kubernetes - Docker UCP 控制器因错误而关闭：不健康的 UCP 管理器：无法访问管理器：连接：连接被拒绝

Question

我已经使用 docker UCP 设置了一个测试环境，几天后，其中一个控制器随机关闭，并在 UCP 中显示主机已关闭且集群不健康的消息。

控制器容器的日志：

{"level":"warning","msg":"Kube controller manager health check error: unable to inspect container: context deadline exceeded","time":"2018-12-18T10:03:10Z"}
{"level":"warning","msg":"Kube controller manager health check timed out","time":"2018-12-18T10:03:19Z"}
{"level":"warning","msg":"Node health error detected during _ping: Kube controller manager health check timed out","time":"2018-12-18T10:03:19Z"}
{"level":"warning","msg":"Kube controller manager health check error: unable to inspect container: context deadline exceeded","time":"2018-12-18T10:03:19Z"}
{"level":"warning","msg":"Kube controller manager health check timed out","time":"2018-12-18T10:03:56Z"}
{"level":"warning","msg":"Node health error detected during _ping: Kube controller manager health check timed out","time":"2018-12-18T10:03:56Z"}
{"level":"warning","msg":"Kube controller manager health check error: unable to inspect container: context deadline exceeded","time":"2018-12-18T10:03:56Z"}
{"level":"warning","msg":"Kube controller manager health check timed out","time":"2018-12-18T10:04:15Z"}
{"level":"warning","msg":"Node health error detected during _ping: Kube controller manager health check timed out","time":"2018-12-18T10:04:15Z"}
{"level":"warning","msg":"Kube controller manager health check error: unable to inspect container: context deadline exceeded","time":"2018-12-18T10:04:15Z"}
{"level":"warning","msg":"Kube controller manager health check timed out","time":"2018-12-18T10:04:32Z"}
{"level":"warning","msg":"Node health error detected during _ping: Kube controller manager health check timed out","time":"2018-12-18T10:04:32Z"}
{"level":"warning","msg":"Kube controller manager health check error: unable to inspect container: context deadline exceeded","time":"2018-12-18T10:04:32Z"}
{"level":"warning","msg":"Kube controller manager health check error: unable to inspect container: context deadline exceeded","time":"2018-12-18T10:05:07Z"}
{"level":"warning","msg":"Kube controller manager health check timed out","time":"2018-12-18T10:05:07Z"}
{"level":"warning","msg":"Node health error detected during _ping: Kube controller manager health check timed out","time":"2018-12-18T10:05:07Z"}
{"level":"warning","msg":"Kube controller manager health check error: unable to inspect container: context deadline exceeded","time":"2018-12-18T10:05:43Z"}
{"level":"warning","msg":"Kube controller manager health check timed out","time":"2018-12-18T10:05:43Z"}
{"level":"warning","msg":"Node health error detected during _ping: Kube controller manager health check timed out","time":"2018-12-18T10:05:43Z"}
{"level":"warning","msg":"Kube controller manager health check timed out","time":"2018-12-18T10:05:51Z"}
{"level":"warning","msg":"Node health error detected during _ping: Kube controller manager health check timed out","time":"2018-12-18T10:05:51Z"}
{"level":"warning","msg":"Kube controller manager health check error: unable to inspect container: context deadline exceeded","time":"2018-12-18T10:05:51Z"}
{"level":"warning","msg":"Kube controller manager health check error: unable to inspect container: context

可以是随机网络连接问题吗？但它应该已经自动恢复了？

标签： kubernetesdocker-ucpdocker-ee

Answer 1

在检查了主机上的 docker daemon 之后，我看到系统出现了这个问题：

https://github.com/docker/for-linux/issues/162