Finalize Kube items 3.2

[liggitt]

Placeholder to track upstream pulls that need to be included in 3.2.
Link to upstream pulls with a comment.
After rebases are completed, we'll look at which ones need to be cherry-picked and track them here.

Upstream release branch cherry pick candidates:
https://github.com/kubernetes/kubernetes/pulls?utf8=%E2%9C%93&q=is%3Apr+label%3Acherrypick-candidate+milestone%3Av1.2+

Other items:

• Carry for SCC describer: Add SCC Describer #7127 (fork pull: add scc describer kubernetes#5)
• Track the resolution of Master pods in a permanent "Terminating" status after restarting multiple times in a row kubernetes/kubernetes#22025 Reconcile graceful termination kubernetes/kubernetes#20497
• Doc upgrading a node involves draining or restarting containers (because of cgroup changes, and because of Upgrading a node from kubelet 1.1.3 to 1.2.0 results in containers getting destroyed and recreated kubernetes/kubernetes#23104)
• Fix for Use OomScoreAdj in kubelet for newer docker api kubernetes/kubernetes#21741 (origin PR UPSTREAM: 23894: OOM errors when processes exit rapidly #8412)

In master upstream:

• support CIDRs in NO_PROXY kubernetes/kubernetes#23003 (Origin PR UPSTREAM: 23003: support CIDRs in NO_PROXY #8016) fixes How to avoid proxies interferring with /api/v1/namespaces/{namespace}/pods/{name}/proxy/ #7694
• kubectl: print errors that wont be reloaded in the editor kubernetes/kubernetes#22634 (Origin PR UPSTREAM: 22634: kubectl: print errors that wont be reloaded in the editor #7829) fixes Edit with an admission failure drops user out of edit without printing error #7811
• Origin PR UPSTREAM: 23078: Check claimRef UID when processing a recycled PV #8100 (upstream PR Check claimRef UID when processing a recycled PV kubernetes/kubernetes#23078 and Check claimRef UID when processing a recycled PV, take 2 kubernetes/kubernetes#23548)

Proposed for v1.2.3:

• Honor starting resourceVersion in watch cache kubernetes/kubernetes#24208 (origin PR UPSTREAM: 24208: Honor starting resourceVersion in watch cache #8514)
• Fixed mounting with containerized kubelet kubernetes/kubernetes#23435 (origin PR UPSTREAM: 23435: Fixed mounting with containerized kubelet #8501)

In v1.2.2:

• Automated cherry pick of #21140 #23769 upstream release 1.2 kubernetes/kubernetes#23943 (fix volume GetCloudProvider, origin PR Upstream 1.2.2 picks #8492)
• Automated cherry pick of #23793 kubernetes/kubernetes#23935 (make configmap volume readable, origin PR UPSTREAM: 23793: Make ConfigMap volume readable as non-root #8411)
• Automated cherry pick of #23929 #23463 kubernetes/kubernetes#23934 (fix daemonset pod shouldRun, add events for daemonset failures, origin PR Upstream 1.2.2 picks #8492)
• A pod never terminated if a container image registry was unavailable kubernetes/kubernetes#23746 (Origin PR UPSTREAM: 23746: A pod never terminated if a container image registry was unavailable #8378)

In v1.2.1:

• Automated cherry pick of #23530 #23467 kubernetes/kubernetes#23586 (validate daemonsets don't have empty selectors, Origin PR Upstream picks #8308)
• Use versioned object when computing patch kubernetes/kubernetes#23145 (Origin PR UPSTREAM: 23145: use versioned object as schema when computing patch on conflict retry #8098) fixes issue Flake in cmd test #6435
• Fix controller-manager race condition issue which cause endpoints flush during restart kubernetes/kubernetes#23035 (race causing endpoints flush during restart, origin PR Upstream cherry picks #8277)
• Validate minimum CPU limits to be >= 10m kubernetes/kubernetes#23143 (cpu limit >= 10m, origin PR Upstream cherry picks #8277)
• kubelet: send all recevied pods in one update kubernetes/kubernetes#23141 (kubelet update all pods at once, origin PR Upstream cherry picks #8277)
• Add a rate limiter to the GCE cloudprovider kubernetes/kubernetes#23019 (rate limit GCE cloudprovider, origin PR Upstream cherry picks #8277)?
• Fix hairpin mode kubernetes/kubernetes#23325 (fix hairpin mode, origin PR Upstream cherry picks #8277)
• Automated cherry pick of #23446 kubernetes/kubernetes#23457 (remove usage metric on hostpath volume, origin PR Upstream cherry picks #8277)
• Automated cherry pick of #23223 kubernetes/kubernetes#23456 (don't sync daemonsets with no selector, origin PR Upstream cherry picks #8277)

