-
Notifications
You must be signed in to change notification settings - Fork 1.2k
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
Unique name tag per executor causes unbounded cardinality metrics #2182
Comments
hey @chringwer I'm sorry this fell through the cracks. one big issue with your idea is that it is not that simple, because the metrics decorator is installed through a public API based on |
there are other caveats that are potential blockers:
|
Is there any solution or workaround for this? |
@lotusnow In upcoming 3.5.0 (the next major version scheduled around November this year), we're introducing a dedicated See #3109, which will be made available in the soon-to-be-published |
The way executor metrics are tagged using an ever increasing integer per executor seems to cause problems with elastic schedulers. At least when using prometheus as a monitoring backend we were facing OOMs due to the high cardinality of time series produced by assigning unique names to each new executor which is spawned when the "elastic" scheduler scales out.
reactor-core/reactor-core/src/main/java/reactor/core/scheduler/SchedulerMetricDecorator.java
Lines 65 to 68 in 638de1f
Maybe this could be solved by assigning the executorId based on its "slot" in the scheduler so that the cardinality is bound by the scheduler's maxSize value?
The text was updated successfully, but these errors were encountered: