You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Workaround partially missing prometheus metrics
Workaround partially missing metrics.
e.g. if we have a list of 4 partitions with metrics missing for
2 out of 4, like [10,11,0,0], resulting "expectedConsumption will be sum
of [10,11,11,11]
Expose more metrics
Expose "lag as seen by konsumerator" metric.
Expose number of samples received from prometheus for
production/consumption/offsets to know if some timeseries are missing.
fix "expected" counter. fix glimiter value in state
"Expected" value prints number of partitions instead of the number of
instances. Fix this.
state.glimitResources used to contain amount of resources that we
"asked" from glimiter, instead of the actual resources allocated after
the limit was applied. fix this.
Compare only "requests" during estimateDeploy
Ignore `resource.limits` block during comparison of current vs
estimated, it's always just a rounding of the `request.request` and
causes weird logging when deployment was created with custom values.
Print memory in the logline, since it's also used in the comparison of
resources.