Use KubeletPort reported in NodeStatus instead of cluster-wide master config #12919

gmarek · 2015-08-19T13:42:41Z

Fixes to #12912. Depends on #12655, currently includes #12913

cc @dchen1107 @yujuhong @wojtek-t @fgrzadkowski @bgrant0607 @davidopp

k8s-bot · 2015-08-19T14:10:40Z

GCE e2e build/test failed for commit dc13a043f66d5a880cf23e415861b61bf569c56d.

k8s-bot · 2015-08-19T14:12:27Z

GCE e2e build/test failed for commit 9ec22afab43f3fc712419f132e7c35e5f1685d68.

roberthbailey · 2015-08-19T18:47:31Z

pkg/api/types.go

What still uses the read-only port? When I created it (for heapster), @vishh told me that heapster would move to use the secured port and that we could quickly deprecated the RO port. I know that monit was using it as well -- does supervisord need it or can it make an https connection?

(If it wasn't clear, I'd really hate for this to become part of the stable v1 API so that we can't remove it)

Why is the var named Porty instead of Port?

I don't see the ReadOnlyPort used anywhere else in this CL (only the KubeletPort). Nor do I see the CadvisorProxy port used. Why are they being added to the node info?

Porty is a typo. The idea behind this change is to make Node object everything's that needed to interact with 'things' (currently Kubelet/CAdvisor, I may add KubeProxy there as well) running on the Node. Adding RO and CAdvisor was pretty straightforward, so I did this. KubeProxy is a bit more tricky, so I left it for later. Those informations are generally for external consumption.

Opened #12968

gmarek · 2015-08-20T05:34:44Z

Fixes #9325

k8s-bot · 2015-08-20T12:14:45Z

GCE e2e build/test failed for commit 3c18a206668edccfab43f10caf09de72c65c6a1c.

gmarek · 2015-08-20T12:26:47Z

port-forward failure seems related.

k8s-bot · 2015-08-20T12:28:57Z

GCE e2e build/test failed for commit 0e2be56aebf5b0291b6b256c128575581450f724.

k8s-bot · 2015-08-20T13:12:48Z

GCE e2e build/test failed for commit 559c4bbf018e7922ba71db854bd8fbb9241ccf24.

k8s-bot · 2015-08-20T15:09:56Z

GCE e2e build/test failed for commit fe2d46add69aaa80d09bb0af3625032bfdf16042.

k8s-bot · 2015-08-20T15:16:09Z

GCE e2e build/test failed for commit f7b04a9a9c1da7b284927518371b96176256ceea.

bgrant0607 · 2015-08-20T15:54:26Z

api/swagger-spec/v1.json

This is stale.

Yes, but this PR is built on top of #12655, so you needn't worry about the first commit.

gmarek · 2015-08-21T10:03:46Z

I run an upgrade test:

I created a cluster from the head
I upgraded the master to the new version (from this PR): everything kept running correctly
I upgraded the node to the new version (from this PR) and changed the kubelet port to 10260: everything kept running correctly ('old' Kubelet was serving from port 10250, 'new' one from 10260, daemonPorts were filled only in the 'new' Kubelets Node object)

k8s-bot · 2015-10-21T10:52:56Z

GCE e2e test build/test passed for commit f3feff51384e573c03e07a5aa5b5ab35a5d11ba1.

k8s-bot · 2015-10-21T10:53:03Z

GCE e2e test build/test passed for commit 26b2f40331f480f8880b3bae146852e6a36e2061.

k8s-bot · 2015-10-21T11:01:25Z

GCE e2e test build/test passed for commit 90fb7ed433efa78aab45102d363aaecbe27cb3e1.

k8s-bot · 2015-10-21T13:18:07Z

GCE e2e test build/test passed for commit 4e0dea91bbea692e5f7b37c410e122e2b34462db.

