Docker Autoheal

监控并重启不健康的 Docker 容器。此功能曾提议在增加 `HEALTHCHECK` 后包含进来，但最终未能实现。此容器是一个临时替代品，直到有原生支持 `--exit-on-unhealthy`https://github.com/docker/docker/pull/22719。

支持的标签和 Dockerfile 链接

如何使用

UNIX 套接字透传

docker run -d \
    --name autoheal \
    --restart=always \
    -e AUTOHEAL_CONTAINER_LABEL=all \
    -v /var/run/docker.sock:/var/run/docker.sock \
    willfarrell/autoheal

TCP socket

docker run -d \
    --name autoheal \
    --restart=always \
    -e AUTOHEAL_CONTAINER_LABEL=all \
    -e DOCKER_SOCK=tcp://HOST:PORT \
    -v /path/to/certs/:/certs/:ro \
    willfarrell/autoheal

a) 在你的容器上应用标签 `autoheal=true` 以使其被监控。

b) 设置环境变量 `AUTOHEAL_CONTAINER_LABEL=all` 以监控所有运行中的容器。

c) 设置环境变量 `AUTOHEAL_CONTAINER_LABEL` 为具有值 `true` 的现有标签名称。

注意：您必须首先将 HEALTHCHECK 应用于您的 Docker 镜像。详情请参阅 https://docs.docker.com/engine/reference/builder/#healthcheck。有关如何使用 mTLS 配置 TCP，请参阅 https://docs.docker.com/engine/security/https/。

证书和密钥需要以下名称：

ca.pem
client-cert.pem
client-key.pem

更改时区

如果需要时区与本地机器一致，可以将 /etc/localtime 映射到容器中。

docker run ... -v /etc/localtime:/etc/localtime:ro

ENV 默认设置

AUTOHEAL_CONTAINER_LABEL=autoheal
AUTOHEAL_INTERVAL=5   # check every 5 seconds
AUTOHEAL_START_PERIOD=0   # wait 0 seconds before first health check
AUTOHEAL_DEFAULT_STOP_TIMEOUT=10   # Docker waits max 10 seconds (the Docker default) for a container to stop before killing during restarts (container overridable via label, see below)
DOCKER_SOCK=/var/run/docker.sock   # Unix socket for curl requests to Docker API
CURL_TIMEOUT=30     # --max-time seconds for curl requests to Docker API
WEBHOOK_URL=""    # post message to the webhook if a container was restarted (or restart failed)

可选容器标签

autoheal.stop.timeout=20        # Per containers override for stop timeout seconds during restart

测试

docker build -t autoheal .

docker run -d \
    -e AUTOHEAL_CONTAINER_LABEL=all \
    -v /var/run/docker.sock:/var/run/docker.sock \
    autoheal

3.2 KiB Raw Blame History Unescape Escape