Dennis-Mircea commented on code in PR #1070:
URL:
https://github.com/apache/flink-kubernetes-operator/pull/1070#discussion_r3028515541
##########
docs/content/docs/custom-resource/autoscaler.md:
##########
@@ -158,6 +158,15 @@ In general, it's not recommended to set target utilization
close to 100% as perf
In addition to the utilization target we can set a utilization boundary, that
serves as extra buffer to avoid immediate scaling on load fluctuations.
Setting `job.autoscaler.target.utilization.boundary: "0.2"` means that we
allow 20% deviation from the target utilization before triggering a scaling
action.
+For example, with a target utilization of `0.6` and a boundary of `0.2`,
scale-up is triggered when utilization exceeds `0.8` (`0.6` + `0.2`) and
scale-down when it drops below `0.4` (`0.6` - `0.2`).
+
+For more fine-grained control, the `job.autoscaler.utilization.max` and
`job.autoscaler.utilization.min` options can be used to set asymmetric upper
and lower utilization thresholds independently.
+When set, these options take precedence over the values derived from the
boundary. When only one of them is set, the other is still derived from the
boundary.
+If neither is set, the boundary is used to compute both thresholds symmetric.
+
+{{< hint warning >}}
+The `job.autoscaler.target.utilization.boundary` option is **deprecated**.
Users are encouraged to use `job.autoscaler.utilization.max` and
`job.autoscaler.utilization.min` instead, which provide more explicit and
flexible control over the scaling thresholds. The boundary option will continue
to work as a fallback when neither `max` nor `min` is set.
+{{< /hint >}}
Review Comment:
Done! Please let me know what you think!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
