The BackoffManager interface sleeps without considering the caller's context,
i.e. cancellation is not supported. This alone is reason enough to deprecate it
and to replace it with an interface that supports a context parameter.
The other reason is that contextual logging needs that parameter.
Kubernetes-commit: b15a1943d51adfb8c5e0185d58d25e038c3d6ade
The default handler now uses contextual logging. Instead of
warnings.go:106] warning 1
it now logs the caller of client-go and uses structured, contextual
logging
main.go:100] "Warning" message="warning 1"
Users of client-go have the choice whether the handler that they provide uses
the traditional API (no API break!) or contextual logging.
Kubernetes-commit: 48fb886325fce4b16e4067caadb7bcd3044d460f
Logging in rest.Request.Body cannot be made context-aware without an API
change. Such a change is complicated if done in a backwards-compatible
fashion (must change lots of callers in Kubernetes) and prohibitive if not (all
callers of Body would have to pass a context).
Instead, logging of the request body gets moved into the functions which send
the request. This is a change of behavior, but it is limited to log levels >= 8
and thus should have no impact in production.
A request which gets sent multiple times will also log the body multiple
times. This might even be a good thing because it serves as reminder what is
being sent when it is being sent.
While at it, stack backtracing gets enhanced so that the caller of the REST API
is logged and tests for the new behavior get added.
Kubernetes-commit: 57f9b7c7a2412865e7817dbf7638881b00ac9721
The Error method of the error returned from Request.Watch was "unknown"
even the server returned clear message in the Status struct. It was
because Request.Watch used the Result's err member directly, which is an
unstructured error from the response which the Result object may use if
the caller did not return a structured error.
The patch fixes it by calling the Result's Error method instead, which
returns the structured error when it's present.
It also removes the wrong expectation about events.
Kubernetes-commit: 596c5696c64023808af164284263647d795b0ac2
This touches cases where FromInt() is used on numeric constants, or
values which are already int32s, or int variables which are defined
close by and can be changed to int32s with little impact.
Signed-off-by: Stephen Kitt <skitt@redhat.com>
Kubernetes-commit: 94410ee8078971b8894e5c400ce3fc79f02bc510
Requests can accumulate errors with no obvious indication, e.g. if
their primary purpose is to construct a URL: URL() itself doesn't
return an error if r.err is non-nil.
Instead of changing URL() to return an error, which has quite a large
impact, add an Error() function and indicate on URL() that it should
be checked.
Signed-off-by: Stephen Kitt <skitt@redhat.com>
Kubernetes-commit: f69c1c47463ff70ad61adf6f38c4d5b7373e9d0a
The functionality provided by the finalURLTemplate is still used by
certain external projects to track the request latency for requests
performed to kube-apiserver.
Using a template of the URL, instead of the URL itself, prevents the
explosion of label cardinality in exposed metrics since it aggregates
the URLs in a way that common URLs requests are reported as being the
same.
This reverts commit bebf5a608f68523fc430a44f6db26b16022dc862.
Signed-off-by: André Martins <aanm90@gmail.com>
Kubernetes-commit: f8f190cdd2fa76296f8b1b019ac77128b5d40b79
This commit refactors the retry logic to include resetting the
request body. The reset logic will be called iff it is not the
first attempt. This refactor is nescessary mainly because now
as per the retry logic, we always ensure that the request body
is reset *after* the response body is *fully* read and closed
in order to reuse the same TCP connection.
Previously, the reset of the request body and the call to read
and close the response body were not in the right order, which
leads to race conditions.
This commit also adds a test that verifies the order in which
the function calls are made to ensure that we seek only after
the response body is closed.
Co-authored-by: Madhav Jivrajani <madhav.jiv@gmail.com>
Kubernetes-commit: 68c8c458ee8f6629eef806c48c1a776dedad3ec4
The restclient metrics were updated to track only the host field of the
url, the finalURLTemplate is not longer needed, its only goal was to
replace name and namespace in the path to avoid cardinality.
Kubernetes-commit: bebf5a608f68523fc430a44f6db26b16022dc862
the test TestHTTP1DoNotReuseRequestAfterTimeout has to wait for
request to time out to assert that subsequent requests does not
reuse the TCP connection.
It seems that current value of 100ms causes issues on some CI
environments and bumping the timeout seems to solve this flakiness,
We can bump the timeout value because is really low compared to real
scenarios and the bump still keeps it in the millisecond order.
Kubernetes-commit: 85797eba7075d83b116b5c91ff3b17c2d5118d01
It iterates over the wrapped transports until it finds one
that implements the CloseIdleConnections method and executes it.
add test for closeidle http1 connections
add test for http1.1 reconnect with inflight request
add test to reuse connection request
add test for request connect after timeout
add test for client-go request concurrency
Kubernetes-commit: b9d865a8185b62d83e9ff81b0e3499a26ac6960d
This reverts commit 5a59a43957c6743995dac67fdda42bf8e0a9ca77, reversing
changes made to 81b9789eaa7bc067f417b5e74d5695dd6dd88a46.
Kubernetes-commit: 892d4fabb845e2461e3655aa414beb6ac322fc99
In some environments, where url base is "/", it can cause all paths to
be presented in metrics with "{prefix}" as `groupIndex` is with the wrong
index. To fix the behavior in such environments, it was added a
conditional branch to check if the URL base is "/" and, thus, print the
metrics with the correct path, for example "api/v1/nodes/{name}" instead
of "{prefix}".
Fixes: 99248b8fe1fe ("Rewrite finalURLTemplate used only for metrics because of dynamic client change")
Signed-off-by: André Martins <aanm90@gmail.com>
Kubernetes-commit: c039b02fa7281fc061455e23b6530ed8b4d19645
This commit performs two refactors and fixes a bug.
Refactor 1 changes the signature of Request to take a RESTClient, which
removes the extra copy of everything on RESTClient from Request. A pair
of optional constructors are added for testing. The major functional
change is that Request no longer has the shim HTTPClient interface and
so some test cases change slightly because we are now going through
http.Client code paths instead of direct to our test stubs.
Refactor 2 changes the signature of RESTClient to take a
ClientContentConfig instead of ContentConfig - the primary difference
being that ClientContentConfig uses ClientNegotiator instead of
NegotiatedSerializer and the old Serializers type. We also collapse
some redundancies (like the rate limiter can be created outside the
constructor).
The bug fix is to negotiate the streaming content type on a Watch()
like we do for requests. We stop caching the decoder and simply
resolve it on the request. We also clean up the dynamic client
and remove the extra WatchSpecificVersions() method by providing
a properly wrapped dynamic client.
Kubernetes-commit: 3b780c64b89606f4e6b21f48fb9c305d5998b9e5
Migrated code that checks for common programmer errors to a separated
function and added test coverage for it. Wrong comment stating that a
typed error is returned was also removed.
Kubernetes-commit: ad5fafd6ade2838098890a4e7727c8e347686867
Normal files should have permissions 644 by default,
and does not require the last bit to be
executable
Signed-off-by: Odin Ugedal <odin@ugedal.com>
Kubernetes-commit: 35cb87f9cf71776e99a970dfff751cd29ba7ebfb
