I had the a very similar requirement. It was a tricky query to build from
scratch, but simple when you've worked it out, so I'm happy to share :-)
- name: DiskRate12h
interval: 1h
rules:
# Warn if rate of growth over last 12 hours means filesystem will fill in
7 days
- alert: DiskFilling7d
expr: |
node_filesystem_avail_bytes / (node_filesystem_avail_bytes -
(predict_linear(node_filesystem_avail_bytes,fstype!~"fuse.*|nfs.*"}[12h],
604800) < 0)) * 604800
for: 24h
labels:
severity: warning
annotations:
summary: 'Filesystem will be full in {{ $value | humanizeDuration }}
at current 12h growth rate'
I'm using "node_filesystem_avail_bytes" rather than
"disk_usage_percentage", but as they both trend down to zero, you should be
able to replace it. Replace the time periods as appropriate.
The logic goes something like this: say V is the variable you're interested
in (node_filesystem_avail_bytes in this case)
* we take the current value of V; call it V1
* predict_linear(V[12h], 604800) is the expected value in 7 days time based
on the trend over the last 12 hours; call it V2
* filter that with < 0, so we get no value unless it's predicted to be
below 0 in 7 days
^ V1
| \
| \
+--0---x--7----> time
\
V2
To find where the cut is on the time axis, you note that V1 is to (V1 +
(-V2)) as x is to 7 days. That is, V1/(V1-V2) is the ratio of the lines
V1...x and V1...V2. And therefore that's also the fraction of 604800
seconds to the zero crossing point x.
Your problem is slightly different: you want to know when the free space
percentage will fall below 20, not when it falls below zero. I'll leave
that as an exercise :-) I think just substituting
(disk_space_percentage-20) everywhere in place of the variable is a good
starting point, but you have to be careful what happens if the current
value is already below 20.
HTH,
Brian.
On Thursday, 18 August 2022 at 06:31:08 UTC+1 [email protected] wrote:
> Hello all,
>
> I have the percentage of disk usage on a metric. I can use
> predict_linear(disk_usage_percentage[30d], 30*24*60*60) to give me a
> prediction in 1 month from the past month of metrics. fine
>
> but how could I retrieve the date on which the predict_linear function
> will reach 80% for instance ? if that's possible :-)
>
> Thank you
>
> regards
> ++ Jerome
>
>
>
--
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/836079a5-3e0e-457f-95db-724c024b2f99n%40googlegroups.com.
