apiserver_flowcontrol_request_wait_duration_seconds
apiserver_flowcontrol_request_concurrency_in_use
apiserver_flowcontrol_request_concurrency_limit
apiserver_flowcontrol_rejected_requests_total
apiserver_flowcontrol_dispatched_requests_total
apiserver_flowcontrol_current_inqueue_requests
apiserver_flowcontrol_current_executing_requests
Signed-off-by: Andrew Sy Kim <andrewsy@google.com>
Kubernetes-commit: 0bb419b1498a664d1dda3b487e9f15fd220ea363
Also make some design changes exposed in testing and review.
Do not remove the ambiguous old metric
`apiserver_flowcontrol_request_concurrency_limit` because reviewers
though it is too early. This creates a problem, that metric can not
keep both of its old meanings. I chose the configured concurrency
limit.
Testing has revealed a design flaw, which concerns the initialization
of the seat demand state tracking. The current design in the KEP is
as follows.
> Adjustment is also done on configuration change … For a newly
> introduced priority level, we set HighSeatDemand, AvgSeatDemand, and
> SmoothSeatDemand to NominalCL-LendableSD/2 and StDevSeatDemand to
> zero.
But this does not work out well at server startup. As part of its
construction, the APF controller does a configuration change with zero
objects read, to initialize its request-handling state. As always,
the two mandatory priority levels are implicitly added whenever they
are not read. So this initial reconfig has one non-exempt priority
level, the mandatory one called catch-all --- and it gets its
SmoothSeatDemand initialized to the whole server concurrency limit.
From there it decays slowly, as per the regular design. So for a
fairly long time, it appears to have a high demand and competes
strongly with the other priority levels. Its Target is higher than
all the others, once they start to show up. It properly gets a low
NominalCL once other levels show up, which actually makes it compete
harder for borrowing: it has an exceptionally high Target and a rather
low NominalCL.
I have considered the following fix. The idea is that the designed
initialization is not appropriate before all the default objects are
read. So the fix is to have a mode bit in the controller. In the
initial state, those seat demand tracking variables are set to zero.
Once the config-producing controller detects that all the default
objects are pre-existing, it flips the mode bit. In the later mode,
the seat demand tracking variables are initialized as originally
designed.
However, that still gives preferential treatment to the default
PriorityLevelConfiguration objects, over any that may be added later.
So I have made a universal and simpler fix: always initialize those
seat demand tracking variables to zero. Even if a lot of load shows
up quickly, remember that adjustments are frequent (every 10 sec) and
the very next one will fully respond to that load.
Also: revise logging logic, to log at numerically lower V level when
there is a change.
Also: bug fix in float64close.
Also, separate imports in some file
Co-authored-by: Han Kang <hankang@google.com>
Kubernetes-commit: feb42277884bc7cfbd6f0bb1d875cc63b1b6caac
Fix the one path where boundNextDispatchLocked was not being called
after modifying a queue.
Also check for negative work in a request.
These are motivated by
https://github.com/kubernetes/kubernetes/issues/112169 but I do not
have a way to reproduce it and so can not check that these changes
actually remove that symptom. But these changes are good anyway.
Kubernetes-commit: 6ee93e2cee695203a6ce4935da1b9a807b624260
- Run hack/update-codegen.sh
- Run hack/update-generated-device-plugin.sh
- Run hack/update-generated-protobuf.sh
- Run hack/update-generated-runtime.sh
- Run hack/update-generated-swagger-docs.sh
- Run hack/update-openapi-spec.sh
- Run hack/update-gofmt.sh
Signed-off-by: Davanum Srinivas <davanum@gmail.com>
Kubernetes-commit: a9593d634c6a053848413e600dadbf974627515f
Track the introduction of FinalSeats.
Give up on calculating expected results for tests with added latency,
because I did not find an easy and obvious way to do it.
Kubernetes-commit: 0fc595e03360ba7fc4c3e251d4b41f39172aca72
New anti-windup technique: use the request arrival time as the floor
on the virtual dispatch time. Prevent bound violations where they
might arise rather than fixing up just one queue at dispatch time,
so that the fixed up dispatch times figure into the dispatching choice.
Two tweaks to the shuffle sharding. Take seats of executing requests
into account as well as seats of waiting requests. Do not always
consider the generated hand in the same order.
Rename the queueset methods that do shuffle sharding and finding the
queue to dispatch from, because the old names were confusingly
similar.
