Debian Wheezy, Postgresql 9.1

i believe this wouldn't really work for how zabbix is designed - we would need timestamps of all those values anyway to know at what time they were collected...

a very similar feature request asks for skipping sending of identical values for active agents : ZBXNEXT-113

i think here is a good solution for this, to reduce data: http://blog.endpoint.com/2014/04/filling-gaps-in-cumulative-sum-in.html

Hi, i have implement something like by sripts, and i think it can be easy implemented in zabbix itself.
In my realization:
For ping, we have 2 items:
1. ping_trap sender
2. ping_trap reciver

ping_trap sender - run every 30-60s and ping remote host, save this value somewhere (i use files in tmpfs)
if new value is different to latest, then send it, to ping_trap reciver
As addition, i think about add periodic heartbeats, i.e. send value, without check, to ping reciver if it not send any data in 20m & etc
—
It's can be easy implemented as type of check and can reduce storage needs/DB performance
I.E. on WebUI add new type of checks: ~zabbix agent (active trapper)
With 2 time keys:
Frequency (30s)
Max time between forced send (20m) (barrier)
On zabbix agent side:
1. Load items from ServerActive
2. For items with that type:
2.1 Check Last value not NULL
2.2 save in memory with timestamp
2.3 send it to server
—
And in loop renew value, if it changed -> send it to server, if it get time barrier -> send to server
—
On WebUI side for Nice graps, logic for trapped graphs can be used.
—
For nice trapped graphs (i.e. with ping example, host will be fall in graph up to time barrier value)
Just send 2 values if it changed: old with timestamp = (timestamp - Frequency/2) + new value.
—
This logic can be also used for simple checks on server side

P.S. i just think what it can fix this issue for some time without big code changes in zabbix
And i don't want maintain own version of code for this feature
P.S.S.
By using this method i can increase frequency of all checks and decrease size of history tables(indexes) by 2-10x

I think this could also easily achieved by ZBXNEXT-1836 resp. ZBXNEXT-714 (e.g. by using a module for a proper time series database).

ZBXNEXT-3544 is similar

ZBXNEXT-4016 hints about implementing this as a new preprocessing option.

It could also be implemented as a preprocessing option that would compare new value with the last one. If they are equal then the new value will be ignored.

Hey all! Please have a look at the attached acceptance criteria document that describes proposed enhancements for preprocessing that will cover this functionality as well as ZBXNEXT-4016 and ZBXNEXT-3544. It has all chances to be implemented in Zabbix 4.2. Let me know what you think. Is there anything to improve? I appreciate any feedback.

I recall that sometimes, when putting in some trigger conditions, there is no support in negative matching, i.e. !~, and I have to set it up in the global regex section and use it in the trigger. The thing is, if I am not mistaken, those global stuff are not easily exportable and kept consistent across multiple zabbix instances. Is it possible to add this feature?

Daniel.

I read the document, and it's really a good way to reduce the database size.

Support of throttling has been implemented under https://support.zabbix.com/browse/ZBXNEXT-4724, I am closing this ZBXNEXT.