[大数据]k8s容器驱逐之ephemeral-storage

1.容器被驱逐，首先比较直观的是查看Event

The node was low on resource: ephemeral-storage.?
Container zk was using 910316Ki, which exceeds its request of 0.

2.原因

cat /etc/docker/daemon.json
{
    "exec-opts": ["native.cgroupdriver=systemd"],
    "log-driver": "json-file",
    "log-opts": {
        "max-size": "100m",
        "max-file": "10"
    },
    "bip": "xxxxxx/24",
    "oom-score-adjust": -1000,
    "registry-mirrors": ["https://xxx.mirror.aliyuncs.com"],
    "storage-driver": "overlay2",
    "storage-opts":["overlay2.override_kernel_check=true"],
    "live-restore": true
}

k8s容器应用使用console以方便日志采集，日志尽管有清理，但是ephemeral-storage是整体概念

达到驱逐限制，则bbq

3.像zk或者Kafka中间件核心的有状态的虽然方便运维且有HA，但是存在驱逐后有一定风险。

即使单独使用打标签、打污点的专属节点，但只是缓解风险来临，因此建议优先选择非k8s环境。