Tighten up some request margins.
Name the test cases in TestNoRestraint and TestWindup.
Kubernetes-commit: 4b9cba85874158b25b5c994773a4ec04343820c2
Canonicalize listing of test cases.
Make TestNoRestraint try both cases: competition and none.
Kubernetes-commit: 0ee1a7b4ff9012b050bd447055ad5e1e8c57c30e
Make TestNoRestraint verify that fairness is NOT achieved
when there is real competition.
Make TestWindup run two cases, to show that 0.1 is too narrow
a margin and 0.26 is wide enough.
Kubernetes-commit: c4945fdf0c14ba2032a5c8edf192678d9fe00374
These behavioral unit tests of queueset were failing because the
evaluation criteria were too strict.
Kubernetes-commit: 59d319ec06bb33289a87036418b4a61ed3bb215f
So that the width estimate has some effect but not a grossly excessive
one.
Added the fifo::Peek method to simplify the fifo client code.
Also renamed the queueSet::estimatedServiceTime field to
estimatedServiceSeconds to make the units clear.
Kubernetes-commit: a0c161f2f6908ee424ea888ff40f75ff071bd20a
Added missing dispatching after delayed release of seats.
Updated logging for all six situations of execution completion and
seat release.
Added behavioral tests for non-zero extra latency and non-unit width.
Also added two tests for baseline functionality.
Also improved some comments and other logging in `queueset.go`.
Kubernetes-commit: d2a27a58f0af20c6185fa1c21890d666e9d3746b
Add comment outlining TestContextCancel.
Stop calling `t.Errorf` from wrong goroutine.
Package up queueNoteFn expectation checking.
Add counting of goroutine in req1 exec fn.
Remove unnecessary assignment to `_`.
Make TestContextCancel wait on fake clock, to insulate timing check
from scheduler noise.
Factor goroutine counting out of queueset.go, into queueset_test.go,
where it matters.
Refactor promise: Use a simple channel-based implementation for normal
code, a mutex-based one for testing code.
Took all the panics out of queueset.go
Shrink the timeouts in promise tests to 1 second.
Kubernetes-commit: 1db36ae3b30e30d70972998a22987a7db470479b
Rename from `clock` to `eventclock`.
Simplify by removing the prohibition on an EventFunc suspending and
resuming activity.
Remove "EventClock" from names to avoid stuttering.
Start to consolidate test code under fairqueuing/testing/.
Kubernetes-commit: 80ca6a4ae6ff571c32962a7155efd55edefff9e6
So we can move off of the apimachinery clock package.
Switch queueset to new clocks.
Removed event clocks based on apimachinery clocks,
because this PR introduces ones based on k8s.io/utils/clock .
Removed interface that is implemented by only one interesting type.
Simplify RealEventClock::EventAfterTime.
Kubernetes-commit: dcb298c9552de44e27ed52f5e2b58a0dd7cd8d54
- add plumbing that allows us to estimated "width" of a request
- the default implementation returns 1 as the "width" of all
incoming requests, this is in keeping with the current behavior.
Kubernetes-commit: 9b72eb1929a64b9d5a5234090a631ba312fb4d41
Introduce min, average, and standard deviation for the number of
executing mutating and readOnly requests.
Introduce min, max, average, and standard deviation for the number
waiting and number waiting per priority level.
Later:
Revised to use a series of windows
Use three individuals instead of array of powers
Later:
Add coarse queue count metrics, removed windowed avg and stddev
Add metrics for number of queued mutating and readOnly requests,
to complement metrics for number executing.
Later:
Removed windowed average and standard deviation because consumers can
derive such from integrals of consumer's chosen window.
Also replaced "requestKind" Prometheus label with "request_kind".
Later:
Revised to focus on sampling
Make the clock intrinsic to a TimedObserver
... so that the clock can be read while holding the observer's lock;
otherwise, forward progress is not guaranteed (and violations were
observed in testing).
Bug fixes and histogram buckets revision
SetX1 to 1 when queue length limit is zero, beause dividing by zero is nasty.
Remove obsolete argument in gen_test.go.
Add a bucket boundary at 0 for sample-and-water-mark histograms, to
distinguish zeroes from non-zeros.
This includes adding Integrator test.
Simplified test code.
More pervasively used "ctlr" instead of "ctl" as abbreviation for
"controller".
Kubernetes-commit: 57ecea22296797a93b0157169db0ff2e477f58d0
Abu Kashem