Closed mthompson-kx closed 3 years ago
Summary and histogram metrics are set to _seconds
_seconds
Change to after:{} to return seconds
after:{}
q)tm1:`timestamp$.z.d q)tm2:tm1+0D00:00:05 q) q)(10e-9)*tm2-tm1 50f q)(10e-10)*tm2-tm1 5f q)
Prometheus histogram buckets are required to be cumulative, with the last bucket matching the value of _count
_count
Additionally a +Inf le is required.
+Inf
le
Without the above changes the PromQL queries fail, returning NaN
NaN
histogram_quantile(0.9, rate(kdb_sync_histogram_seconds{service="kx-tp"}[10s]) )
Without changes
{container="tp-metrics", endpoint="metrics", instance="10.0.1.22:8080", job="kx-tp", namespace="kx", pod="kx-tp-566978d9fc-g4z4r", service="kx-tp"} NaN
With changes
{container="tp-metrics", endpoint="metrics", instance="10.0.0.39:8080", job="kx-tp", namespace="kx", pod="kx-tp-688fbf59cc-9l9v5", service="kx-tp"} 0.225
Summary and histogram metrics are set to
_seconds
Change to
after:{}
to return secondsPrometheus histogram buckets are required to be cumulative, with the last bucket matching the value of
_count
Additionally a
+Inf
le
is required.Without the above changes the PromQL queries fail, returning
NaN
Without changes
With changes