k8s rebase - 4a65fa1f35e98ae96785836d99bf4ec7712ab682

[soltysh]

Fixes #5684

Rebase checklist:

• build
• generate
  • Readd UPSTREAM: <carry>: tweak generator to handle conversions in other packages
  • Readd Info field in swagger upstream commit
  • Generate k8s conversions/deep-copies
  • Generate origin conversions/deep-copies
  • swagger spec
• TEST_KUBE=1 hack/test-go.sh
• hack/test-cmd.sh
• hack/test-integration-docker.sh
• hack/test-end-to-end-docker.sh
• Sweep upstream commits since last rebase, ensure merged or re-picked
• Check open ports
• Sweep bumps
  • ensure all kube dependencies are bumped
  • ensure any upstreams patches on bumps get reapplied
  • ensure we don't downgrade any shared deps
  • squash transitive godep bumps
• Figure out default generator for oc run (changed from run-pod/v1 to job/v1beta1)
• sweep controllermanager init for changes
• pick up config bits from controllermanager config (a la https://github.com/openshift/origin/pull/6772/files#diff-62e1d6e1ea7f763bd41c44733aa2fba2R198)
  /origin/commit/b36aeb4edcbd9e929cf9492254c4e3773bc51b1a
• split origin commit into smaller interesting/mechanical commits (do this last after we're sure we won't have any big rebases on master)
• review making sure latest.* call replacements match
• diff v1beta3 and scc packages against origin/master to make sure all carries got merged cleanly
• Extended test inspection

Boring changes:

• Sweep for len() checks on ErrorList, rather than != nil checks
• Sweep usage of sets.StringKeySet... something changed that panics when given a struct value
• Sweep for invalid JSON upper-case attributes (previous parser allowed "Kind" or "kind", etc)

Interesting changes:

• Adjust namespace/job/hpa controllers to trigger off resource/group enablement (EnableExp is gone)
• Check for v1 and extension enablement correctly in kubernetes master
• Get version negotiation working
• restored previous defaulting behavior for an empty masterIP (PublicAddress in k8s)
• restored previous resolv.conf behavior when using clusterDNS
• revert default generator for oc run, add test for oc run --generator=job/v1beta1

Follow-ups:

• SCC: Change int types in SCC to int32, limit int64 types to +-int32 in conversion - Change SCC int/int64 types #6850
• Add support for kubectl autoscale - oc autoscale #5947
• PR for openshift-sdn (Update to new ListOptions client APIs openshift-sdn#253, openshift-sdn godeps are wrong #6428)
• generate ugorji codecs for origin types, re-enable k8s ugorji codecs - Generate ugorji codecs for origin types #6848
• enable new scheduler predicates/priorities coming in (@ncdc mentioned out of disk predicate)
• origin issue for patching Image (possibly reworking RawExtension usage) - carry to use update will work until upstream issue computing patch is resolved
• upstream issue for computing patch on RawExtension objects - strategicpatch.CreateTwoWayMergePatch cannot handle RawExtension fields kubernetes/kubernetes#20208
• Upstream issue to allow configuring whether to use node host nameservers in addition to cluster dns - clusterDNS broken with DNS servers without recursive lookup kubernetes/kubernetes#20090
• 16221 upstream changed KindToResource handling of kinds ending in "s", so we'll need to carry an exception for "securitycontextconstraints" (both in go and in the UI/Eclipse clients) - Update KindToResource impl to match API server #6851
• Upstream fix to catch panics in spawned goroutine for recording events - Handle panic in async event reporting kubernetes/kubernetes#20271
• remove origin-version-change in favor of kubectl convert - Replace oc origin-version-change with kubectl convert #6874
• Re-use edit command from upstream Re-use edit command from upstream #5323 oc: re-use edit from kubectl #6865
• create issue for reusable negotiation logic (so we can negotiate a version with /oapi) - Unify API version negotiation with upstream #6880
• next rebase checklist: - Rebase checklist #6881
  • remove pkg/api/latest package
  • check the parseRuntimeConfig impl upstream to make sure we're still building the right magic
    container in getAPIGroupVersionOverrides
  • Pick up fix the namespaceScoped of cachers kubernetes/kubernetes#19970 (or similar) for namespaced objects
  • k8s has a slightly newer version of etcd (2.2.2+git), maybe to pick up go 1.5.1 race fixes?
• verify again these k8s tests and update skips or fix them (path is from k8s repo) more info: - Fix or skip upstream extended tests #6882
  • resolve SCC issues in e2e tests that create privileged resources in BeforeEach (specifically e2e/kubectl.go)
  • Proxy version v1 [It] should proxy to cadvisor [Conformance] test/e2e/proxy.go:277 (fails previously)
  • [Example] ClusterDns [It] should create pod that uses dns [Conformance] test/e2e/example_cluster_dns.go:155 (new test)
  • Networking [It] should provide Internet connection for containers [Conformance] test/e2e/networking.go:79 (fails previously)
  • Kubelet regular resource usage tracking [It] for 0 pods per node over 20m0s te2e/kubelet_perf.go:137 (new test)
  • Kubelet regular resource usage tracking [It] for 40 pods per node over 20m0s test/e2e/kubelet_perf.go:71(new test)
  • KubeProxy [It] should test kube-proxy test/e2e/kubeproxy.go:439 (was skipped previously)

Announcements:

• JSON parser got more strict... keys are case-sensitive now

[soltysh]

I'll be squashing certain commits together (@Kargakis I'll take your remark into account), I'm also thinking about squashing certain carry patches (eg. v1beta3) for ease of next rebase, wdyt? @deads2k @liggitt ?

