I like it, should be implemented at some point.

+1

+1!

ZBXNEXT-1521 asks for this in aggregate items

+1

please do not clutter issues with such comments - see https://zabbix.org/wiki/Docs/bug_reporting_guidelines#Reporting_an_issue

+1

+1

ZBXNEXT-1521 could help to achieve some mentioned task.

ZBXNEXT-2707 asks for aggregate item based on lld.

+1

+1

please do not clutter issues with such comments - see https://zabbix.org/wiki/Docs/bug_reporting_guidelines#Reporting_an_issue

If you want to show that you also suffer from a particular problem and would like to see it fixed, vote on it.

see the vote feature in the upper right corner of this page.

Hey,
guys, is there a chance that this is going to be impelemented? Or maybe some kind of variation on aggregate check that would calculate items on one host (instead of group of hosts) by their, for example, application? Having access to this functionality natively would be awesome and for sure it would be 'clean and non-hackish'

I'd upgraded server to version 3.2.4 and all lld calculated items, which where summing dynamic number of other lld items within one host just died (case similar to ZBX-11700 or ZBX-11411), due to newly implemented limitations

if lack of a specific feature is important to you, please consider financed development.

until this feature appears, this example script to create graphs could be easily adapted to create calculated items
http://www.zabbixbook.com/2017/02/07/zabbix-api-example-creating-a-graph-with-lld-generated-items/

Thanks for a quick response!
I'm afraid that amount of funds needed to develop one feature is just out of my range .
Until I figure this example script out, I guess downgrading server would be the fastest solution for me.

had similar issue, to workaround the issue, I create a script which take a regex and create a calculated item or graph based on found item.
Check out https://github.com/cgasp/zabbix-aggregate-items

would it make sense to implement this as - for example -
sum(last("vfs.fs.size[*,total]"))
where in this case the discovered filesystem is a wildcard, or maybe even a regexp to make it even more versatile ....
?

Strongly +1

before this implemented, I have to update the workaround  "last(item[0]) + last(item[1]) + last(item[2])...“ frequently, because the item[x] will change frequently.

The following SQL-query shows a result for discovered items created by prototypes from this project:

https://github.com/Scatty81/zabbix-windows.git

It queries all Hosts, where Java Runtime

Name.{26A24AE4%

has multiple installations - which often leads to problems.

select ap.name
, ho.name
, substring(it.name, 1, 29)
, string_agg(distinct hi.value::text, ' ')
from hosts ho
join items it on it.hostid = ho.hostid
join items_applications ia on ia.itemid = it.itemid
join applications ap on ia.applicationid = ap.applicationid
join history_text hi on hi.itemid = it.itemid
where ap.name like 'Software'
and it.name like 'Name.{26A24AE4%'
group by ho.name, ap.name, substring(it.name, 1, 29)
having count(*) > 1
order by 1, 2, 3

This would of course be better achieved with this Issue.

Why this still not implemented?

This is obvious that it is a needed and a must be feature.

We have the same problem here: We use a discovery to discover network interfaces and create items for in/out traffic per interface. A very common demand is to sum the in/out traffic of every bonded interface to get the actual traffic.

Are there any workarounds since this issue is open for 7 years?

Hi sirs!

I do think about what if merge this issue with ZBXNEXT-1521 and ZBXNEXT-2707 by add to calculated items ability use functions from aggregated items and regexs or wildcards. As results can be key with custom name type of calculated items, for example:

grpsum["host group","host","item key or standard function eg last()"]

This may be like:

grpsum[,the.host.dom","my_discovery_item_key[eth$\{MACROS}.last()"]

${MACROS}=regex([0-9])

+1

+1

+1

+1

this is especially useful, for monitoring discovered parallel jobs, where number of jobs on hosts can differ, but if they fail, one trigger over all of them would be sufficient

or one action over multiple triggers, that when i have 20 parallel jobs, I don't need to receive 20 emails, just one would be enough even if multiple jobs fail.

+1

I can't believe this is not part of Zabbix already!

+1

I can't believe this is not part of Zabbix already! (me too!)

+1

+1

We'll try to have it in 5.4 but we need a nice and clean implementation design for this one.

Thank you.

is the much needed ,please help to develop this

+1

+1, need this too

It's now possible in Zabbix 5.4.

Suppose you have a discovery rule that returns LLD macro {#INDEX} and creates items like discovered_key[{#INDEX}]. Then you can define a calculated item with the formula sum(last_foreach(//discovered_key[*])) which will calculate the sum of last values for all discovered items.

Are there any use cases that are not covered by this approach?

Hello, I tested this new feature. I'm trying to add following item:

sum(last_foreach(/*/net.if.in.discards[ifInDiscards.*]?[group="DC/Switches"]))

Returns error "Cannot evaluate function: no items matching query at ...". Host group specified correctly. Item name copied from original template (Interfaces SNMP). I not yet understand what I've done wrong.

The following one worked for me:

sum(last_foreach(/*/titanDiameterNgPeerServerResultCount[*,*,*],5m))

I have a problem. I'm doing an LLD via the ssh.run method. Meaning my item keys end up like:

• ssh.run[myvalue_item1]
• ssh.run[myvalue_item2]
• ssh.run[myvalue_item3]

I would like to do

sum(last_foreach(//ssh.run[myvalue_*]))

but end up with the error

Cannot evaluate function: item "/test_host/ssh.run[myvalue_*]" does not exist at "last_foreach(//ssh.run[myvalue_*]))

I can do ssh.run[*] but that would include everything I do via SSH on that host which is not what I want.

So how (if possible) do I fix this?

RenHoek, asterisks '*' in the item filters work like placeholders for hosts, keys or argument positions. Thus these are not capable of partial argument matching.

Alex Kalimulin : So how do I do it then? Is there some sort of regex function I can use? Should I put in a feature request?

RenHoek, there is no regex processing for this use case. Please feel free to add a new feature request.

This is resolved since 5.4, closing.