bhouse-nexthop commented on PR #14111: URL: https://github.com/apache/cloudstack/pull/14111#issuecomment-5683404181
> @bhouse-nexthop nice work. There is some overlap with #14043, which adds a `both` value for `drs.metric` that scores a migration on the worse of its cpu and memory imbalance, so a cluster only counts as balanced when both are. > > The `weighted` algorithm here goes further than that. If the maintainers prefer this approach, I am happy to drop the `both` metric from #14043 and keep only the event-driven part (running DRS on VM power-state events), which does not overlap with this PR. > > Both PRs change `ClusterDrsServiceImpl` and `ClusterDrsAlgorithm`, so it would be good to agree on an order to avoid conflicts. Yeah, this reused the fairly complex algorithm I'm introducing in #14109 because of some crazy stuff we do in our clusters (mostly over-provisioning combined with the fact that we spawn thousands of GHA runners per hour) so were hitting a combination of misbalanced clusters and thundering herd issues. At least in our environment, its unlikely this DRS change would actually help us, but its a natural consumer of that same algorithm so its worthwhile. This PR is stacked on top of both #14108 and #14109 so I have a feeling it'll be a while before it can be merged which is why its in Draft mode. -- 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]
