Kube 1.4 cherry picks

[soltysh]

Resovles #11565.

@liggitt @smarterclayton ptal
@pweil- @mfojtik fyi

The summary (master #PR: title (cherry-pick #PR) - why):

• 33052: Bump cAdvisor Godeps version to official release tag - Only updated Godep.json to match 1.4 release, no actual change happened.
• 33086: Fix possible panic in PodAffinityChecker (33166) - Small fix, but can save us from nil panic
• 33170: Remove closing audit log file and add error check when writing to audit (33172) - Improve audit logging, especially when log operation fails
• 33346: disallow user to update loadbalancerSourceRanges (33513) - Changes Service validation, not sure we want to wait, but there's a TODO about being
  able to modify that
• 32807: Fix race condition in setting node statusUpdateNeeded flag (34038) - Race fix
• 32914: Limit the number of names per image reported in the node status (34157) - Fix OOM master etcd on big clusters
• 34000: Set deserialization cache size based on target memory usage (34213) - To reduce memory usage to reasonable levels in smaller clusters
• 33796: Fix issue in updating device path when volume is attached multiple times (34215) - P0 fixing wrong AWS volume mount issue
• 33735: Fixes in HPA: consider only running pods; proper denominator in avg (34224) - Update which pods are calculated for HPA
• 33968: scheduler: initialize podsWithAffinity (34266) - Fix panic in scheduler
• 34076: Remove headers that are unnecessary for proxy target (34280) - P1 issue
• 34694: Handle DeletedFinalStateUnknown in NodeController (34803) - Fix panic in NodeController caused by receiving DeletedFinalStateUnknown object from the cache.
• 34809: NodeController waits for informer sync before doing anything (34813) - P0 issue
• 34251: Fix nil pointer issue when getting metrics from volume mounter (34296) - Possible nil pointer panic
• 34851: Only wait for cache syncs once in NodeController (34861) - P0 in branch 1.4, P1 in master
• 34895: Fix non-starting node controller in 1.4 branch - P0 in branch 1.4
• 34955: HPA: fixed wrong count for target replicas calculations (35406) - P0 issue
• 35071: Change merge key for VolumeMount to mountPath (35208) - Fixes VolumeMounts edit operation
• 35273: Fixed mutation warning in Attach/Detach controller (35633) - Fixes cache mutation
• 34368: Node status updater should SetNodeStatusUpdateNeeded if it fails to (35653) - Fix node status update

Not sure about:

• 35205 - contains fixes to etcd3
• 33699 - moves informer to client

[smarterclayton]

Are any of those cherry picks not from 1.4?

On Fri, Oct 28, 2016 at 10:28 AM, Maciej Szulik notifications@github.com
wrote:

Resovles #11565 #11565.

@liggitt https://github.com/liggitt @smarterclayton
https://github.com/smarterclayton ptal
@pweil- https://github.com/pweil- @mfojtik https://github.com/mfojtik
fyi
The summary (master #PR: title (cherry-pick #PR) - why):

33052: Bump cAdvisor Godeps version to official release tag - Only
updated Godep.json to match 1.4 release, no actual change happened.

33086: Fix possible panic in PodAffinityChecker (33166) - Small fix,
but can save us from nil panic

33170: Remove closing audit log file and add error check when writing
to audit (33172) - Improve audit logging, especially when log operation
fails

33346: disallow user to update loadbalancerSourceRanges (33513) -
Changes Service validation, not sure we want to wait, but there's a TODO
about being
able to modify that

32807: Fix race condition in setting node statusUpdateNeeded flag
(34038) - Race fix

32914: Limit the number of names per image reported in the node status
(34157) - Fix OOM master etcd on big clusters

34000: Set deserialization cache size based on target memory usage
(34213) - To reduce memory usage to reasonable levels in smaller clusters

33796: Fix issue in updating device path when volume is attached
multiple times (34215) - P0 fixing wrong AWS volume mount issue

33735: Fixes in HPA: consider only running pods; proper denominator in
avg (34224) - Update which pods are calculated for HPA

33968: scheduler: initialize podsWithAffinity (34266) - Fix panic in
scheduler

34076: Remove headers that are unnecessary for proxy target (34280) -
P1 issue

34694: Handle DeletedFinalStateUnknown in NodeController (34803) - Fix
panic in NodeController caused by receiving DeletedFinalStateUnknown object
from the cache.

