This commit adds a new volume manager in kubelet that synchronizes
volume mount/unmount (and attach/detach, if attach/detach controller
is not enabled).
This eliminates the race conditions between the pod creation loop
and the orphaned volumes loops. It also removes the unmount/detach
from the `syncPod()` path so volume clean up never blocks the
`syncPod` loop.
Automatic merge from submit-queue
volume integration: wait for PVs before creating PVCs
The test should wait until all volumes are processed by volume controller (i.e. in the controller cache) before creating a PVC.
Without that, the "best" matching PV could not be in the cache and controller might bind the PVC to suboptiomal one.
This fixes integration test flake "Bind mismatch! Expected pvc-2 capacity 50000000000 but got pvc-2 capacity 52000000000".
Fixes#27179 (together with #26894)
Automatic merge from submit-queue
Fix integration pv flakes
There are two fixes in this PR:
- run tests in separarate functions and use objects with different names, otherwise events from the beginning of the function are caught later when we watch for events of a different PV/PVC
- don't set PV.Spec.ClaimRef.UID of pre-bound PVs. PVs with UID set are considered as bound and they are deleted/recycled when appropriate PVC does not exists yet.
Fixes#26730 and probably also ~~#26894~~ #26256
The test should wait until all volumes are processed by volume controller (i.e.
in the controller cache) before creating a PVC.
Without that, the "best" matching PV could not be in the cache and controller
might bind the PVC to suboptiomal one.
This fixes integration test flake "Bind mismatch! Expected pvc-2 capacity
50000000000 but got pvc-2 capacity 52000000000".
Automatic merge from submit-queue
Considering all nodes for the scheduler cache to allow lookups
Fixes the actual issue that led me to create https://github.com/kubernetes/kubernetes/issues/22554
Currently the nodes in the cache provided to the predicates excludes the unschedulable nodes using field level filtering for the watch results. This results in the above issue as the `ServiceAffinity` predicate uses the cached node list to look up the node metadata for a peer pod (another pod belonging to the same service). Since this peer pod could be currently hosted on a node that is currently unschedulable, the lookup could potentially fail, resulting in the pod failing to be scheduled.
As part of the fix, we are now including all nodes in the watch results and excluding the unschedulable nodes using `NodeCondition`
@derekwaynecarr PTAL
Automatic merge from submit-queue
Enable WatchCache in test/integration/ tests
We already run cmd/integration/ with watch cache on. We should also run tests in test/integration/ with watch cache on.
@wojtek-t @lavalamp
Automatic merge from submit-queue
Add a custom main instead of the standard test main, to reduce stack …
Adds a custom test main handler (see: `TestMain` in https://golang.org/pkg/testing/ for details)
Partial fix for https://github.com/kubernetes/kubernetes/issues/25965
This does the standard timeout, but strips non-kubernetes stacks out of the stack trace (e.g. it filters things like:
```
goroutine 466 [IO wait, 7 minutes]:
net.runtime_pollWait(0x7fd74c4672c0, 0x72, 0xc821614000)
/usr/local/go/src/runtime/netpoll.go:160 +0x60
net.(*pollDesc).Wait(0xc8215c21b0, 0x72, 0x0, 0x0)
/usr/local/go/src/net/fd_poll_runtime.go:73 +0x3a
net.(*pollDesc).WaitRead(0xc8215c21b0, 0x0, 0x0)
/usr/local/go/src/net/fd_poll_runtime.go:78 +0x36
net.(*netFD).Read(0xc8215c2150, 0xc821614000, 0x1000, 0x1000, 0x0, 0x7fd74c491050, 0xc820014058)
/usr/local/go/src/net/fd_unix.go:250 +0x23a
net.(*conn).Read(0xc820a5a090, 0xc821614000, 0x1000, 0x1000, 0x0, 0x0, 0x0)
/usr/local/go/src/net/net.go:172 +0xe4
net/http.noteEOFReader.Read(0x7fd74c465258, 0xc820a5a090, 0xc8215f0068, 0xc821614000, 0x1000, 0x1000, 0x405773, 0x0, 0x0)
/usr/local/go/src/net/http/transport.go:1687 +0x67
net/http.(*noteEOFReader).Read(0xc8215ae1a0, 0xc821614000, 0x1000, 0x1000, 0xc82159ad1d, 0x0, 0x0)
<autogenerated>:284 +0xd0
bufio.(*Reader).fill(0xc8202a2b40)
/usr/local/go/src/bufio/bufio.go:97 +0x1e9
bufio.(*Reader).Peek(0xc8202a2b40, 0x1, 0x0, 0x0, 0x0, 0x0, 0x0)
/usr/local/go/src/bufio/bufio.go:132 +0xcc
net/http.(*persistConn).readLoop(0xc8215f0000)
/usr/local/go/src/net/http/transport.go:1073 +0x177
created by net/http.(*Transport).dialConn
/usr/local/go/src/net/http/transport.go:857 +0x10a6
```
We may want to get even more aggressive in the future.
