[英]Is HA really useful if ActiveMQ Artemis cluster is a Kubernetes StatefulSet?
If for any reason a live-node (pod) goes down Kubernetes will create another one in place of it.如果由于某种原因一个活动节点(pod)出现故障,Kubernetes 将创建另一个节点来代替它。 This would definitely cause intermittent issues at the clients' end, but the client should start working once live-node (pod) is back up.
这肯定会导致客户端出现间歇性问题,但是一旦实时节点(pod)备份,客户端应该开始工作。 On the other hand, if HA was in place I am unsure how quick the movement of clients (connections/subscriptions/sessions) from live-node to backup-node and back to live-node (ie fail-over and fail-back) would work.
另一方面,如果 HA 到位,我不确定客户端(连接/订阅/会话)从活动节点到备份节点再回到活动节点(即故障转移和故障恢复)的移动速度有多快会工作。
I think only if all the backup-nodes are in another data-center they would be useful for disaster recovery.我认为只有当所有备份节点都在另一个数据中心时,它们才会对灾难恢复有用。 However, I am unable to think of a scenario where the backup-nodes would be useful if they exist in the same data-center as the live-nodes.
但是,我无法想到如果备份节点与活动节点存在于同一个数据中心中,它们会很有用的场景。 I may be missing something, please advise on an ideal approach here.
我可能遗漏了一些东西,请在这里提供理想的方法建议。
If Kubernetes is managing a broker and will restart it if it fails then configuring brokers as active/passive for HA fail-over is not really useful.如果 Kubernetes 正在管理一个代理并且在它失败时将重新启动它,那么将代理配置为主动/被动以进行 HA 故障转移并没有真正的用处。 This, of course, assumes that all the relevant persistent data (eg the
data
directory) is safe from loss (eg on another pod).当然,这假设所有相关的持久数据(例如
data
目录)都不会丢失(例如在另一个 pod 上)。
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.