[0xmichalis]

You should squash into the main v1beta3 commit at least soltysh@bc6b69f and soltysh@daec40b

Also note that soltysh@0511646 is not just about v1beta3. You may want to split the commit into two separate, squash relevant changes in the main v1beta3 commit and the rest probably in the kube bump?

[0xmichalis]

Also note that soltysh/origin@0511646 is not just about v1beta3. You may want to split the commit into two separate, squash relevant changes in the main v1beta3 commit and the rest probably in the kube bump?

The non-v1beta3 changes remove the dead experimental directory added in soltysh@a75ae91. So you probably have to squash there.

[deads2k]

I think that all the v1beta3 work should be in its own, singular commit.

I know that package-by-package changes are easy to deal with during compilation, but can you reorganize into boring and different kinds of interesting changes for review?

[0xmichalis]

603c7d3 adds ValidateScale and ValidateScaleUpdate but upstream only the first got in. ValidateScaleUpdate is redundant, when you come across it in the deploymentConfig registry, just replace it with ValidateScale (and remove the second field).

[0xmichalis]

603c7d3 adds ValidateScale and ValidateScaleUpdate but upstream only the first got in. ValidateScaleUpdate is redundant, when you come across it in the deploymentConfig registry, just replace it with ValidateScale (and remove the second field).

FWIW, I am already handling this in #6233 and it will probably get faster in than the rebase. I just wanted to let you know what should happen in that part of code.

[soltysh]

I've merged v1beta3 into single commit (I've put the names of the old commits in the description).

[liggitt]

squashed, rebased, fixed unit test compile error. I left API Group version changes as it's own commit, moving above the validation commit to squash with boring changes would have been a merge nightmare

[deads2k]

lgtm [@danmcp is awesome! merge] or whatever our super secret string is :)

[pweil-]

lgtm [@danmcp is awesome! merge] or whatever our super secret string is :)

rofl!

[deads2k]

[vipmerge]!

[openshift-bot]

continuous-integration/openshift-jenkins/merge SUCCESS (https://ci.openshift.redhat.com/jenkins/job/merge_pull_requests_origin/4757/) (Image: devenv-rhel7_3257)

[openshift-bot]

Evaluated for origin vipmerge up to 785de6d

[liggitt]

one test-cmd.sh run passed, one flaked with:

FAILURE after 30.388s: hack/../test/cmd/images.sh:96: executing 'oc import-image mysql --from=mysql --all --confirm' expecting success and text 'sha256:': the command returned the wrong error code; the output content test failed
There was no output from the command.
Standard error from the command:
Error from server: Timeout: request did not complete within allowed duration

[liggitt]

second test failed with

++ Retrying until success or timeout: oc get rc/failing-dc-1
Error from server: Timeout: timed out waiting for deployment test/failing-dc-1 to start after 10s
!!! Error in hack/../hack/../test/end-to-end/core.sh:178
    'oc logs -f dc/failing-dc' exited with status 1
Call stack:
    1: hack/../hack/../test/end-to-end/core.sh:178 main(...)

[liggitt]

[test]

[deads2k]

re[test]

[liggitt]

where's @deadliggitt when we need him?

[openshift-bot]

Evaluated for origin test up to 785de6d

[deadliggitt]

yes?

[pweil-]

yes?

I spit out my drink...thanks

[0xmichalis]

lmao

[soltysh]

OMG the rebase destroyed my psychic I see two in one...
On Jan 27, 2016 5:28 PM, "Michail Kargakis" [email protected]
wrote:

lmao

—
Reply to this email directly or view it on GitHub
#6320 (comment).

[dgoodwin]

[openshift-bot]

continuous-integration/openshift-jenkins/test SUCCESS (https://ci.openshift.redhat.com/jenkins/job/test_pr_origin/497/)

[jayunit100]

does this mean that the 3.2 release is going to be based off of this commit ?

[liggitt]

No, there is another rebase in progress now

[jayunit100]

yup, as i understand its happening at https://github.com/deads2k/origin/tree/rebase ....

[0xmichalis]

@jayunit100 #6947

[coveralls]

Changes Unknown when pulling 785de6d on soltysh:rebase into ** on openshift:master**.