@kubernetes/sig-testing
When we create a PV, we should created it withoud Spec.ClaimRef.UID.
In rare cases, when 'PV added' event with UID is processed before 'PVC
added' (created by for loop few lines above), the controller does not know
a PVC with this UID and considers the PV as released. Reclaim policy is
then executed and the PV is deleted and it's never bound.
With UID="", the controller waits for the PVC to get created and binds
it.
Different tests should use different objects and watchers - I noticed
sometimes an event from old tests leaked into subsequent test in the
same function.
And add some logs.
The test tries to bind configured nr. of PVs to the same nr. of PVCs.
'100' is used by default, which should take ~1-3 seconds (depends on log level).
Periodic sync is needed in rare cases, which may add another 10 seconds. - cache
from #25881 will help here and sync should not be needed at all.
The test is configurable and may be reused to measure binder performance.
Set KUBE_INTEGRATION_PV_* env. variables as described in
persistent_volume_test.go and run the tests:
# compile
$ cd test/integration
$ godep go test -tags 'integration no-docker' -c
# run the tests
$ KUBE_INTEGRATION_PV_SYNC_PERIOD=10s KUBE_INTEGRATION_PV_OBJECTS=1000 time ./integration.test -test.run TestPersistentVolumeMultiPVsPVCs -v 2
Log level '2' is useful to get timestamps of various events like
'TestPersistentVolumeMultiPVsPVCs: start' and 'TestPersistentVolumeMultiPVsPVCs:
claims are bound'.
- add more logs
- wait both for volume and claim to get bound
When binding volumes to claims the controller saves PV first and PVC right
after that. In theory, this saved PV could cause waitForPersistentVolumePhase
to finish and PVC could be checked in the test before the controller saves it.
So, wait for both PVC and PV to get bound and check the results only after
that. This is only a theory, there are no usable logs in integration tests.
Automatic merge from submit-queue
Use pause image depending on the server's platform when testing
Removed all pause image constant strings, now the pause image is chosen by arch. Part of the effort of making e2e arch-agnostic.
The pause image name and version is also now only in two places, and it's documented to bump both
Also removed "amd64" constants in the code. Such constants should be replaced by `runtime.GOARCH` or by looking up the server platform
Fixes: #22876 and #15140
Makes it easier for: #25730
Related: #17981
This is for `v1.3`
@ixdy @thockin @vishh @kubernetes/sig-testing @andyzheng0831 @pensu
Automatic merge from submit-queue
Attach Detach Controller Business Logic
This PR adds the meat of the attach/detach controller proposed in #20262.
The PR splits the in-memory cache into a desired and actual state of the world.
Split controller cache into actual and desired state of world.
Controller will only operate on volumes scheduled to nodes that
have the "volumes.kubernetes.io/controller-managed-attach" annotation.
