Skip to content

Commit 9360aa1

Browse files
committed
sample-and-watermark replaced with timing histograms
1 parent 1939378 commit 9360aa1

File tree

1 file changed

+30
-66
lines changed

1 file changed

+30
-66
lines changed

content/en/docs/concepts/cluster-administration/flow-control.md

Lines changed: 30 additions & 66 deletions
Original file line numberDiff line numberDiff line change
@@ -508,26 +508,15 @@ poorly-behaved workloads that may be harming system health.
508508
last window's high water mark of number of requests actively being
509509
served.
510510

511-
* `apiserver_flowcontrol_read_vs_write_request_count_samples` is a
512-
histogram vector of observations of the then-current number of
513-
requests, broken down by the labels `phase` (which takes on the
514-
values `waiting` and `executing`) and `request_kind` (which takes on
515-
the values `mutating` and `readOnly`). The observations are made
516-
periodically at a high rate. Each observed value is a ratio,
517-
between 0 and 1, of a number of requests divided by the
518-
corresponding limit on the number of requests (queue length limit
519-
for waiting and concurrency limit for executing).
520-
521-
* `apiserver_flowcontrol_read_vs_write_request_count_watermarks` is a
522-
histogram vector of high or low water marks of the number of
523-
requests (divided by the corresponding limit to get a ratio in the
524-
range 0 to 1) broken down by the labels `phase` (which takes on the
525-
values `waiting` and `executing`) and `request_kind` (which takes on
526-
the values `mutating` and `readOnly`); the label `mark` takes on
527-
values `high` and `low`. The water marks are accumulated over
528-
windows bounded by the times when an observation was added to
529-
`apiserver_flowcontrol_read_vs_write_request_count_samples`. These
530-
water marks show the range of values that occurred between samples.
511+
* `apiserver_flowcontrol_read_vs_write_current_requests` is a
512+
histogram vector of observations, made at the end of every
513+
nanosecond, of the number of requests, broken down by the labels
514+
`phase` (which takes on the values `waiting` and `executing`) and
515+
`request_kind` (which takes on the values `mutating` and
516+
`readOnly`). Each observed value is a ratio, between 0 and 1, of a
517+
number of requests divided by the corresponding limit on the number
518+
of requests (queue volume limit for waiting and concurrency limit
519+
for executing).
531520

532521
* `apiserver_flowcontrol_current_inqueue_requests` is a gauge vector
533522
holding the instantaneous number of queued (not executing) requests,
@@ -542,52 +531,27 @@ poorly-behaved workloads that may be harming system health.
542531
holding the instantaneous number of occupied seats, broken down by
543532
the labels `priority_level` and `flow_schema`.
544533

545-
* `apiserver_flowcontrol_priority_level_request_count_samples` is a
546-
histogram vector of observations of the then-current number of
547-
requests broken down by the labels `phase` (which takes on the
548-
values `waiting` and `executing`) and `priority_level`. Each
549-
histogram gets observations taken periodically, up through the last
550-
activity of the relevant sort. The observations are made at a high
551-
rate. Each observed value is a ratio, between 0 and 1, of a number
552-
of requests divided by the corresponding limit on the number of
553-
requests (queue length limit for waiting and concurrency limit for
554-
executing).
555-
556-
* `apiserver_flowcontrol_priority_level_request_count_watermarks` is a
557-
histogram vector of high or low water marks of the number of
558-
requests (divided by the corresponding limit to get a ratio in the
559-
range 0 to 1) broken down by the labels `phase` (which takes on the
560-
values `waiting` and `executing`) and `priority_level`; the label
561-
`mark` takes on values `high` and `low`. The water marks are
562-
accumulated over windows bounded by the times when an observation
563-
was added to
564-
`apiserver_flowcontrol_priority_level_request_count_samples`. These
565-
water marks show the range of values that occurred between samples.
566-
567-
* `apiserver_flowcontrol_priority_level_seat_count_samples` is a
568-
histogram vector of observations of the utilization of a priority
569-
level's concurrency limit, broken down by `priority_level`. This
570-
utilization is the fraction (number of seats occupied) /
571-
(concurrency limit). This metric considers all stages of execution
572-
(both normal and the extra delay at the end of a write to cover for
573-
the corresponding notification work) of all requests except WATCHes;
574-
for those it considers only the initial stage that delivers
575-
notifications of pre-existing objects. Each histogram in the vector
576-
is also labeled with `phase: executing` (there is no seat limit for
577-
the waiting phase). Each histogram gets observations taken
578-
periodically, up through the last activity of the relevant sort.
579-
The observations
580-
are made at a high rate.
581-
582-
* `apiserver_flowcontrol_priority_level_seat_count_watermarks` is a
583-
histogram vector of high or low water marks of the utilization of a
584-
priority level's concurrency limit, broken down by `priority_level`
585-
and `mark` (which takes on values `high` and `low`). Each histogram
586-
in the vector is also labeled with `phase: executing` (there is no
587-
seat limit for the waiting phase). The water marks are accumulated
588-
over windows bounded by the times when an observation was added to
589-
`apiserver_flowcontrol_priority_level_seat_count_samples`. These
590-
water marks show the range of values that occurred between samples.
534+
* `apiserver_flowcontrol_priority_level_request_utilization` is a
535+
histogram vector of observations, made at the end of each
536+
nanosecond, of the number of requests broken down by the labels
537+
`phase` (which takes on the values `waiting` and `executing`) and
538+
`priority_level`. Each observed value is a ratio, between 0 and 1,
539+
of a number of requests divided by the corresponding limit on the
540+
number of requests (queue volume limit for waiting and concurrency
541+
limit for executing).
542+
543+
* `apiserver_flowcontrol_priority_level_seat_utilization` is a
544+
histogram vector of observations, made at the end of each
545+
nanosecond, of the utilization of a priority level's concurrency
546+
limit, broken down by `priority_level`. This utilization is the
547+
fraction (number of seats occupied) / (concurrency limit). This
548+
metric considers all stages of execution (both normal and the extra
549+
delay at the end of a write to cover for the corresponding
550+
notification work) of all requests except WATCHes; for those it
551+
considers only the initial stage that delivers notifications of
552+
pre-existing objects. Each histogram in the vector is also labeled
553+
with `phase: executing` (there is no seat limit for the waiting
554+
phase).
591555

592556
* `apiserver_flowcontrol_request_queue_length_after_enqueue` is a
593557
histogram vector of queue lengths for the queues, broken down by

0 commit comments

Comments
 (0)