For ObjectMeta pruning, we round trip through marshalling and
unmarshalling. If the ObjectMeta contained any strings with "" (or other
fields with empty values) _and_ the respective fields are omitempty,
those fields will be lost in the round trip process.
This makes ObjectMeta after the no-op write different from the one
before the write.
Resource version is incremented every time data is written to etcd.
Writes to etcd short-circuit if the bytes being written are identical
to the bytes already present. So this ends up incrementing the
resourceVersion even on no-op writes.
The zero values are set in BeforeCreate and BeforeUpdate. This commit
updates BeforeUpdate such that zero values are only set when the
object does not have a zero value for the respective field.
Kubernetes-commit: d691748aa64376e4b90ae2c01aca14d1be2b442c
This ensures that request cancellation will be propagated properly to
the client used to create the stream. Without this fix, the apiserver
and the kubelet may leak resources (e.g., goroutine, inotify watches).
One such example is that if user run `kubectl logs -f <container that
don't produce new logs)` and then enter ctrl-c, both kubelet and
apiserver will hold on to the connection and resources indefinitely.
Kubernetes-commit: 31d1607a514b62ef46452e402f5438d827314b98
Make apiserver pass connectRequest.Options directly to the admission layer. All
the information in rest.ConnectRequest is present in admission attributes.
Kubernetes-commit: 355691d310803ea3a0cd8ff284a39ead38857602
The registry abstraction is unnecessary and adds direct coupling to the
core types. By using a wrapper, we carry through the default
implementations of the non-mutating operations. The DeleteCollection
method is explicitly patched out since it cannot be correctly
implemented on the storage currently.
As a result, TableConvertor is now exposed.
A few other minor refactorings
* Corrected the case of some variables
* Used functions instead of methods for several helper methods
* Removed the legacy Deleter - service was the only remaining consumer
Kubernetes-commit: 110b064d630ca39220696225dd597e7d33b95f4f
Tables should be a mapping from lists, so if the incoming object has
these add them to the table. Allows paging over server side tables.
Add tests on the generic creater and on the resttest compatibility.
Kubernetes-commit: d2a62fd42234a96cbab2dbcf402c168c59b41784
The DestroyFunc functions returned by generic.NewRawStorage is never
called when we do a StartTestServer() in the test suite. For a quick
hack for now, added TrackStorageCleanup/RegisterStorageCleanup and
CleanupStorage. Note that unless TrackStorageCleanup is called (which
is called only from the test suite) the other two methods are
no-ops essentially. So no change in behavior at runtime. This vastly
brings down the number of goroutines that are left behind when this
test is executed and should reduce if not eliminate the flakiness
of TestCRD
Kubernetes-commit: 00bcbd1311af711f70c771d790137b93ce48309a
Currently setting watch cache size for a given resource does not disable
the watch cache. This commit adds a new `default-watch-cache-size` flag
to map to the existing field, and refactors how watch cache sizes are
calculated to bring all of the code into one place. It also adds debug
logging to startup to allow us to verify watch cache enablement in
production.
Kubernetes-commit: fc2d201e155296f311ae0a9278b00dcae2d68708
Add a feature gate in the apiserver to control whether paging can be
used. Add controls to the storage factory that allow it to be disabled
per resource. Use a JSON encoded continuation token that can be
versioned. Create a 410 error if the continuation token is expired.
Adds GetContinue() to ListMeta.
Kubernetes-commit: 8952a0cb722b77459cf2701632a30f5b264f5aba
Decouple finalizer processing from garbage collection configuration.
Finalizers should be effective even when garbage collection is disabled
for a given store.
Fixes https://github.com/kubernetes/kubernetes/issues/50528.
Kubernetes-commit: ed5b5bb94e7c75f22a7fc302e47dade6c0d1662d
Adding myself as a reviewer for the following areas:
- API
- auth
- registry
- storage (etcd)
Signed-off-by: Monis Khan <mkhan@redhat.com>
Kubernetes-commit: dd06794bc20ef1e0889af576c7a4f7a2f607e49d
The strategy used for the secret store defined custom export logic, and
had accompanying unit tests. However the secret storage did not actually
wire this up by setting an ExportStrategy and thus the code was never
used in the real world.
This change fixes the missing assignment and adds testing at a higher
level to ensure any uses of the generic registry.Store that we expect to
have an ExportStrategy do, and no others.
Several other strategies in the RBAC package also appeared to have
unwired Export logic, however their implementations were all empty
leading me to believe that these are not considered exportable. The
empty methods have now been removed.
Kubernetes-commit: 855a1c17131f92fca6de33279a02eca3893ca374
These are googlers who don't work on the project anymore but are still
getting reviews assigned to them:
- bprashanth
- rjnagal
- vmarmol
Kubernetes-commit: c201553f2776ac401549d561485f9a5cb4841ae8
Users who delete a collection expect all resources to be deleted, and
users can also delete an uninitialized resource. To preserve this
expectation, DeleteCollection selects all resources regardless of
initialization.
The namespace controller should list uninitialized resources in order to
gate cleanup of a namespace.
Kubernetes-commit: 9ad1f80fdcd77edcdd53abec3641c04c80fd9b1e
Print a better error from the response. Performs validation to ensure it
does not regress in alpha state.
Kubernetes-commit: ce972ca47591cc24a3a24362478dc61ec8e91278
Add support for creating resources that are not immediately visible to
naive clients, but must first be initialized by one or more privileged
cluster agents. These controllers can mark the object as initialized,
allowing others to see them.
Permission to override initialization defaults or modify an initializing
object is limited per resource to a virtual subresource "RESOURCE/initialize"
via RBAC.
Initialization is currently alpha.
Kubernetes-commit: 331eea67d8000e5c4b37e2234a90903c15881c2f
All generic registries expose metadata output, and refactor endpoints to
allow negotiation to handle those responses. Add support for
PartialObjectMetadata being returned for objects as well.
Kubernetes-commit: f203e42cb98ed4bac7ad8ebbed717d3bd42f55b6
All registry.Store objects already set a non-nil DeleteStrategy.
This change ensures that all future objects do so as well.
Signed-off-by: Monis Khan <mkhan@redhat.com>
Kubernetes-commit: 08fcd79e1f4b9d3efe0a20ea4ce4fdf5ffea0531
Add support for following redirects to the SpdyRoundTripper. This is
necessary for clients using it directly (e.g. the apiserver talking
directly to the kubelet) because the CRI streaming server issues a
redirect for streaming requests.
Also extract common logic for following redirects.
Kubernetes-commit: 715d5d9c91c669cf33c0bf9a9c9d352c6c4228a6
All Stores in Kubernetes follow the same logic for determining the name
of an object. This change makes it so that CompleteWithOptions defaults
the ObjectNameFunc if it is not specified. Thus a user does not need to
remember to use ObjectMeta.Name. Using the wrong field as the name can
lead to an object which has a name that bypasses normal object name
validation.
Signed-off-by: Monis Khan <mkhan@redhat.com>
Kubernetes-commit: ed35deb69d6fe480adc9178c30b4b9c9e03ca1a9
Tim Allclair