(Thinks: maybe it's *not* necessary to apply distinct labels? This feels
wrong somehow, but I can't pinpoint exactly why it would be bad)
On Tuesday 9 January 2024 at 14:43:51 UTC Brian Candler wrote:
> Unfortunately, the timeout can't be longer than the scrape interval,
> firstly because this would require overlapping scrapes, and secondly the
> results could be returned out-of-order: e.g.
>
> xx:yy:00 scrape 1: takes 25 seconds, gives result at xx:yy:25
> xx:yy:15 scrape 2: takes 5 seconds, gives result at xx:yy:20
>
> > If I run two blackbox_probes in parallel with scrape_interval: 30s and
> scrape_timeout: 30s this will work but both probes will start more or less
> at the same time.
>
> Actually I think you'll find they'd be evenly spread out over the scrape
> interval - try it.
>
> For example, make a single scrape job with a 60 second scrape interval,
> and list the same target 4 times - but make sure you apply some distinct
> label to each instance, so that they generate 4 separate timeseries. You
> can then look at the raw timestamps in the database to check the actual
> scrape times: easiest way is by using the PromQL web interface and
> supplying a range vector query, like probe_success{instance="foo"}[5m].
> This has to be in table view, not graph view. Don't mix any other targets
> into that scrape job, because they'll be spread together.
>
> Alternatively, KISS: use a 15 second scrape interval, and simply accept
> that "scrape failed" = "took longer than 15 seconds". Does it really matter
> whether it was 20 seconds or 25 seconds? Can you get that information from
> somewhere else if needed, e.g. web server logs?
>
> On Tuesday 9 January 2024 at 10:04:42 UTC Alexander Wilke wrote:
>
>> Hello,
>> I want to use blackbox_exporter and http prober to login to an API.
>>
>> My goal is to do the login every 15s which could be
>>
>> xx:yy:00
>> xx:yy:15
>> xx:yy:30
>> xx:yy:45
>>
>> I could solve this with scrape_interval: 15s.
>> But in addition I want to allow a scrape timeout of 30s which is longer
>> than the scrape_timeout.
>>
>> If I run two blackbox_probes in parallel with scrape_interval: 30s and
>> scrape_timeout: 30s this will work but both probes will start more or less
>> at the same time.
>>
>> xx:yy:00
>> xx:yy:30
>>
>> The idea behind tha is:
>> In general the API response for login is very fast. For whatever reason
>> sometimes it takes 30s or more. I do not want the probe to just fail after
>> 15s but want to see and understand how long a login request takes.
>>
>> If I abort a long lasting request or do a parellel login this may work
>> very fast. So it is probably not a problem with the API in general but with
>> specific user session or other unknown circumstances. So I want many scrape
>> intervals but the imeout sometimes needs to be higher OR I need several
>> blackbox_probes which do not start at the same time but are spread equally.
>>
>> Any ideas?
>> Is this possible with prometheus 2.48.1 and blackbox_exporter 0.24.0 ?
>>
>>
>>
>>
--
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/7cef9bcc-ef76-408e-9c91-689014ec9e42n%40googlegroups.com.
