Fix DeltaFIFO bug that caused the sync delta created by a relist
for object ID X to be dropped if the DeltaFIFO already stored a
Delete delta for X. This caused SharedIndexInformer to miss create
notifications. Also, add unit test to expose the bug.
Kubernetes-commit: 9e4be54895e8c0655408b4766cd78de17b6d5b7c
It was previously possible to instantiate `Reflector` with
`*unstructured.Unstructured` as the expected type but this did not
support checking that event objects were of the correct API
type (e.g. if event object was `v1.Pod` in `Unstructured` form but
`v1.Service` was expected). This commit adds support for providing a
GVK via an `Unstructured` expected type to compare with the GVK of
event objects. The GVK will also be used in reflector log output.
Kubernetes-commit: 237dbfd8ad322dfcad4bd4d5345368480c22d82f
If a cache was already synced, cache.WaitForCacheSync would
always take 100ms to complete because the PollUntil method will
sleep first before checking the condition. Switching to
PollImmediateUntil will ensure already synced caches will return
immediately. For code that has, for example, 20 informers, the time
to check the cache was in sync would take at least 2 seconds, but with
this change it can be as fast as you can actually load the data.
Signed-off-by: Darren Shepherd <darren@rancher.com>
Kubernetes-commit: 8a58f332dd62b68ae22c67585002defe6aeb4b04
Added definition of "collection states" and noted that informer cache
queries are answered against these.
Also added an explicit note that the concept of object identity used
in the contract here does _not_ include ObjectMeta.UID.
Also updated the description of the notifications received by a
client, to account for the fact that a client can be added after the
informer starts running.
Added references to `MetaNamespaceKeyFunc` and
`SplitMetaNamespaceKey`.
Noted that a client must process each notification promptly and added
a referral to workqueue.
Made a couple other minor clarifications and corrections.
Kubernetes-commit: 848301e11a8558f6a90e2fada870a91baace812a
If user specifies a nil callback, then error out
rather than crashing at runtime due to a nil reference.
Kubernetes-commit: 9dbbc652efee21bd421db1b1f52235ed20ffb7eb
2. Remove staging/src/k8s.io/client-go/tools/cache from .golint_failures;
3. Fix some typo from comments.
Kubernetes-commit: 0e0e1f7daba0a6ae6dd59df0a1bb643c323ad8cb
* Kubectl user exec should accept zero-length environment values #652
* Changing TestValidateAuthInfoExecInvalidEnv to allow for empty strings as Exec values
Kubernetes-commit: f30af9dd6da46f0f01e38b477d455907da9f1b6c
* Added custom error message when wrong file is provided with KUBECONFIG
* Modified test case
* Updated the code to warn the missing files
* Renamed the variable
Kubernetes-commit: a5eedcde611658c220c56d2819bf0420aded4ed6
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
* Fixed and clarified comments and parameter names in index.go
Fixed the comment on IndexFunc to say that it returns multiple indexed values.
Clarified the comments and parameter names in the Indexer interface to
consistently use a dichotomy between "storage keys" and "indexed
values".
* Updated comments in index.go based on review by Liggitt
Kubernetes-commit: 4160909495626262ae514bab30fe81a3ff869d39
This warning comes from Reflector watchHandler, from the apiserver error that
indicates a watch was restarted. This happens when etcd drops the connection
and resources are relisted. This informs the user that the watchers are
operating properly, so should be logged as Info rather than Warning.
Kubernetes-commit: a953d1542c06456c6fe97121abd99acd352545df
The cache mutation detector must use a diff function that is
designed to show differences that cause reflect.DeepEqual to return
false.
Signed-off-by: Monis Khan <mkhan@redhat.com>
Kubernetes-commit: dd4f53cd45499687e768de46a15038d61935ab96
There was some weird queuing going on. The queue was implement by
spawning goroutines that would block on the "ticketer" until it was
their turn to write to the output channel. If N events where in the
watch when the consumer of the watch stopped reading events, N
goroutines would leak. In unit tests of the certificate manager, this
was causing ~10k goroutines to leak.
Fix it with a buffering event processor that uses only one routine and
cancels correctly.
Kubernetes-commit: cafc640bfa0f7362b178b1b896085962d018afe3
