- when we forward the request to the aggregated server, set the audit
ID in the new request header. This allows audit logs from aggregated
apiservers to be correlated with the kube-apiserver.
- use the audit ID in the current tracer
- use the audit ID in httplog
- when a request panics, log an error with the audit ID.
Kubernetes-commit: b607ca1bf3e1cf6152c446ea61ac7fdd9014e1f1
Manage the audit ID early in the request handling logic so that it can
be used by different layers to improve correlation.
- If the caller does not specify a value for Audit-ID in the request
header, we generate a new audit ID
- If a user specified Audit-ID is too large, we truncate it
- We echo the Audit-ID value to the caller via the response
Header 'Audit-ID'
Kubernetes-commit: 31653bacb9b979ee2f878ebece7e25f79d3f9aa6
- as soon as a request is received by the apiserver, determine the
timeout of the request and set a new request context with the deadline.
- the timeout filter that times out non-long-running requests should
use the request context as opposed to a fixed 60s wait today.
- admission and storage layer uses the same request context with the
deadline specified.
we use the default timeout enforced by the apiserver:
- if the user has specified a timeout of 0s, this implies no timeout on the user's part.
- if the user has specified a timeout that exceeds the maximum deadline allowed by the apiserver.
Kubernetes-commit: e416c9e574c49fd0190c8cdac58322aa33a935cf
- as soon as a request is received by the apiserver, determine the
timeout of the request and set a new request context with the deadline.
- the timeout filter that times out non-long-running requests should
use the request context as opposed to a fixed 60s wait today.
- admission and storage layer uses the same request context with the
deadline specified.
Kubernetes-commit: 83f869ee1350da1b65d508725749fb70d0f535f2
Aborted requests are the ones that were disrupted with http.ErrAbortHandler.
For example, the timeout handler will panic with http.ErrAbortHandler when a response to the client has been already sent
and the timeout elapsed.
Additionally, a new metric requestAbortsTotal was defined to count aborted requests. The new metric allows for aggregation for each group, version, verb, resource, subresource and scope.
Kubernetes-commit: 057986e32c1bb7284b0edbc161f0380f1548492f
The MaxInFlight and PriorityAndFairness apiserver filters maintain
watermarks with histogram metrics that are observed when requests
are handled. When a request is received, the watermark observer
needs to fill out observations for the entire time period since the
last request was received. If it has been a long time since a
request has been received, then it can take an inordinate amount of
time to fill out the observations, to the extent that the request
may time out. To combat this, these changes will have the filters
fill out the observations on a 10-second interval, so that the
observations never fall too far behind.
This follows a similar approach taken in
9e89b92a92c02cdd2c70c0f52a30936e9c3309c7.
https://github.com/kubernetes/kubernetes/issues/95300
The Priority-and-Fairness and Max-in-Flight filters start goroutines to
handle some maintenance tasks on the watermarks for those filters. Once
started, these goroutines run forever. Instead, the goroutines should
have a lifetime tied to the lifetime of the apiserver.
These changes move the functionality for starting the goroutines to
a PostStartHook. The goroutines have been changed to accept a stop channel
and only run until the stop channel is closed.
Kubernetes-commit: 6c9b86646871f13a4431361310ba6a0785372053
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
Currently we record request metrics during the normal request flow and
we also manually invoke `Record` in the timeout handler to record
timeouts. This means that we effectively double count whenever we
timeout. This PR renames the `Record` function to `RecordRequestError`
to more accurately reflect the intended side-effect of the function
call.
Change-Id: Ie37fd0c1e501bd525640a434433d364a5fd6dde2
Kubernetes-commit: 4c6e7247878477a1f2efc26df7f141258010374f
The previous HTTP compression implementation functioned as a filter, which
required it to deal with a number of special cases that complicated the
implementation.
Instead, when we write an API object to a response, handle only that one
case. This will allow a more limited implementation that does not impact
other code flows.
Also, to prevent excessive CPU use on small objects, compression is
disabled on responses smaller than 128Kb in size.
Kubernetes-commit: 4ed2b9875d0498b5c577095075bda341e96fcec2
Abu Kashem