Reduce API calls for Azure instance metadata

[feiskyer]

What this PR does / why we need it:

Azure cloud provider gets a lot of "Too many requests" error when getting availability zones, instance types and node addresses. Hence kubelet won't be able to initialize itself sometimes.

This PR reduces such calls and alos switches to json API which is more stable.

Which issue(s) this PR fixes (optional, in fixes #<issue number>(, fixes #<issue_number>, ...) format, will close the issue(s) when PR gets merged):
Fixes Azure/acs-engine#3681

Special notes for your reviewer:

Release note:

Reduce API calls for Azure instance metadata.

cc @ritazh @khenidak @andyzhangx

[feiskyer]

/sig azure

[feiskyer]

/kind bug

[andyzhangx]

Do we have any rate limit doc about instance metadata feature @khenidak , it's a little weird for this fix: az.metadata.Object call has more rate limit than az.metadata.Text call, so we use az.metadata.Object call for this fix, I doubt in the future, we may hit such issue again...

Related calls are under here:

kubernetes/pkg/cloudprovider/providers/azure/azure_instance_metadata.go

Lines 75 to 76 in 59fdc02

	func (i *InstanceMetadata) Object(path string, obj interface{}) error {
	data, err := i.queryMetadataBytes(path, "json")

kubernetes/pkg/cloudprovider/providers/azure/azure_instance_metadata.go

Lines 84 to 85 in 59fdc02

	func (i *InstanceMetadata) Text(path string) (string, error) {
	data, err := i.queryMetadataBytes(path, "text")

[ritazh]

@feiskyer why is this causing kube-scheduler to not even start?

[khenidak]

I honestly think we should cache this call. Zone, ip, vm name etc will not change that frequently. I do second @andyzhangx comments wrt not understanding the difference between Text and Object - the number of calls is the same either way, right?

[feiskyer]

@khenidak This PR reduces the original two calls to one: isCurrentInstance() will also make a API call while it's removed in this PR. And from my testing, Object way did reduce "Too many requests" error in kubelet logs. While with Text() way, there're a few such errors on kubelet start.

@ritazh Kubelet won't initialize itself successfully, hence some operations are not working as expected.

[andyzhangx]

@feiskyer I am ok if there is no better fix for this issue since v1.12 code freeze date is approaching.
@khenidak @ritazh any comments on this?

[ritazh]

will test this first thing tomorrow morning

[andyzhangx]

@ritazh that's great, then we will wait for your comment.

[feiskyer]

@andyzhangx We may introduce caches for them in the future (which needs more work because we should cache all metadata requests), but now changing to json API does work to fix current issues.

@ritazh Thanks.

[ritazh]

confirming this fixes Azure/acs-engine#3681

[andyzhangx]

/lgtm

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: andyzhangx, feiskyer

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• pkg/cloudprovider/providers/azure/OWNERS [andyzhangx,feiskyer]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[k8s-github-robot]

/test all

Tests are more than 96 hours old. Re-running tests.

[k8s-github-robot]

/test all [submit-queue is verifying that this PR is safe to merge]

[feiskyer]

/retest

[k8s-github-robot]

Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions here.