…r config.

k8s-bot · 2015-10-21T15:46:03Z

GCE e2e test build/test passed for commit 60404a5.

lavalamp · 2015-10-21T23:26:22Z

OK, looks like you got rid of the imports I didn't like, so LGTM

k8s-github-robot · 2015-10-21T23:33:02Z

Continuous integration appears to have missed, closing and re-opening to trigger it

k8s-bot · 2015-10-22T00:00:38Z

GCE e2e test build/test passed for commit 60404a5.

k8s-github-robot · 2015-10-22T18:18:47Z

@k8s-bot test this [submit-queue is verifying that this PR is safe to merge]

k8s-bot · 2015-10-22T18:48:18Z

GCE e2e build/test failed for commit 60404a5.

wojtek-t · 2015-10-22T18:50:59Z

Jenkins failure is flake:
"Kubectl client Guestbook application should create and stop a working application [Conformance]"

@k8s-bot test this please

k8s-bot · 2015-10-22T19:17:00Z

GCE e2e test build/test passed for commit 60404a5.

k8s-github-robot · 2015-10-23T02:22:53Z

@k8s-bot test this [submit-queue is verifying that this PR is safe to merge]

k8s-bot · 2015-10-23T02:49:08Z

GCE e2e test build/test passed for commit 60404a5.

k8s-github-robot · 2015-10-23T02:53:39Z

Automatic merge from submit-queue

Auto commit by PR queue bot

saad-ali · 2015-10-23T04:25:31Z

I suspect this may have broken the build:

20:00:12 Flag shorthand -c has been deprecated, please use --client instead.
20:00:16 +++ [1022 20:00:16] Staging release artifacts to /jenkins-master-data/jobs/kubernetes-build/workspace/_output/gcs-stage
20:00:19 tar: node.yaml: Cannot stat: No such file or directory
20:00:19 tar: Exiting with failure status due to previous errors
20:00:26 +++ [1022 20:00:26] Staging release artifacts to /jenkins-master-data/jobs/kubernetes-build/workspace/_output/gcs-stage
20:00:28 tar: node.yaml: Cannot stat: No such file or directory
20:00:28 tar: Exiting with failure status due to previous errors
20:00:35 +++ [1022 20:00:35] Staging release artifacts to /jenkins-master-data/jobs/kubernetes-build/workspace/_output/gcs-stage
20:00:36 tar: node.yaml: Cannot stat: No such file or directory
20:00:36 tar: Exiting with failure status due to previous errors

I'm going to revert it.

saad-ali · 2015-10-23T04:39:40Z

Verified that this broke the build. PR has been reverted.
Please see these builds for details:

Jenkins kubernetes-build/5708
Jenkins kubernetes-build/5709

liggitt · 2015-10-23T05:01:41Z

pkg/registry/node/etcd/etcd.go

sorry for commenting late, only saw this PR when it merged... I think the ConnectionInfoGetter interface that is already responsible for returning the kubelet port should be updated to provide this info...

The problem is that we actually need to read data from etcd to get the Port for a given Kubelet, as it's no longer 'static' (in a sense that each Kubelet may be able to choose it's own port to run on), while we have a single 'KubeletClient' in master. We'll need to remove Port from it eventually, but it needs to stay there for backward compatibility.

gmarek added team/master sig/node Categorizes an issue or PR as relevant to SIG Node. labels Aug 19, 2015

googlebot added the cla: yes label Aug 19, 2015

gmarek force-pushed the use_api_ports branch from dc13a04 to 9ec22af Compare August 19, 2015 13:45

gmarek changed the title ~~Use KubeletPort reporeted in NodeStatus instead of cluster-wide master config.~~ WIP: Use KubeletPort reporeted in NodeStatus instead of cluster-wide master config. Aug 19, 2015

