Thanks Brian for your reply.
In my use case, if i want sent the error log into the alarm generated, i
should add the error message as label of my metric. The metric created by
mtail :
test_dbms_error[$container,$namespace,$pod_name,$domain,$productname,$setname,$message]
As the error message is present in the metric, i can't created my sample
with value 0 at the start. Indeed, the content of error message is
dynamically registered from the log and i can't created the metric sample
before.
This is why i would like use a alertmanager or prometheus parameter for
auto-resolv my rule. But it's not possible?
Loïc
Le mercredi 22 juin 2022 à 12:11:40 UTC+2, Brian Candler a écrit :
> > When my alarm is firing, i would like auto-resolved it
>
> Alerts are generated by a PromQL expression ("expr:"). For as long as
> this returns a non-empty instance vector, the alert is firing. When the
> result is empty, the alert stops.
>
> For example: I want to get an alert whenever the metric
> "megaraid_pd_media_errors" increases by more than 200. But if it has been
> stable for 72 hours, I want the alert to go away. This is what I do:
>
> - alert: megaraid_pd_media_errors_rate
> expr: increase(megaraid_pd_media_errors[72h]) > 200
> for: 5m
> labels:
> severity: warning
> annotations:
> summary: 'Megaraid Physical Disk media error count increased by
> {{$value | humanize}} over 72h'
>
> Every time the expr is evaluated, it's looking over the most recent 72
> hours. "increase" is like "rate", but its output is scaled up to the time
> period in question - i.e. instead of rate per second, it gives rate per 72
> hours in this case.
>
> > i tried to use the promql function rate but in this case my first
> occurence is missing.
>
> "rate" (and "increase") calculate the rate between two data points. If
> the timeseries has only one data point, it cannot give a result. It cannot
> assume that the previous data point was zero, because in general that may
> not be the case: prometheus could have been started when the counter was
> already above zero.
>
> You should make your timeseries spring into existence with value 0 at the
> start.
>
> On Wednesday, 22 June 2022 at 09:27:52 UTC+1 Loïc wrote:
>
>> Hi,
>>
>> I use an exporter mtail to alerting when a pattern match into the
>> kubernetes logs. When my alarm is firing, i would like auto-resolved it. I
>> search how to use tje endsat parameter in my rule but i don't found.
>>
>> Also, i tried to use the promql function rate but in this case my first
>> occurence is missing.
>>
>> Have you an idea ?
>>
>> Thanks
>> Loïc
>>
>
--
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/ee94ec73-8714-46b5-b6cd-1ec1cabcf93en%40googlegroups.com.
