best-practices/autoscaling/hpa-with-custom-metrics/prometheus-adapter #115
Replies: 5 comments 3 replies
-
生产中的prom压力很大,基础指标和业务指标都是塞一套prom集群中,,偶尔的聚合查询会导致prom oom;如果要用基于custom/external metrics,是否要独立一套prom呢?roc大佬有啥建议不 |
Beta Was this translation helpful? Give feedback.
-
如果不嫌运维麻烦,可以独立一套 也可以试试 victoriametrics,相比原生 prom 做了很多性能优化。 |
Beta Was this translation helpful? Give feedback.
-
能否请教下,对于长链接服务,如何自动缩容呢?目前网上都是介绍有状态服务数据卷如何处理,没有发现长链接缩容这块的处理,难道只能写一个程序监控连接数,然后调用 k8s 接口来缩容吗? |
Beta Was this translation helpful? Give feedback.
-
你好,我们有个多用户在线提交任务的业务,等待任务数大于HPA平均值时触发扩容,但是在缩容时会无差别的把运行任务的实例给缩掉,这种情况有办法不缩容正在执行任务的实例吗? |
Beta Was this translation helpful? Give feedback.
-
请教一下,部署完adapter后,获取到的资源只有job,「{"kind":"APIResourceList","apiVersion":"v1","groupVersion":"custom.metrics.k8s.io/v1beta1","resources":[{"name":"jobs.batch/httpserver_requests_qps","singularName":"","namespaced":true,"kind":"MetricValueList","verbs":["get"]}]}」并没有pods资源,导致我hpa无法抓去到相关的指标 |
Beta Was this translation helpful? Give feedback.
-
https://imroc.cc/kubernetes/best-practices/autoscaling/hpa-with-custom-metrics/prometheus-adapter
Beta Was this translation helpful? Give feedback.
All reactions