gmarek mentioned this pull request Aug 19, 2015

Allow Kubelets to run on different ports. #12912

Closed

roberthbailey reviewed Aug 19, 2015
View reviewed changes

gmarek mentioned this pull request Aug 20, 2015

Add information about ports opened by Kubelet to API #12655

Merged

gmarek force-pushed the use_api_ports branch 2 times, most recently from 3c18a20 to 0e2be56 Compare August 20, 2015 12:05

gmarek force-pushed the use_api_ports branch from 0e2be56 to 559c4bb Compare August 20, 2015 12:52

gmarek force-pushed the use_api_ports branch from fe2d46a to f7b04a9 Compare August 20, 2015 14:36

bgrant0607 reviewed Aug 20, 2015
View reviewed changes

k8s-github-robot assigned zmerlynn Aug 21, 2015

gmarek force-pushed the use_api_ports branch 2 times, most recently from 985a6bd to 059ca2d Compare August 21, 2015 10:10

gmarek force-pushed the use_api_ports branch from 90fb7ed to 4e0dea9 Compare October 21, 2015 12:51

gmarek added 2 commits October 21, 2015 17:19

Use KubeletPort reporeted in NodeStatus instead of cluster-wide maste…

6d6cd8e

…r config.

Make Node.yaml use KUBELET_PORT instead of hardcoded 10250

60404a5

gmarek force-pushed the use_api_ports branch from 4e0dea9 to 60404a5 Compare October 21, 2015 15:19

lavalamp added the lgtm "Looks good to me", indicates that a PR is ready to be merged. label Oct 21, 2015

k8s-github-robot closed this Oct 21, 2015

k8s-github-robot reopened this Oct 21, 2015

k8s-github-robot pushed a commit that referenced this pull request Oct 23, 2015

Merge pull request #12919 from gmarek/use_api_ports

a6b8e11

Auto commit by PR queue bot

k8s-github-robot merged commit a6b8e11 into kubernetes:master Oct 23, 2015

saad-ali mentioned this pull request Oct 23, 2015

Revert "Use KubeletPort reported in NodeStatus instead of cluster-wid… #16144

Merged

liggitt reviewed Oct 23, 2015
View reviewed changes

Use KubeletPort reported in NodeStatus instead of cluster-wide master config #12919

Use KubeletPort reported in NodeStatus instead of cluster-wide master config #12919

Uh oh!

Conversation

gmarek commented Aug 19, 2015

Uh oh!

k8s-bot commented Aug 19, 2015

Uh oh!

k8s-bot commented Aug 19, 2015

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

gmarek commented Aug 20, 2015

Uh oh!

k8s-bot commented Aug 20, 2015

Uh oh!

gmarek commented Aug 20, 2015

Uh oh!

k8s-bot commented Aug 20, 2015

Uh oh!

k8s-bot commented Aug 20, 2015

Uh oh!

k8s-bot commented Aug 20, 2015

Uh oh!

k8s-bot commented Aug 20, 2015

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

gmarek commented Aug 21, 2015

Uh oh!

k8s-bot commented Oct 21, 2015

Uh oh!

k8s-bot commented Oct 21, 2015

Uh oh!

k8s-bot commented Oct 21, 2015

Uh oh!

k8s-bot commented Oct 21, 2015

Uh oh!

k8s-bot commented Oct 21, 2015

Uh oh!

lavalamp commented Oct 21, 2015

Uh oh!

k8s-github-robot commented Oct 21, 2015

Uh oh!

k8s-bot commented Oct 22, 2015

Uh oh!

k8s-github-robot commented Oct 22, 2015

Uh oh!

k8s-bot commented Oct 22, 2015

Uh oh!

wojtek-t commented Oct 22, 2015

Uh oh!

k8s-bot commented Oct 22, 2015

Uh oh!

k8s-github-robot commented Oct 23, 2015

Uh oh!

k8s-bot commented Oct 23, 2015

Uh oh!

k8s-github-robot commented Oct 23, 2015

Uh oh!

saad-ali commented Oct 23, 2015

Uh oh!

saad-ali commented Oct 23, 2015

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!