In v1.2.0 (Origin PR #8072):

• Adding FullyLabeledReplicas field to replicaset status kubernetes/kubernetes#22847 for API change
• AWS EBS: Remove the attached volumes cache kubernetes/kubernetes#22791
• Allow up to 39 PVs attached to AWS node kubernetes/kubernetes#22942 helps with bz#1316095
• bump cadvisor to v0.22.1 kubernetes/kubernetes#22803 (contains Cadvisor hangs on startup google/cadvisor#1152 cAdvisor hangs for 20s on startup when not on GCE/AWS. This means that the kubelet also hangs for 20s on startup.)
• Updating cadvisor deps to v0.22.2 kubernetes/kubernetes#23020 (contains Avoid collecting network stats for non root cgroups in raw handler. google/cadvisor#1158 to reduce high cpu load from collecting network stats), part of a round of 1.2 cherrypicks here Release 1.2 cherrypick batch kubernetes/kubernetes#23024
• Add user-agent default for discovery client: add user-agent defaulting for discovery kubernetes/kubernetes#22746 (already picked)

In the v1.2-beta.1 rebase (Origin PR #7975):

• cache resizing add a param(--watch-cache-sizes) of kube-apiserver to set watch cache capabilities kubernetes/kubernetes#19783
• kubecontainer.Cache Turn on kubecontainer.Cache in kubelet kubernetes/kubernetes#19850 , there a number of fixes that are associated with this PR set so I'm hoping next rebase is after this merge.
• Scheduler predicate to cap AWS EBS vols Scheduler predicate to cap AWS EBS volumes mounted on nodes kubernetes/kubernetes#19880
• Updating cAdvisor deps to v0.20.4 kubernetes/kubernetes#19739 Cadvisor update, fixes goroutine leak, jitter on housekeeping, and modifications necessary for metrics api. @ncdc do you have coverage of the metrics api changes?
• There will probably need to be another cadvisor bump too. It's currently part of Ensure kubelet pid is not moved to system container kubernetes/kubernetes#19354. It has fixes for https://bugzilla.redhat.com/show_bug.cgi?id=1300570.
• fix the namespaceScoped of cachers kubernetes/kubernetes#19970 Fixes namescoping in a bunch of cachers.
• Use Openstack os-ext-ips extension to qualify IP address types Use Openstack os-ext-ips extension to qualify IP address types kubernetes/kubernetes#18829 For bug https://bugzilla.redhat.com/show_bug.cgi?id=1303085
• Follow-up from 19354: Do not move pid 1 to /system container kubernetes/kubernetes#20395
• E2E libraryzation - E2E libraryzation (minimal patch) kubernetes/kubernetes#19624 for [DO-NOT-MERGE] Add optional E2E test artifact to binary artifacts for openshift #6541 /cc @jayunit100
• AWS EBS Volume Limit Scheduler Predicate - Scheduler predicate to cap AWS EBS volumes mounted on nodes kubernetes/kubernetes#19880
• Quantity can lose formatting: Add or subtract from zero should not lose format kubernetes/kubernetes#20505
• Make container runtime's cgroup configurable Make container runtime's cgroup configurable. kubernetes/kubernetes#20687 Fixes bootstrapping of the kubelet on systemd to no longer move docker daemon.
• namespace deletion improvements: Further improvements to namespace deletion kubernetes/kubernetes#20076 (for bulk project deletion causes long controller response curve #7166)
• nice to have (allows running oc in pods without any manual setup): Provide current namespace to InClusterConfig kubernetes/kubernetes#21095
• No-op changes: Don't store no-op updates in etcd. kubernetes/kubernetes#20897
• Don't rate limit watches: Never rate limit watches kubernetes/kubernetes#21104 Still pending, but the root of several issues...
• Support the subresource of service proxy kubernetes/kubernetes#17493
• Support subresource of node proxy kubernetes/kubernetes#18010
• nice to have: Jittering periods of some kubelet's sync loops: kubernetes/kubernetes#20726
• Can't replace a generic resource that is cluster scoped kubernetes/kubernetes#21179
• only load kubeconfig files one time kubernetes/kubernetes#21266 needed for new-app spews out way too much cmd gorp #6804
• make kubectl logs work for replication controllers kubernetes/kubernetes#21335 is needed for Can't get logs from replication controller #7329
• update query encode/decode to support custom marshalled types kubernetes/kubernetes#21324 Fix sinceTime pod log options kubernetes/kubernetes#21398 is needed for https://bugzilla.redhat.com/show_bug.cgi?id=1286116
• Tolerate individual NotFound errors in DeleteCollection kubernetes/kubernetes#21340 for namespace deletion races
• kubectl: scale down based on ready during rolling updates kubernetes/kubernetes#21273 for fixing deployments against unready apps EDIT: Already merged in Origin (UPSTREAM: 21273: kubectl: scale down based on ready during rolling up… #7307)
