Instead of endpoints being a flat list, it is now a list of "subsets"
where each is a struct of {Addresses, Ports}. To generate the list of
endpoints you need to take union of the Cartesian products of the
subsets. This is compact in the vast majority of cases, yet still
represents named ports and corner cases (e.g. each pod has a different
port number).
This also stores subsets in a deterministic order (sorted by hash) to
avoid spurious updates and comparison problems.
This is a fully compatible change - old objects and clients will
keepworking as long as they don't need the new functionality.
This is the prep for multi-port Services, which will add API to produce
endpoints in this new structure.
Currently, API server is not aware of the static pods (manifests from
sources other than the API server, e.g. file and http) at all. This is
inconvenient since users cannot check the static pods through kubectl.
It is also sub-optimal because scheduler is unaware of the resource
consumption by these static pods on the node.
This change syncs the information back to the API server by creating a
mirror pod via API server for each static pod.
- Kubelet creates containers for the static pod, as it would do
normally.
- If a mirror pod gets deleted, Kubelet will re-create one. The
containers are sync'd to the static pods, so they will not be
affected.
- If a static pod gets removed from the source (e.g. manifest file
removed from the directory), the orphaned mirror pod will be deleted.
Note that because events are associated with UID, and the mirror pod has
a different UID than the original static pod, the events will not be
shown for the mirror pod when running `kubectl describe pod
<mirror_pod>`.
The integration test will fail if I check in my pending PR
to remove boundPods. Kubernetes eventually does the right
thing, but getting the integration test to check expectations
is hard because the scheduling behavior is unpredictable.
The boundPods removal is needed to fix P1 bug and speeds up
the scheduler considerably.
Revise our code to only call Request.Namespace() if a namespace
*should* be present. For root scoped resources, namespace should
be ignored. For namespaced resources, it is an error to have
Namespace=="".
--sync_nodes=false gives user flexibility to add/remove nodes in the
cluster using REST api/kubectl cli and at the same time can use
cloud provider for other resources like persistent disks, etc.
Currently, the validation logic validates fields in an object and supply default
values wherever applies. This change factors out defaulting to a set of
defaulting callback functions for decoding (see #1502 for more discussion).
* This change is based on pull request 2587.
* Most defaulting has been migrated to defaults.go where the defaulting
functions are added.
* validation_test.go and converter_test.go have been adapted to not testing the
default values.
* Fixed all tests with that create invalid objects with the absence of
defaulting logic.
Allow the master to have pod/node cache timeouts controlled via a config
flag for integration tests.
Move integration test to '127.0.0.1' so that it correctly returns a health
check, and enable health check testing on the integration test.
Break up the monolithic volumes code in kubelet into very small individual
modules with a well-defined interface. Move them all into their own packages
and beef up testing along the way.
Added a kubelet config source for watching pods on apiserver.
The pods are converted to boundpods for merging with other
config sources.
The preferred way to create a kubelet is now to pass an apiserver
client but not an etcd client. Changed cmd/integration to use
apiserver to talk to kubelets. And cmd/kubernetes.
Unit, integration, and e2e tests pass, except for a failure of the pd
e2e test which was unrelated.
Make all kubelet config sources ensure that UID and Namespace are defaulted, if
need be.
We can *almost* disable the "if blank" logic for UID, except for tests that
call APIs that do not run through SyncPods. We really ought to be enforcing
invariants better.
Added basic interface for authorizer implementations.
Added default "authorize everything" and "authorize nothing
implementations.
Added authorization check immediately after authentication check.
Added an integration test of authorization at the HTTP level of
abstraction.
Callsites no longer allocate a mux.
Master now exposes method to install handlers
which use the master's auth code. Not used
but forks (openshift) are expected to use these
methods. These methods will later be a point
for additional plug-in functionality.
Integration tests now use the master-provided
handler which has auth, rather than using the mux,
which didn't. Fix TestWhoAmI now that /_whoami
sits behind auth.
Moved code from cmd/apiserver to pkg/master.
test/integration/client_test made to use a master object,
instead of an apiserver.Handle.
Subsequent PRs will move more handler-installation into
pkg/master, with the goal that every http.Handler of a
standalone apiserver process can also be tested
in a "testing"-style go test.