Automatic merge from submit-queue
Fix panic in auth test failure
I got a spurious failure in the webhook integration test, but couldn't see the error returned because a panic was hit that assumed a body was always returned with the response
Automatic merge from submit-queue
Cache Webhook Authentication responses
Add a simple LRU cache w/ 2 minute TTL to the webhook authenticator.
Kubectl is a little spammy, w/ >= 4 API requests per command. This also prevents a single unauthenticated user from being able to DOS the remote authenticator.
- create 100 PV, ranging from 0 to 99GB; create 1 PVC to claim 50GB. Verify only one PV is bound and rest are pending
- create 2 PVs with different access modes (RWM, RWO), 1 PVC to claim RWM PV. Verify RWM is bound and RWO is not bound.
Signed-off-by: Huamin Chen <hchen@redhat.com>
Automatic merge from submit-queue
Refactor persistent volume controller
Here is complete persistent controller as designed in https://github.com/pmorie/pv-haxxz/blob/master/controller.go
It's feature complete and compatible with current binder/recycler/provisioner. No new features, it *should* be much more stable and predictable.
Testing
--
The unit test framework is quite complicated, still it was necessary to reach reasonable coverage (78% in `persistentvolume_controller.go`). The untested part are error cases, which are quite hard to test in reasonable way - sure, I can inject a VersionConflictError on any object update and check the error bubbles up to appropriate places, but the real test would be to run `syncClaim`/`syncVolume` again and check it recovers appropriately from the error in the next periodic sync. That's the hard part.
Organization
---
The PR starts with `rm -rf kubernetes/pkg/controller/persistentvolume`. I find it easier to read when I see only the new controller without old pieces scattered around.
[`types.go` from the old controller is reused to speed up matching a bit, the code looks solid and has 95% unit test coverage].
I tried to split the PR into smaller patches, let me know what you think.
~~TODO~~
--
* ~~Missing: provisioning, recycling~~.
* ~~Fix integration tests~~
* ~~Fix e2e tests~~
@kubernetes/sig-storage
<!-- Reviewable:start -->
---
This change is [<img src="http://reviewable.k8s.io/review_button.svg" height="35" align="absmiddle" alt="Reviewable"/>](http://reviewable.k8s.io/reviews/kubernetes/kubernetes/24331)
<!-- Reviewable:end -->
Fixes#15632
Automatic merge from submit-queue
Webhook Token Authenticator
Add a webhook token authenticator plugin to allow a remote service to make authentication decisions.
Automatic merge from submit-queue
Let the dynamic client take runtime.Object instead of v1.ListOptions
so that I can pass whatever version of ListOptions to the List/Watch/DeleteCollection methods.
cc @krousey
Automatic merge from submit-queue
enable resource name and service account cases for impersonation
Adds the resource name check since that attribute was added for authorization. Also adds a check against a separate resource for service accounts. Allowing impersonation of service accounts to use a different resource check places control of impersonation with the same users to have the power to get the SA tokens directly.
@kubernetes/kube-iam
@sgallagher FYI
Automatic merge from submit-queue
Abstract node side functionality of attachable plugins
- Create PhysicalAttacher interface to abstract MountDevice and
WaitForAttach.
- Create PhysicalDetacher interface to abstract WaitForDetach and
UnmountDevice.
- Expand unit tests to check that Attach, Detach, WaitForAttach,
WaitForDetach, MountDevice, and UnmountDevice get call where
appropriet.
Physical{Attacher,Detacher} are working titles suggestions welcome. Some other thoughts:
- NodeSideAttacher or NodeAttacher.
- AttachWatcher
- Call this Attacher and call the Current Attacher CloudAttacher.
- DeviceMounter (although there are way too many things called Mounter right now :/)
This is to address: https://github.com/kubernetes/kubernetes/pull/21709#issuecomment-192035382
@saad-ali
Automatic merge from submit-queue
Support persisting config from kubecfg AuthProvider plugins
Plumbs through an interface to the plugin that can persist a `map[string]string` config for just that plugin. Also adds `config` to the AuthProvider serialization type, and `Login()` to the AuthProvider plugin interface.