• change kubelet defaults: Change default --max-pods in kubelet to 110 kubernetes/kubernetes#21361
• Unqualified host:port pairs are valid Docker auth fields kubernetes/kubernetes#21286 handles some types of Docker cfg secrets from being selected by the keyring.
• fix limitranger to handle latent caches without live lookups every time kubernetes/kubernetes#21470 needed for LimitRange admission plugin race #7401
• Fix type for DaemonSet informer in node controller: Fix type for DaemonSet informer kubernetes/kubernetes#21496
• Add a liveness and readiness describer to pods kubernetes/kubernetes#21341 for enhancing the dc describer EDIT: Already merged in Origin (oc: enhance deploymentconfig description #7354)
• Unqualified host:port pairs are valid Docker auth fields kubernetes/kubernetes#21286
• Revert 20202. Use other measures to prevent race in test-cmd.sh kubernetes/kubernetes#21175 <Revert 20202. Use other measures to prevent race in test-cmd.sh kubernetes/kubernetes#21175 (comment)> because it's API incompatible with 3.1
• kubectl: add container ports in pod description kubernetes/kubernetes#21671 for the pod describer EDIT: Already merged in Origin (UPSTREAM: 21671: kubectl: add container ports in pod description #7516)
• Fix duplicate updates on pod delete: kubelet: fix duplicated status updates at pod cleanup kubernetes/kubernetes#21732 This can flood channels in kubelet when doing bulk pod deletion. EDIT(ncdc): THIS WAS REVERTED! EDIT2(ncdc): here's the replacement: kubelet: fix duplicated status updates at pod cleanup kubernetes/kubernetes#21959
• Fix LimitRanger LRU cach: Fix flake in LimitRange e2e kubernetes/kubernetes#21788
• Perf improvement for quota replenishment: Quota replenishment should only add quotas that track the resource kubernetes/kubernetes#22445
• Force pod status update in kubelet if deletionTimestamp is set: kubelet: force status update when the deletion timestamp is set kubernetes/kubernetes#22483
• Resolve fenceposts for deployments together kubernetes/kubernetes#22516 (only the first commit is enough - Origin PR: UPSTREAM: 22516: kubectl: set maxUnavailable to 1 if both fenceposts resolve to zero #7796)
• kubectl: bring the rolling updater on par with the deployments kubernetes/kubernetes#22526 (Origin PR: UPSTREAM: 22526: kubectl: bring the rolling updater on par with the deployments #7828)
• Quota replenishment perf: Don't sync all objects for replenishment as often as ResourceQuota. kubernetes/kubernetes#22628
• fix for bz#1311868: "No tags for dynamic provisioned volume in AWS": Add PV.Name into names of generated GCE/AWS/OpenStack volumes. kubernetes/kubernetes#20900
• fix for bz#1311312 "AWS and GCE Dynamic provisioners do not work": Hot-fix admission of volumes that are being provisioned. kubernetes/kubernetes#21738
• Keep an eye on Kubelet: Always restart container in unknown state. kubernetes/kubernetes#22607. We'll most likely want it, assuming it merges.
• Kubelet stats is wrong on systemd ignore .mount cgroups associated with container on systemd kubernetes/kubernetes#22656
• Pass ResourceVersion in Scale object back to RC before updating RC so that it can be used to check for conflicts. kubernetes/kubernetes#22769 API changes to scale subresources
• Support for both map-based and set-based selectors in extensions/v1beta1.Scale kubernetes/kubernetes#22724 API changes to scale subresources
• ReplicaSetSpec.Template shouldn't be a pointer. kubernetes/kubernetes#22758 API change to replicaset:
• Loosen LimitRange validation @aveshagarwal Fix validation of resources (cpu, memory, storage) for limit range types. kubernetes/kubernetes#22766 This is not targeted 1.2, but is just a small change to validation.
• If allow resource.version.group in kubectl kubernetes/kubernetes#22853 merges, then we'll need to pick it to get consistent client tool capability.
• prevent disallowed secret refs from leaking via the downward API kubernetes/kubernetes#22865 secrets in envvars don't obey SA limits
• Close streams with Reset instead of Close when closing streaming connections kubernetes/kubernetes#22802 ensures that spdystream goroutines are cleaned up when the exec/attach/portforward spdy connection is closed. Requires bump(github.com/docker/spdystream): 449fdfce4d962303d702fec724ef0ad181c92528 kubernetes/kubernetes#22804 to function 100% properly.
• Nice to have so we don't get log spam Ignore empty strings paths while calculting disk usage. google/cadvisor#1122