34809: NodeController waits for informer sync before doing anything
(34813) - P0 issue

34251: Fix nil pointer issue when getting metrics from volume mounter
(34296) - Possible nil pointer panic

34851: Only wait for cache syncs once in NodeController (34861) - P0
in branch 1.4, P1 in master

34895: Fix non-starting node controller in 1.4 branch - P0 in branch
1.4

34955: HPA: fixed wrong count for target replicas calculations (35406)

• P0 issue

  35071: Change merge key for VolumeMount to mountPath (35208) - Fixes

  VolumeMounts edit operation

  35273: Fixed mutation warning in Attach/Detach controller (35633) -

  Fixes cache mutation

  34368: Node status updater should SetNodeStatusUpdateNeeded if it
  fails to (35653) - Fix node status update

Not sure about:

• 35205 - contains fixes to etcd3
• 33699 - moves informer to client

---

You can view, comment on, or merge this pull request online at:

#11642
Commit Summary

• bump(github.com/google/cadvisor): 0cdf4912793fac9990de3790c27334
  2ec31817fb
• UPSTREAM: 33086: Fix possible panic in PodAffinityChecker
• UPSTREAM: 33170: Remove closing audit log file and add error check
  when writing to audit
• UPSTREAM: 33346: disallow user to update loadbalancerSourceRanges
• UPSTREAM: 32807: Fix race condition in setting node
  statusUpdateNeeded flag
• UPSTREAM: 32914: Limit the number of names per image reported in the
  node status
• UPSTREAM: 34000: Set deserialization cache size based on target
  memory usage
• Deserialization cache changes
• UPSTREAM: 33796: Fix issue in updating device path when volume is
  attached multiple times
• UPSTREAM: 33735: Fixes in HPA: consider only running pods; proper
  denominator in avg
• UPSTREAM: 33968: scheduler: initialize podsWithAffinity
• UPSTREAM: 34076: Remove headers that are unnecessary for proxy target
• UPSTREAM: 34694: Handle DeletedFinalStateUnknown in NodeController
• UPSTREAM: 34809: NodeController waits for informer sync before doing
  anything
• UPSTREAM: 34251: Fix nil pointer issue when getting metrics from
  volume mounter
• UPSTREAM: 34851: Only wait for cache syncs once in NodeController
• UPSTREAM: 34895: Fix non-starting node controller in 1.4 branch
• UPSTREAM: 34955: HPA: fixed wrong count for target replicas
  calculations
• UPSTREAM: 35071: Change merge key for VolumeMount to mountPath
• UPSTREAM: 35273: Fixed mutation warning in Attach/Detach controller
• UPSTREAM: 34368: Node status updater should
  SetNodeStatusUpdateNeeded if it fails to

