DirectXMan12
commented
6 years ago Heapster is now deprecated, so no new features will be accepted. Please see https://github.com/kubernetes/heapster/blob/master/docs/deprecation.md for more information.
Closed bmoyles0117 closed 6 years ago
DirectXMan12
commented
6 years ago Heapster is now deprecated, so no new features will be accepted. Please see https://github.com/kubernetes/heapster/blob/master/docs/deprecation.md for more information.
bmoyles0117
commented
6 years ago @DirectXMan12 this change is a stability fix for the Stackdriver sink, when we run Heapster in any environment that is not GCE, Heapster crashes. Having this change allows us to mitigate this crash for our users when the GCE Metadata Server is unavailable. Is there any way I can get your help in approving this change, and validating the case? I really appreciate any input you have on this, and any alternatives you have to offer to help us release this critical stability fix to our users.
bmoyles0117
commented
6 years ago Just clarifying a bit, by "our users", I'm totally being a bit selfish when I say that means "Stackdriver Users", or "Users that have enabled the Stackdriver sink", I didn't mean to conflate that this somehow affects heapster users not using this sink.
kawych
commented
6 years ago /ok-to-test LGTM once comments from Marian are resolved
k8s-ci-robot
commented
6 years ago [APPROVALNOTIFIER] This PR is APPROVED
This pull-request has been approved by: bmoyles0117, loburm
The full list of commands accepted by this bot can be found here.
The pull request process is described here
What this PR does When Heapster is configured to use Stackdriver as its only sink, and the Kubernetes cluster is not running inside of GCP, Heapster repeatedly crashes, and no metrics are sent to Stackdriver, rendering metrics ingestion useless for Stackdriver monitoring customers with configurations like this.
This PR allows users to manually configure the Stackdriver sink to restore metric ingestion and stop the crash loop.