Not in 1.2:

• ThirdPartyResources is still targeted for 1.2 and is big enough to impact whether future patches apply cleanly: Add support for 3rd party objects to kubectl kubernetes/kubernetes#18835 moved to 1.3
• Adapt to Specify Pod hostname by Annotation kubernetes/kubernetes#20688 - opened Update skydns backend to use pod/endpoint hostname annotations #8179 to track
• kubelet: reading cloudinfo from cadvisor kubernetes/kubernetes#21373 (Origin PR UPSTREAM: 21373: kubelet: reading cloudinfo from cadvisor #8062, reverted in UPSTREAM: revert: d54ed4e: 21373: kubelet: reading cloudinfo from cad… #8197 because of https://bugzilla.redhat.com/show_bug.cgi?id=1319439 until cloudinfo: bugfix openstack identified as aws google/cadvisor#1171 is picked up)
• fix edit on list kubernetes/kubernetes#20918 fixes a regression in oc edit to handle lists.
• provide best effort generation for struct field types even if parent types are asymmetrical kubernetes/kubernetes#23213 (origin PR UPSTREAM: 23213: provide best effort generation even if types are asy… #8194) - needed before next rebase

[liggitt]

From #6881 (comment)

• cache resizing add a param(--watch-cache-sizes) of kube-apiserver to set watch cache capabilities kubernetes/kubernetes#19783
• kubecontainer.Cache Turn on kubecontainer.Cache in kubelet kubernetes/kubernetes#19850 , there a number of fixes that are associated with this PR set so I'm hoping next rebase is after this merge.
• Scheduler predicate to cap AWS EBS vols Scheduler predicate to cap AWS EBS volumes mounted on nodes kubernetes/kubernetes#19880

[ncdc]

[timothysc]

• Updating cAdvisor deps to v0.20.4 kubernetes/kubernetes#19739 Cadvisor update, fixes goroutine leak, jitter on housekeeping, and modifications necessary for metrics api. @ncdc do you have coverage of the metrics api changes?

[ncdc]

There will probably need to be another cadvisor bump too. It's currently part of kubernetes/kubernetes#19354. It has fixes for https://bugzilla.redhat.com/show_bug.cgi?id=1300570.

@timothysc which metrics changes are you referring to? We aren't going to need cadvisor changes for custom metrics until 3.3.

[0xmichalis]

kubernetes/kubernetes#19970

Fixes namescoping in a bunch of cachers.

[ncdc]

Use Openstack os-ext-ips extension to qualify IP address types kubernetes/kubernetes#18829

For bug https://bugzilla.redhat.com/show_bug.cgi?id=1303085

[ncdc]

Follow-up from 19354: kubernetes/kubernetes#20395

[timothysc]

E2E libraryzation - kubernetes/kubernetes#19624 for #6541 /cc @jayunit100

[DirectXMan12]

AWS EBS Volume Limit Scheduler Predicate - kubernetes/kubernetes#19880

[derekwaynecarr]

Quantity can lose formatting: kubernetes/kubernetes#20505

[pweil-]

Carry for SCC describer: #7127 (fork pull: openshift/kubernetes#5)

[timothysc]

@pweil - if there is a final rebase I believe you can drop most of the items I have listed above.

[derekwaynecarr]

Make container runtime's cgroup configurable
kubernetes/kubernetes#20687

Fixes bootstrapping of the kubelet on systemd to no longer move docker daemon.

