When the agent is turned on/off, the information about its availability (becoming available/unavailable) is displayed after 30-40 seconds. Is it possible to make the information about the node's availability change displayed instantly?
-
-
The knowledge of availability can only be determined when item data is expected.
In the agent template an item agent.ping is present and by default it's set on 1m
You can lower this number to now the value quicker, however don't forget to change the trigger as well (change the agent no data macro)
That said if agent availability is that important I would expect you configure a 2-way check to prevent false positives
-
-
-
Not knowing which version you run....
You know that problem creation vs alert notification are 2 different things. 1 is evaluated at the 0s of a minute in v6 (not 100% sure stil the same in v7) . So Does the problem appear in problems section of zabbix front-end.
Details about this is provided in the zabbix traning ZCS. However it was mentioned I can really find it in de slides.
It's just not realistic to do 1s interval check on agent AND also trigger on it.
So depending on what you think it's useful for I would advise you to reconsider your approach.Comment
-
From documentation: nodata(/host/key,sec,<mode>)
Check for no data received.
Supported value types: Integer, Float, Character, Text, Log.
Returns: 1 - if no data received during the defined period of time; 0 - otherwise.
Parameters:- See common parameters;
- sec - the period should not be less than 30 seconds because the history syncer process calculates this function only every 30 seconds; nodata(/host/key,0) is disallowed.
- mode - if set to strict (double-quoted), this function will be insensitive to proxy availability (see comments for details).
Any reason cutting it to this low of a value? In all monitoring implementations I've seen so far, the most delay were always caused by human reaction rather then by systems raising alerts. It depends on actual escalation chain and processes in place, but most often it goes something like 1 minute to raise alert and 5-15 minutes for p1 evaluation, escalation to p2 engineer and his initial investigation and mitigation start.
Do you really have such a strict SLA that 1 min alert raise time causes it to be persistently breached?Comment
-
I wrote a simple interface in QML that displays the status of monitored devices in Zabbix. The information about the access to each agent in the interface is taken from the Zabbix database from the hosts table. I am testing my application by turning the agent on and off. Everything works, but the problem is that after turning the agent on or off, my interface and the Zabbix web interface only update the data after a minute, sometimes longer.
I have tried various ways to reduce the delay: I edited the trigger templates (ICMP ping, Agent ping) for the nodes, setting the intervals to 1 second. I also set all timeouts in the zabbix_server.conf file to 1 second - it didn't help.
I also tried to get information about the availability of agents not through the database, but through the Zabbix API, and it didn't help; the delays are the same.
Zabbix version: 4.0.3Comment
Comment