-
Notifications
You must be signed in to change notification settings - Fork 1.2k
⚠️ Drop rest client request latecy metric #1586
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: 2uasimojo The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
Hi @2uasimojo. Thanks for your PR. I'm waiting for a kubernetes-sigs member to verify that this patch is reasonable to test. If it is, they should reply with Once the patch is verified, the new status will be reflected by the I understand the commands that are listed here. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is going to break a bunch of people who alert off of this metric, is that going to be okay?
This is where it would help me to understand what the actual effect of kubernetes/kubernetes#83836 was. Wouldn't that have already broken such consumers? Also: They could restore by registering the metric themselves, like this, right? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
/ok-to-test
We actually have a deprecation policy, we announce a metric deprecation, we use our prometheus wrapper to annotate our metrics, our framework, on a version bump will auto-hide the metric (which yes breaks consumers). However, they can re-enable it for a release in order to properly migrate off of it. After that, the metric is deleted. We offer a staged way to softly push people from using a metric that we intend to delete, in this PR, it'll just abruptly break people... |
@logicalhan is that an assumption or do you personally know of a case where someone uses this? The metric includes the whole url as a label, which is obviously a cardinality problem |
I'm assuming people use recording rules to aggregate the metric into something more useable. |
FYI (We also have a way to enforce cardinality bounds): kubernetes/enhancements#1692 |
We try really, really hard not to break people upstream... |
Sure but deactivating a metric and forcing ppl to do sth to get it back is always going to break them. I am also fine with adding an explicit opt-in for this and removing it in a future release, but the only value add I see in that is for cases where ppl use this metric and don't know how to properly copy the code that is currently being removed. |
Remove the `rest_client_request_latency_seconds` metric. This metric has been deprecated since kube 1.14 and [turned off in kube 1.17](kubernetes/kubernetes#83836). Co-Authored-By: suhanime
Okay; if either
or
allows us to effectively nix this metric from taking up space, we're happy to do that. What does the former look like? /me starts reading the latter... |
like this? #1587 |
Yes but in order to unbreak in this case is to roll-back a version. In our case, you just need to enable a flag. This means you can bump the version and use it in an unbroken way, but you must now explicitly acknowledge that this metric is not one you can depend on in a future release. It is a forcing function which is softer than, you can't use this release if you can't tolerate not having this metric. |
Yeah I probably prefer the optional approach, it's softer and friendly to users. |
Okay, let's pursue this via #1587 (or something like it). Thanks! |
Remove the
rest_client_request_latency_seconds
metric.This metric has been deprecated since kube 1.14 and turned off in kube 1.17.
fixes #1423
Co-Authored-By: @suhanime