[pweil-]

namespace deletion improvements: kubernetes/kubernetes#20076 (for #7166)

[liggitt]

nice to have (allows running oc in pods without any manual setup): kubernetes/kubernetes#21095

[smarterclayton]

No-op changes: kubernetes/kubernetes#20897

[timothysc]

Don't rate limit watches: kubernetes/kubernetes#21104

Still pending, but the root of several issues...

[jsafrane]

Fixes for BZ#1305546: #7258 #7260 Already merged

[liggitt]

subresource proxy for nodes/services:
kubernetes/kubernetes#17493
kubernetes/kubernetes#18010

[timothysc]

nice to have: kubernetes/kubernetes#20726

[smarterclayton]

kubernetes/kubernetes#21179

[deads2k]

kubernetes/kubernetes#21266 needed for #6804

[derekwaynecarr]

Kubelet stats is wrong on systemd
kubernetes/kubernetes#22656

[ncdc]

Add user-agent default for discovery client: kubernetes/kubernetes#22746

[derekwaynecarr]

In case this does not make kube 1.2, we should carry the following for cadvisor:

google/cadvisor#1152

cAdvisor hangs for 20s on startup when not on GCE/AWS.

This means that the kubelet also hangs for 20s on startup.

[liggitt]

API changes to scale subresources
kubernetes/kubernetes#22769
kubernetes/kubernetes#22724

API change to replicaset:
kubernetes/kubernetes#22758

[deads2k]

ThirdPartyResources is still targeted for 1.2 and is big enough to impact whether future patches apply cleanly: kubernetes/kubernetes#18835

[derekwaynecarr]

Loosen LimitRange validation @aveshagarwal

kubernetes/kubernetes#22766

This is not targeted 1.2, but is just a small change to validation.

[deads2k]

If kubernetes/kubernetes#22853 merges, then we'll need to pick it to get consistent client tool capability.

[deads2k]

kubernetes/kubernetes#22865 secrets in envvars don't obey SA limits

[ncdc]

kubernetes/kubernetes#22802 ensures that spdystream goroutines are cleaned up when the exec/attach/portforward spdy connection is closed. Requires kubernetes/kubernetes#22804 to function 100% properly.

[ncdc]

google/cadvisor#1158 to reduce high cpu load from collecting network stats

[liggitt]

kubernetes/kubernetes#22847 for API change

[deads2k]

kubernetes/kubernetes#21373 would help integration with manageiq and is contained. I'd like to pick it even though kube 1.2 will not have it. Testing has been discussed.

[0xmichalis]

oc edit fix: kubernetes/kubernetes#22634. related issue: #7811

[deads2k]

kubernetes/kubernetes#23003 fixes #7694

[jsafrane]

kubernetes/kubernetes#22942 helps with bz#1316095

[ncdc]

kubernetes/kubernetes#23020 (contains cadvisor 1158), part of a round of 1.2 cherrypicks here kubernetes/kubernetes#23024

[childsb]

kubernetes/kubernetes#21758

[liggitt]

anything in the kube v1.2.0 release tag is in origin as of #8072

[smarterclayton]

Potentially kubernetes/kubernetes#23035 -
controller failover can flush the endpoints list (make them zero)

On Wed, Mar 16, 2016 at 3:01 PM, bradley childs [email protected]
wrote:

kubernetes/kubernetes#21738
kubernetes/kubernetes#21738

—
You are receiving this because you commented.
Reply to this email directly or view it on GitHub
#6766 (comment)

[ghost]

#8100
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1310587

[derekwaynecarr]

cpu limits must be >= 10m

kubernetes/kubernetes#23143

[pweil-]

provide best effort generation for struct fields (required for SCC generation)

kubernetes/kubernetes#23213

pick #8194

[smarterclayton]

Need to be aware of kubernetes/kubernetes#21741 - I'm hitting this on a daily basis

[derekwaynecarr]

Pods stuck in terminating state if the registry is unavailable:

kubernetes/kubernetes#23746

pick #8378

[jsafrane]

Containerized OpenShift should be able to mount volumes

kubernetes/kubernetes#23435
https://bugzilla.redhat.com/show_bug.cgi?id=1313210

Now tracked as merged as #8501

[liggitt]

kubernetes/kubernetes#24208 for etcd watch cache event delivery

[liggitt]

^^^^ everything above this point it now in the checklist ^^^^

[liggitt]

closing, will track items individually from here on