In your code, you just add deltas to a counter metric. This tracks the
local known state of how many lines it's seen. Prometheus tracks this over
time, allowing you to use rate() and increase() functions to calculate the
deltas over time.
If you're looking to observe log files, I highly recommend mtail. It's a
well tested and reliable option for producing metrics about and from log
files.
On Mon, May 25, 2020 at 1:48 PM tomeri <tom...@liveperson.com> wrote:
> You right, i care about "how many lines have I added to this file", and
> yes I'm actually want to check the rate of log files (append-only).
> Let's say counter is the best fit here, on the code side how do i get the
> diff between the current total value to the previous sample? do i need to
> hold it in my app memory? is it something that Prometheus can solve?
>
> On Sunday, May 24, 2020 at 8:42:12 PM UTC+3, Julius Volz wrote:
>>
>> On Sun, May 24, 2020 at 5:02 PM tomeri <tom...@liveperson.com> wrote:
>>
>>> Hi Julius,
>>>
>>> Thanks for the fast response.
>>>
>>> Do you think Gauge is the right metric kind for this purpose?
>>>
>>
>> A gauge tracks anything that can naturally go up or down. The length of
>> lines in a file can in principle go up or down, so normally that would be a
>> gauge.
>>
>> If you have files that are basically append-only (like log files), and
>> you want to track the rate at which they grow, then it would be better to
>> think of the metric conceptually not as "how many lines does this file
>> have", but "how many lines have I added to this file" (can only go up, and
>> you count lines as they are being added, not just proxying some third-party
>> line count), which would be more fitting for a counter metric.
>>
>>
>>> How do I handle resets? i see drops with deriv/delta
>>>
>>
>> If it's actually a counter (see above) where drops truly indicate a
>> reset, then you will need to use rate() or increase(). In your case
>> "increase(my_lines_total[1m])". This will handle counter resets, but will
>> still have the same extrapolation caveat as I mentioned earlier.
>>
>>
>>> I tried both drive and delta with 1m time range multiple by 60 and got
>>> wrong results, here is my query:
>>> sum by(datacenter)(delta(files_total_lines_gauge{datacenter="xxx"}[1m]))
>>> * 60 > 0
>>>
>>> I use >0 to workaround the drops
>>>
>>> I basically try to get the same result as I were to use graphite with
>>> nonNegativeDerivative
>>>
>>
>> It sounds like the closest thing in PromQL is indeed "increase(foo[1m])",
>> but it sounds like the Graphite function doesn't do the extrapolation bit.
>>
>>
>>> On Sunday, May 24, 2020 at 5:39:02 PM UTC+3, Julius Volz wrote:
>>>>
>>>> Since it's a gauge (and at least theoretically line counts can
>>>> decrease, not only increase), you'll want either the delta() (
>>>> https://prometheus.io/docs/prometheus/2.18/querying/functions/#delta) or
>>>> the deriv() (
>>>> https://prometheus.io/docs/prometheus/2.18/querying/functions/#deriv
>>>> <https://www.google.com/url?q=https%3A%2F%2Fprometheus.io%2Fdocs%2Fprometheus%2F2.18%2Fquerying%2Ffunctions%2F%23deriv&sa=D&sntz=1&usg=AFQjCNHJKYuM-jz4RtrYTPmKad4hicnl3w>)
>>>> function, multiplied by 60 (to get from per-second to per-minute).
>>>>
>>>> Note that both functions can give you non-integer results even if the
>>>> line numbers only change by integer increments/decrements, as delta()
>>>> extrapolates the observed slope to the edges of the provided time window,
>>>> and deriv() does a linear regression to estimate how fast a gauge is going
>>>> up or down.
>>>>
>>>> Another thing you could do (if you care about integer results) is:
>>>>
>>>> my_lines_total - my_lines_total offset 1m
>>>>
>>>> ...to give you the absolute difference between the last sample value
>>>> seen 1m ago and the currently last-seen sample value. Note that while this
>>>> returns you an integer result, it might be further away from the "true"
>>>> rate due to the lack of extrapolation, because the two samples you will be
>>>> comparing will not be exactly 1m apart.
