* Added a godep-update-k8s script for upgrading.
$ KUBE_VERSION ./hack/godeps/godep-update-k8s.sh
* Added client-go version ldflags to localkube and minikube. The
version package is now split in both kubernetes and client-go, so we
need to supply both ldflags.
* Change cross builds to use BUILD_IN_DOCKER
* Add a patch for kube-proxy
The setup of kube-proxy recently changed, making the registering of the
component config a fatal error instead of a warning. This patch
reverts it to be a warning, which is a no-op for us anyways since we
don't supply a config file.
Bad tests that panic won't clean themselves up properly. By putting
these cleanup commands before we run the minikube integration tests, we
ensure that minikube will have a clean environment to run.
These all run in common.sh to make sure that even hypervisors that they
arent testing don't interfere with their own test. OS specific
commands will fail silently.
Kubernetes vendors these repos directly from the staging directory even
though they now have their own repositories on github. We should use
the code from the staging/ directory since it is guaranteed to be
compatible with the version of kube we are vendoring.
Reference: https://github.com/kubernetes/kubernetes/pull/44771
Fixes https://github.com/kubernetes/minikube/issues/1252
TPRs are incorrectly coupled with the RestMapper right now. The real
solution is for TPRs to not register themselves with the RestMapper.
This is a short term patch for minikube until the work is done
upstream. On start/stop, the namespace controller and the garbage
collector controller both call this code and panic since TPRs have
registered themselves with enabled versions but have no group metadata.
* We should be calling the hack/godep-restore script in kubernetes
instead of the normal godep restore ./...
* There was a typo in the case where a user didn't already have the
kubernetes repo cloned.
When new iso changes are commited to master, don't count them as diffs
with old PRs that haven't been rebased. That way, we don't rebuild the
iso for PRs that don't contain the upstream fixes yet.
When changes are made to the ISO folder, rebuild and upload the newly
modified image. Use this uploaded image as the default in the e2e test
binaries, so that the e2e test suite runs against the PR's changes.
TODO: Subsequent PRs (no changes to ISO) will still run against the
default version in the makefile. We should figure out if we want to
always run tests against the "latest" image.
Sometimes gsutil cp -r can miss files. We also don't need to copy over
every file. Additionally, I started to abstract the OS_ARCH away so
that we can more easily template the jenkins OS tests.
Matt Rickard
Matt Rickard
Arve Knudsen
Budh Ram Gurung
Aaron Prindle
dlorenc
Giancarlo Rubio
Xinbo Weng
dlorenc