File Changes

• M Godeps/Godeps.json
  https://github.com/openshift/origin/pull/11642/files#diff-0 (156)
• M pkg/cmd/server/kubernetes/master_config.go
  https://github.com/openshift/origin/pull/11642/files#diff-1 (21)
• M pkg/cmd/server/kubernetes/master_config_test.go
  https://github.com/openshift/origin/pull/11642/files#diff-2 (2)
• M vendor/k8s.io/kubernetes/cmd/kube-apiserver/app/server.go
  https://github.com/openshift/origin/pull/11642/files#diff-3 (22)
• M vendor/k8s.io/kubernetes/pkg/api/v1/types.go
  https://github.com/openshift/origin/pull/11642/files#diff-4 (2)
• M vendor/k8s.io/kubernetes/pkg/api/validation/validation.go
  https://github.com/openshift/origin/pull/11642/files#diff-5 (3)
• M vendor/k8s.io/kubernetes/pkg/api/validation/validation_test.go
  https://github.com/openshift/origin/pull/11642/files#diff-6 (19)
• M vendor/k8s.io/kubernetes/pkg/apiserver/audit/audit.go
  https://github.com/openshift/origin/pull/11642/files#diff-7 (12)
• M vendor/k8s.io/kubernetes/pkg/auth/handlers/handlers.go
  https://github.com/openshift/origin/pull/11642/files#diff-8 (6)
• M vendor/k8s.io/kubernetes/pkg/auth/handlers/handlers_test.go
  https://github.com/openshift/origin/pull/11642/files#diff-9 (10)
• M vendor/k8s.io/kubernetes/pkg/controller/framework/shared_
  informer.go
  https://github.com/openshift/origin/pull/11642/files#diff-10 (31)
• M vendor/k8s.io/kubernetes/pkg/controller/node/nodecontroller.go
  https://github.com/openshift/origin/pull/11642/files#diff-11 (191)
• M vendor/k8s.io/kubernetes/pkg/controller/podautoscaler/
  horizontal.go
  https://github.com/openshift/origin/pull/11642/files#diff-12 (27)
• M vendor/k8s.io/kubernetes/pkg/controller/podautoscaler/
  horizontal_test.go
  https://github.com/openshift/origin/pull/11642/files#diff-13 (41)
• M vendor/k8s.io/kubernetes/pkg/controller/podautoscaler/
  metrics/metrics_client.go
  https://github.com/openshift/origin/pull/11642/files#diff-14 (34)
• M vendor/k8s.io/kubernetes/pkg/controller/podautoscaler/
  metrics/metrics_client_test.go
  https://github.com/openshift/origin/pull/11642/files#diff-15 (41)
• M vendor/k8s.io/kubernetes/pkg/controller/volume/
  attachdetach/cache/actual_state_of_world.go
  https://github.com/openshift/origin/pull/11642/files#diff-16 (57)
• M vendor/k8s.io/kubernetes/pkg/controller/volume/
  attachdetach/cache/actual_state_of_world_test.go
  https://github.com/openshift/origin/pull/11642/files#diff-17 (161)
• M vendor/k8s.io/kubernetes/pkg/controller/volume/
  attachdetach/statusupdater/node_status_updater.go
  https://github.com/openshift/origin/pull/11642/files#diff-18 (39)
• M vendor/k8s.io/kubernetes/pkg/genericapiserver/genericapiserver.go
  https://github.com/openshift/origin/pull/11642/files#diff-19 (3)
• M vendor/k8s.io/kubernetes/pkg/genericapiserver/options/
  server_run_options.go
  https://github.com/openshift/origin/pull/11642/files#diff-20 (6)
• M vendor/k8s.io/kubernetes/pkg/kubelet/kubelet.go
  https://github.com/openshift/origin/pull/11642/files#diff-21 (3)
• M vendor/k8s.io/kubernetes/pkg/kubelet/kubelet_node_status.go
  https://github.com/openshift/origin/pull/11642/files#diff-22 (16)
• M vendor/k8s.io/kubernetes/pkg/kubelet/kubelet_node_status_test.go
  https://github.com/openshift/origin/pull/11642/files#diff-23 (10)
• M vendor/k8s.io/kubernetes/pkg/kubelet/kubelet_test.go
  https://github.com/openshift/origin/pull/11642/files#diff-24 (2)
• M vendor/k8s.io/kubernetes/pkg/kubelet/kubelet_volumes.go
  https://github.com/openshift/origin/pull/11642/files#diff-25 (6)
• M vendor/k8s.io/kubernetes/pkg/kubelet/volumemanager/cache/
  actual_state_of_world.go
  https://github.com/openshift/origin/pull/11642/files#diff-26 (9)
• M vendor/k8s.io/kubernetes/pkg/util/wait/wait.go
  https://github.com/openshift/origin/pull/11642/files#diff-27 (5)
• M vendor/k8s.io/kubernetes/plugin/pkg/scheduler/
  algorithm/predicates/predicates.go
  https://github.com/openshift/origin/pull/11642/files#diff-28 (2)
• M vendor/k8s.io/kubernetes/plugin/pkg/scheduler/
  schedulercache/cache.go
  https://github.com/openshift/origin/pull/11642/files#diff-29 (4)

