Thank you Brians for your responses!
I was hopping this problem could be solved with some dark magic promQL
queries. I'd probably go with the vm_tag solution, as it is a general
purpose metric that allows queries with group_right vector matching,
solving more use cases around "tagging".
FYI:
To give you some context about why I chose
vm_info{name="AAAAA" tags=",big,azure,"} 1
is inspiration in prometheus code itself, in the ec2 service discovery code
(and others)
https://github.com/prometheus/prometheus/blob/1a6c2283a3e37f5ef6e0df78b72e67d3b2995a56/discovery/aws/ec2.go#L307-L310
So it seems this "tagging" pattern is already in the community, but
unfortunately does not solves all use cases by itself.
Best!
Javier
El jueves, 26 de mayo de 2022 a las 18:08:01 UTC+2, Brian Candler escribió:
> ... although that doesn't help you with your count-by-arbitrary-tag
> problem, sorry. For that as Brian B says, you need separate series:
>
> vm_tag{name="AAAAA",tag="big"} 1
> vm_tag{name="AAAAA",tag="azure"} 1
>
> On Thursday, 26 May 2022 at 16:57:41 UTC+1 Brian Candler wrote:
>
>> Or separate the tags into individual labels:
>>
>> vm_info{name="AAAAA",tag_big="1",tag_azure="1"} 1
>>
>> On Thursday, 26 May 2022 at 13:51:43 UTC+1 Brian Brazil wrote:
>>
>>> On Thu, 26 May 2022 at 13:42, Javier Mendiara Cañardo <[email protected]>
>>> wrote:
>>>
>>>> Hi,
>>>>
>>>> I have an info metric with all possible tag values
>>>> tag_info{tag="small"} 1
>>>> tag_info{tag="medium"} 1
>>>> tag_info{tag="big"} 1
>>>> tag_info{tag="azure"} 1
>>>> tag_info{tag="aws"} 1
>>>>
>>>> I have an info metric with a multi-value label called "tags", as a
>>>> csv with commas envelope
>>>> vm_info{name="AAAAA" tags=",big,azure,"} 1
>>>> vm_info{name="BBBBB" tags=",big,aws,"} 1
>>>> vm_info{name="CCCCC" tags=",small,aws,"} 1
>>>>
>>>> I want to get how many machines have that specific tag:
>>>> {tag="big"} 2
>>>> {tag="small"} 1
>>>> {tag="azure"} 1
>>>> {tag="aws"} 2
>>>>
>>>> The tags values are not known when coding the query
>>>>
>>>> Any hints? It's a kind-of label_split missing function.
>>>>
>>>
>>> I don't believe this is possible with PromQL in a general form. You can
>>> figure out the value for a single known tag with a regex matcher, but not
>>> in the general case. Such a function would also be a footgun, as it could
>>> greatly increase the cardinality of the output relative to the input. As it
>>> stands you can only increase input to output cardinality via
>>> absent/absent_over_time, and that's only from 0 to 1.
>>>
>>> Thanks!
>>>>
>>>> PD: I have under the control the exporter to make a metrics refactor.
>>>>
>>>
>>> This is what you'd need to do. You want a separate gauge metric with a
>>> sample per tag present.
>>>
>>> Brian
>>>
>>>
>>>>
>>>>
>>>> --
>>>> You received this message because you are subscribed to the Google
>>>> Groups "Prometheus Users" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send
>>>> an email to [email protected].
>>>> To view this discussion on the web visit
>>>> https://groups.google.com/d/msgid/prometheus-users/885bc38a-9950-4100-a247-68d1e52f17e7n%40googlegroups.com
>>>>
>>>> <https://groups.google.com/d/msgid/prometheus-users/885bc38a-9950-4100-a247-68d1e52f17e7n%40googlegroups.com?utm_medium=email&utm_source=footer>
>>>> .
>>>>
>>>
>>>
>>> --
>>> Brian Brazil
>>> www.robustperception.io
>>>
>>
--
You received this message because you are subscribed to the Google Groups
"Prometheus Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/prometheus-users/e5d34bb9-9915-4d3e-8f0d-f0a269d460e9n%40googlegroups.com.