In particular, a subsequent PR will test
authorization.
* Allows consumers to provide their own transports for common cases.
* Supports KUBE_API_VERSION on test cases for controlling which
api version they test against
* Provides a common flag registration method for CLIs that need
to connect to an API server (to avoid duplicating flags)
* Ensures errors are properly returned by the server
* Add a Context field to client.Config
* Defaults to v1beta1
* apiserver takes -storage_version which controls etcd storage version
and the version of the client used to connect to other apiservers
* Changed signature of client.New to add version parameter
* All controller code and component code prefers the oldest (most common)
server version
* Make Codec separate from Scheme
* Move EncodeOrDie off Scheme to take a Codec
* Make Copy work without a Codec
* Create a "latest" package that imports all versions and
sets global defaults for "most recent encoding"
* v1beta1 is the current "latest", v1beta2 exists
* Kill DefaultCodec, replace it with "latest.Codec"
* This updates the client and etcd to store the latest known version
* EmbeddedObject is per schema and per package now
* Move runtime.DefaultScheme to api.Scheme
* Split out WatchEvent since it's not an API object today, treat it
like a special object in api
* Kill DefaultResourceVersioner, instead place it on "latest" (as the
package that understands all packages)
* Move objDiff to runtime.ObjectDiff
Convert host:port and URLs passed to client.New() into the proper
values, and return an error if the value is invalid. Change CLI
to return an error if -master is invalid. Remove Client.rawRequest
which was not in use, and fix the involved tests. Add NewOrDie
Preserves the behavior of the client to not auth when a non-https
URL is passed (although in the future this should be corrected).
Also rename some to other names that make better reading. There are still a
bunch of "make" functions but they do things like assemble a string from parts
or build an array of things. It seemed that "make" there seemed fine. "New"
is for "constructors".
Prepare for running multiple API versions on the same HTTP server
by decoupling some of the mechanics of apiserver. Define a new
APIGroup object which represents a version of the API.
The sync frequency should be part of the syncLoop and resync no
less often than every X seconds. The current implementation runs
even if a config update was delivered less than X seconds ago.
To make sure the etcd watcher works, I changed the replication
controller to use watch.Interface. I made apiserver support watches on
controllers, so replicationController can be run only off of the
apiserver. I made sure all the etcd watch testing that used to be in
replicationController is now tested on the new etcd watcher in
pkg/tools/.
Setting up a new master.Master instance requires passing
around too many arguments.
Add a master.Config type and group related master configs.
Refactor all commands to instantiate new masters using a
master.Config struct.
The current integration tests do not return after delegating
HTTP requests, as a result an extra call to response.WriteHeader
is made for every request.
Fix the issue by returning after delegating HTTP requests.
Also transfer the Kubelet from using ContainerManifest.ID to source specific
identifiers with namespacing. Move goroutine behavior out of kubelet/ and
into integration.go and cmd/kubelet/kubelet.go for better isolation.
The -etcd_servers flag is used inconsistently by the Kubernetes commands,
both externally and internally.
This patch fixes the issue by using the same type to represent a list of
etcd servers internally, and declares the -etcd_servers flag consistently
across all commands.
This patch should be 100% backwards compatible with no changes in behavior.
Improve apiserver/logger.go's interface (it's pretty cool now).
Improve apiserver's error reporting to clients.
Improve client's handling of errors from apiserver.
Make failed PUTs return 409 (conflict)-- http status codes are amazingly
well defined for what we're doing!
1) imported glog to third_party (previous commit)
2) add support for third_party/update.sh to update just one pkg
3) search-and-replace:
s/log.Printf/glog.Infof/
s/log.Print/glog.Info/
s/log.Fatalf/glog.Fatalf/
s/log.Fatal/glog.Fatal/
4) convert glog.Info.*, err into glog.Error*
Adds some util interfaces to logging and calls them from each cmd, which
will set the default log output to write to glog. Pass glog-wrapped
Loggers to etcd for logging.
Log files will go to /tmp - we should probably follow this up with a
default log dir for each cmd.
The glog lib is sort of weak in that it only flushes every 30 seconds, so
we spin up our own flushing goroutine.
Tim Hockin