>>>>
>>>> On Sun, May 24, 2020 at 4:05 PM tomeri <tom...@liveperson.com> wrote:
>>>>
>>>>> Hi,
>>>>>
>>>>> I run an application that export metrics about the total number of
>>>>> lines of each file in my directory, basically the product of `wc -l`
>>>>> So on each interval (every 1m) my app counts the total number of lines
>>>>> and then updates the Gague metric.
>>>>>
>>>>> For example:
>>>>> First iteration: 1000 total lines
>>>>> Second iteration: 1300 total lines
>>>>> Third iteration: 1900 total lines
>>>>> Fourth iteration: 2400 total lines
>>>>> ...
>>>>>
>>>>> Prometheus scraps my app's metric every 15s/25s (depends on the env)
>>>>> What i want to plot is a graph that will show the rate per minute -
>>>>> how many lines produced in each file for that last 1 minute.
>>>>>
>>>>> No matter what I tried, I couldn't make the graph to show to correct
>>>>> results.
>>>>>
>>>>> This message may contain confidential and/or privileged information.
>>>>> If you are not the addressee or authorized to receive this on behalf
>>>>> of the addressee you must not use, copy, disclose or take action based on
>>>>> this message or any information herein.
>>>>> If you have received this message in error, please advise the sender
>>>>> immediately by reply email and delete this message. Thank you.
>>>>>
>>>>> --
>>>>> 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 promethe...@googlegroups.com.
>>>>> To view this discussion on the web visit
>>>>> https://groups.google.com/d/msgid/prometheus-users/4d971809-72e0-4f36-b0cf-0e9914a2d251%40googlegroups.com
>>>>> <https://groups.google.com/d/msgid/prometheus-users/4d971809-72e0-4f36-b0cf-0e9914a2d251%40googlegroups.com?utm_medium=email&utm_source=footer>
>>>>> .
>>>>>
>>>>
>>>>
>>>> --
>>>> Julius Volz
>>>> PromLabs - promlabs.com
>>>>
>>>
>>> This message may contain confidential and/or privileged information.
>>> If you are not the addressee or authorized to receive this on behalf of
>>> the addressee you must not use, copy, disclose or take action based on this
>>> message or any information herein.
>>> If you have received this message in error, please advise the sender
>>> immediately by reply email and delete this message. Thank you.
>>>
>>> --
>>> 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 promethe...@googlegroups.com.
>>> To view this discussion on the web visit
>>> https://groups.google.com/d/msgid/prometheus-users/ca9dc1da-c8c2-4fa7-85d1-ebff2dc393da%40googlegroups.com
>>> <https://groups.google.com/d/msgid/prometheus-users/ca9dc1da-c8c2-4fa7-85d1-ebff2dc393da%40googlegroups.com?utm_medium=email&utm_source=footer>
>>> .
>>>
>>
>>
>> --
>> Julius Volz
>> PromLabs - promlabs.com
>>
>
> This message may contain confidential and/or privileged information.
> If you are not the addressee or authorized to receive this on behalf of
> the addressee you must not use, copy, disclose or take action based on this
> message or any information herein.
> If you have received this message in error, please advise the sender
> immediately by reply email and delete this message. Thank you.
>
> --
> 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 prometheus-users+unsubscr...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/prometheus-users/5fbe9bc7-8b5c-48c7-b4bd-0429dcf011e6%40googlegroups.com
> <https://groups.google.com/d/msgid/prometheus-users/5fbe9bc7-8b5c-48c7-b4bd-0429dcf011e6%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
--
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 prometheus-users+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/prometheus-users/CABbyFmrfivduyD85KaZbL-hJ1tMdDqCxc5jS1CG1SO8GJTtCSw%40mail.gmail.com.
