is ok and dependency works, but if first chceck "slave" device and then "master", in webinterace looks all ok, but send out 2 notifications.

same issue in 1.8.3 version.

[qix]

Yes, that is how it works. As long as the master isn't checked Zabbix only knows the slave has a problem.

Try increasing the item polling interval on the master or decrease on the slave.

[quantim]

yes, but action is delayed for 120 seconds, in this time zabbix check master and disable slave trigger. In correct state, action for slave didn't fired, but i receive both messages. Is this any way to fired only master action, if slave is checked first?

[qix]

I haven't tried delayed event notifications yet. In fact, the whole escalation option is still new to me. Maybe somebody else can answer your question?

[Surge]

Yes, that is how it works. As long as the master isn't checked Zabbix only knows the slave has a problem.

Try increasing the item polling interval on the master or decrease on the slave.

I'm having the same or similar problem but not sure how to fix it.
Scenario:
Zabbix -- Router 1 -- Router 2 -- Host

Router 1
Trigger: ifOperStatus=2 (2 = serial is down on Cisco routers)
Interval: 60 secs

Router 2
Trigger: icmping.max(240)=0
Interval: 60 secs

Host
Trigger: icmping.max(240)=0
Interval: 60 secs

Router 2's trigger is dependent on Router 1 trigger.
Host's trigger is dependent on Router 1 and Router 2 triggers.

When the line between router 1 and 2 goes down three notifications are sent (Router 1, Router 2 and Host).
However Router 2 and Host have to be unreachable for 4 minutes and Router 1's trigger is evaluated 4 times (every 60 seconds) so I don't understand why the dependencies don't work.
The parent trigger (Router 1) is checked at least three times (possibly 4) before the children are checked.

[qix]

Did you use 'Trigger value = "PROBLEM"' in your actions?

[regi]

ok, I make some test and problem is simple. If zabbix first check "master" device, everything is ok and dependency works, but if first chceck "slave" device and then "master", in webinterace looks all ok, but send out 2 notifications.

same issue in 1.8.3 version.

Same here

Delays are set and ... nothing.

Host trigger status stay PROBLEM
u can't see it in webinterace (it's hidden -> fine )
but u can see it in database and in mailbox - so ...

... if documentation says somthing else

... If Zabbix had been given more time then it would have only sent out the notification about the “router”

this is just a bug (for me).

trigger status should be changed to UNKNOWN or OK if dependency trigger is PROBLEM

[regi]

ok solved

force zabbix to set trigger status to OK if depedns on trigger with status = PROBLEEM

--- src/libs/zbxdbhigh/db.c.old 2011-08-04 10:23:14.000000000 +0200
+++ src/libs/zbxdbhigh/db.c     2011-08-08 14:11:16.000000000 +0200
@@ -494,15 +494,19 @@
                zabbix_log(LOG_LEVEL_DEBUG,"In %s() triggerid:" ZBX_FS_UI64 " old:%d new:%d now:%d reason:'%s'",
                                __function_name, triggerid, trigger_value, new_value, now, reason);
 
+       if (trigger_dependent(triggerid) == SUCCEED)
+               new_value = TRIGGER_VALUE_FALSE;
+
        switch (trigger_type)
        {
                case TRIGGER_TYPE_MULTIPLE_TRUE:
                        update_status = (trigger_value != new_value || new_value == TRIGGER_VALUE_TRUE);
-                       update_status = update_status && trigger_dependent(triggerid) == FAIL;
+                       //update_status = update_status && trigger_dependent(triggerid) == FAIL;
                        break;
                case TRIGGER_TYPE_NORMAL:
                default:
-                       update_status = (trigger_value != new_value && trigger_dependent(triggerid) == FAIL);
+                       //update_status = (trigger_value != new_value && trigger_dependent(triggerid) == FAIL);
+                       update_status = (trigger_value != new_value);
                        break;
        }

[fakenick]

Looks fine but the db code is completely different in 1.8.10 yet it still has the same problem. Could anyone please give me a hand and rewrite the patch (that was originally meant for 1.8.6) to work with 1.8.10?

[fakenick]

Anyone got a clue whether this behavior has changed in 2.0 ?

[sasskinn12]

Tryed 2.0rc1 still same shit!! I dont know, is that so difficult to add this function in zabbix? Why it is working in front end but still sending me notifications... Waited vers 2.0 one year and hoped this problem will be solved..still nothing...

[eugene.istomin]

We have the same problem.

Trigger dependencies are working in webinterface correctly, but notifications are send despite trigger deps.

[eugene.istomin]

bug described in https://support.zabbix.com/browse/ZBX-4344

[sirtech]

Something for the manual perhaps?

Yes, that is how it works. As long as the master isn't checked Zabbix only knows the slave has a problem.

Try increasing the item polling interval on the master or decrease on the slave.

I think something along these lines should go in the manual in the section on trigger dependencies.

I'm still testing Zabbix, and encountered a similar problem, and suspected this was the cause. It confused me a little when the manual made no mention of this behavior though so would be nice to see a little "caution / warning" box in the manual about this.