WXJACKIE

Results 4 issues of WXJACKIE

Recently occurred such a problem in our project: There are some redis clusters on K8s deployed and managed by us, after the business service runs for a period of time...

will not fix

**What would you like to be added**: PUB(PodUnavailableBudget)提供额外字段,来支持1个PUB资源对多个工作负载资源(Sts/Deployment)独立生效。 **Why is this needed**: 在我们的Operator场景中,一个CR会包含很多组StatefulSet,这些StatefulSet对应需要的PUB配置完全相同。假设我有200个Sts,那么就需要创建出对应的200个PUB资源,同时创建/管理这些配置完全相同的PUB比较繁琐,相应的Operator对APIServer的请求数也会提高。可以考虑引入类似group by的语义字段

kind/feature-request

**What would you like to be added**: PodUnavailableBudget提供配置,供使用者选择是否跳过NotReady的Pod。当前Kruise的源码中,针对非Ready的Pod,会直接跳过PUB的防护。 **Why is this needed**: 大型故障场景兜底,比如出现控制面与Kubelet不兼容、或者大量Node网络分区导致全量/大量K8s Node同时进入NotReady,同时Pod也进入NotReady触发驱逐。假如我有一个3副本的Sts,PUB设置的minAvailable=1,但是由于网络分区或控制面升级兼容性问题,3个Pod所在的Node都NotReady,进而Pod NotReady,此时PUB将失效。

kind/feature-request

https://redis.io/docs/management/admin/ In this document, it is mentioned “Ensured that swap is enabled“ , but in the actual production experience and the suggestion of Redis commercial version, SWAP should be turned...