Modified the gcp AuthProvider to cache short-term access tokens in the kubecfg file.
Builds on #23066
@bobbyrullo @deads2k @jlowdermilk @erictune
Automatic merge from submit-queue
Reimplement 'pause' in C - smaller footprint all around
Statically links against musl. Size of amd64 binary is 3560 bytes.
I couldn't test the arm binary since I have no hardware to test it on, though I assume we want it to work on a raspberry pi.
This PR also adds the gcc5/musl cross compiling image used to build the binaries.
@thockin
The codec factory should support two distinct interfaces - negotiating
for a serializer with a client, vs reading or writing data to a storage
form (etcd, disk, etc). Make the EncodeForVersion and DecodeToVersion
methods only take Encoder and Decoder, and slight refactoring elsewhere.
In the storage factory, use a content type to control what serializer to
pick, and use the universal deserializer. This ensures that storage can
read JSON (which might be from older objects) while only writing
protobuf. Add exceptions for those resources that may not be able to
write to protobuf (specifically third party resources, but potentially
others in the future).
- Expand Attacher/Detacher interfaces to break up work more
explicitly.
- Add arguments to all functions to avoid having implementers store
the data needed for operations.
- Expand unit tests to check that Attach, Detach, WaitForAttach,
WaitForDetach, MountDevice, and UnmountDevice get call where
appropriet.
Automatic merge from submit-queue
Provide flags to use etcd3 backed storage
ref: #24405
What's in this PR?
- Add a new flag "storage-backend" to choose "etcd2" or "etcd3". By default (i.e. empty), it's "etcd2".
- Take out etcd config code into a standalone package and let it create etcd2 or etcd3 storage backend given user input.
Automatic merge from submit-queue
add user.Info.GetExtra
I found myself wanting this field (or something like it), when trying to plumb the information about which scopes a particular token has.
Only the token authenticators have that information and I don't want tokens to leak past the authenticator. I thought about extending the `authenticator.Token` interface to include scopes (`[]string`), but that felt a little specific for what I wanted to do. I came up with this as an alternative.
It allows the token authenticator to fill in the information and authorizers already get handed the `user.Info`. It means that implementors can choose to tie the layers together if they wish, using whatever data they think is best.
@kubernetes/kube-iam
Add tests to watch behavior in both protocols (http and websocket)
against all 3 media types. Adopt the
`application/vnd.kubernetes.protobuf;stream=watch` media type for the
content that comes back from a watch call so that it can be
distinguished from a Status result.
Automatic merge from submit-queue
Make etcd cache size configurable
Instead of the prior 50K limit, allow users to specify a more sensible size for their cluster.
I'm not sure what a sensible default is here. I'm still experimenting on my own clusters. 50 gives me a 270MB max footprint. 50K caused my apiserver to run out of memory as it exceeded >2GB. I believe that number is far too large for most people's use cases.
There are some other fundamental issues that I'm not addressing here:
- Old etcd items are cached and potentially never removed (it stores using modifiedIndex, and doesn't remove the old object when it gets updated)
- Cache isn't LRU, so there's no guarantee the cache remains hot. This makes its performance difficult to predict. More of an issue with a smaller cache size.
- 1.2 etcd entries seem to have a larger memory footprint (I never had an issue in 1.1, even though this cache existed there). I suspect that's due to image lists on the node status.
This is provided as a fix for #23323
Automatic merge from submit-queue
genericapiserver: Moving InstallSwaggerAPI to Run
Ref https://github.com/kubernetes/kubernetes/pull/21190#discussion_r57494673
Moving InstallSwaggerAPI() from InstallAPIGroups() to Run(). This allows the use of InstallAPIGroups() multiple times or using InstallAPIGroup() directly.
cc @jianhuiz @kubernetes/sig-api-machinery