Patch Links:

• https://github.com/openshift/origin/pull/11642.patch
• https://github.com/openshift/origin/pull/11642.diff

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#11642, or mute the thread
https://github.com/notifications/unsubscribe-auth/ABG_p3H-V5YI0xSiIOqkDc2ADA8jVCajks5q4gaRgaJpZM4Kjf6r
.

[smarterclayton]

[test]

[soltysh]

Are any of those cherry picks not from 1.4?

Nope, just 1.4 branch.

[timothysc]

Just out of curiosity, does anything break if we just updated the godep ref? It seems like we're applying a wand here.

Also, a +1 on the etcd3 changes those are fixes to issues we had upstream.

[smarterclayton]

I did a patch rebase but it was failing upstream e2e and is too risky right now. We have to rebase 90 patches upstream right now in order to pick a new bump (which if this had been started two weeks ago would have been the right choice).

[timothysc]

/cc @mffiedler

[derekwaynecarr]

This needs to pull in kubernetes/kubernetes#33806 to fix #11199

[soltysh]

@derekwaynecarr I've cherry picked that, but it was a bit tricky, mind having a look if that's ok?
@smarterclayton b0c9f3b and 9607748 are dropped (1st - manually reverted, 2nd - reverted during upgrading cAdvisor)
@bparees I had to revert s2i bump @PI-Victor did, just fyi since it was pointing to non-existent commit

[bparees]

@soltysh did you talk to @PI-Victor about that? those changes are important.

[soltysh]

@bparees yes, if there's a requirement to have those changes in, we need to create a branch and have a official commit that is reachable.

[soltysh]

@derekwaynecarr I had to split the cadvisor cherry-pick into two, one is the bump and the other is that cherry-pick itself. (the commit checker was complaining otherwise).

[ncdc]

I thought we stated that it was rather late to be pulling in 34000?

[soltysh]

I thought we stated that it was rather late to be pulling in 34000

Haven't heard about it, but if you say so it can be easily dropped.

[ncdc]

@soltysh see #10719 (comment)

[soltysh]

@ncdc @mfojtik pointed me to the comment before about having those in, I'll blame him ;) but sure I can remove it.

[soltysh]

I dropped 34000, updated client-go to latest 1.4 and updated generated clientset. Hopefully that should be good.

[derekwaynecarr]

@soltysh -- the cadvisor bump looks good to me. @sjenning fyi this would fix your issue if you want to verify against this PR.

[sjenning]

@derekwaynecarr on it

[sjenning]

@derekwaynecarr I confirmed this does fix #11199

[soltysh]

Rebased against latest master and fixed the errors that showed up in tests.

[openshift-bot]

Evaluated for origin test up to f7027fd

[openshift-bot]

continuous-integration/openshift-jenkins/test FAILURE (https://ci.openshift.redhat.com/jenkins/job/test_pr_origin/10898/) (Base Commit: 63642d6)

[mwringe]

Is this being tracking in a bugzilla for OCP anywhere?

[smarterclayton]

Tests are panicking

[ncdc]

I'll start looking into the panics.

[ncdc]

The panic is because our vendored copies of k8s.io/kubernetes/pkg/client/restclient.Config and k8s.io/client-go/1.4/rest.Config don't have the same fields. They're identical except the former adds 1: Timeout. And vendor/k8s.io/kubernetes/test/e2e/framework/framework.go#getClientRepoConfig expects the structs to match.

[timothysc]

ETA on fixes? This is blocking a bunch of stuff.

[ncdc]

Working on it

[ncdc]

This is why the former struct has Timeout and the latter doesn't: #11104

[ncdc]

Testing a fix in #11709

[smarterclayton]

Moving to #11709