@@ -468,6 +468,30 @@ IPVS 为将流量均衡到后端 Pod 提供了更多选择:
468468* ` nq ` (永不排队):流量被发送到一台空闲服务器(如果有的话),而不是等待一台快速服务器;
469469 如果所有服务器都忙碌,算法将退回到 ` sed ` 行为。
470470
471+ <!--
472+ * `mh` (Maglev Hashing): Assigns incoming jobs based on
473+ [Google's Maglev hashing algorithm](https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/44824.pdf),
474+ This scheduler has two flags: `mh-fallback`, which enables fallback to a different
475+ server if the selected server is unavailable, and `mh-port`, which adds the source port number to
476+ the hash computation. When using `mh`, kube-proxy always sets the `mh-port` flag and does not
477+ enable the `mh-fallback` flag.
478+ In proxy-mode=ipvs `mh` will work as source-hashing (`sh`), but with ports.
479+
480+ These scheduling algorithms are configured through the
481+ [`ipvs.scheduler`](/docs/reference/config-api/kube-proxy-config.v1alpha1/#kubeproxy-config-k8s-io-v1alpha1-KubeProxyIPVSConfiguration)
482+ field in the kube-proxy configuration.
483+ -->
484+ * ` mh ` (Maglev Hashing):基于 [ Google 的 Maglev 哈希算法] ( https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/44824.pdf )
485+ 来分配接收的任务。此调度器有两个标志:
486+ ` mh-fallback ` 允许在选定的服务器不可用时回退到另一台服务器;
487+ ` mh-port ` 将源端口号添加到哈希计算中。
488+ 在使用 ` mh ` 时,` kube-proxy ` 始终会设置 ` mh-port ` 标志,但不会启用 ` mh-fallback ` 标志。
489+ 在代理模式为 ipvs 时,` mh ` 的工作方式与源哈希(` sh ` )类似,但会包含端口信息。
490+
491+ 这些调度算法是通过 kube-proxy 配置中的
492+ [ ipvs.scheduler] ( /zh-cn/docs/reference/config-api/kube-proxy-config.v1alpha1/#kubeproxy-config-k8s-io-v1alpha1-KubeProxyIPVSConfiguration )
493+ 字段进行配置的。
494+
471495{{< note >}}
472496<!--
473497To run kube-proxy in IPVS mode, you must make IPVS available on
0 commit comments