☰

使用 docker compose v2 构建失败，但旧版 docker-compose v1.29.2 和 docker build 成功

tag-icon • tag-icon docker ulimit docker-compose
使用 docker compose v2 构建失败，但旧版 docker-compose v1.29.2 和 docker build 成功

我有这个 Dockerfile

FROM alpine/git as clone
WORKDIR /keycloak
RUN git clone https://github.com/maturbet-volta/nexus3-keycloak-plugin.git
WORKDIR /cargo
RUN git clone https://github.com/sonatype-nexus-community/nexus-repository-cargo.git

FROM maven:3.5-jdk-8-alpine as build
WORKDIR /keycloak
COPY --from=clone /keycloak/nexus3-keycloak-plugin /keycloak
RUN mvn -PbuildKar package -Dmaven.test.skip=true

WORKDIR /cargo
COPY --from=clone /cargo/nexus-repository-cargo /cargo
RUN mvn -PbuildKar package
  
FROM sonatype/nexus3:latest
COPY --from=build /keycloak/target/*.kar /cargo/target/*.kar /opt/sonatype/nexus/deploy/

使用 docker build 或 docker-compose (v1.29.2) build 构建此 Dockerfile 没有任何问题

但是当使用 docker compose 版本 2.16.0 时，它在 mvn package 步骤中失败

library initialization failed - unable to allocate file descriptor table - out of memory

所以我猜这是达到了某种限制，但我不知道如何更改 docker compose build 的 ulimits

那么有人知道如何配置 docker compose v2 吗？

谢谢。

以下是 docker info 和 docker compose version 的输出

$ docker compose version
Docker Compose version v2.16.0
$ docker info
Client:
 Context:    default
 Debug Mode: false
 Plugins:
  compose: Docker Compose (Docker Inc., v2.16.0)

Server:
 Containers: 8
  Running: 3
  Paused: 0
  Stopped: 5
 Images: 251
 Server Version: 20.10.23
 Storage Driver: btrfs
  Build Version: Btrfs v6.1.3
  Library Version: 102
 Logging Driver: journald
 Cgroup Driver: systemd
 Cgroup Version: 2
 Plugins:
  Volume: local
  Network: bridge host ipvlan macvlan null overlay
  Log: awslogs fluentd gcplogs gelf journald json-file local logentries splunk syslog
 Swarm: inactive
 Runtimes: io.containerd.runc.v2 io.containerd.runtime.v1.linux runc
 Default Runtime: runc
 Init Binary: /usr/libexec/docker/docker-init
 containerd version: 
 runc version: 
 init version: 
 Security Options:
  seccomp
   Profile: default
  selinux
  cgroupns
 Kernel Version: 6.1.18-200.fc37.x86_64
 Operating System: Fedora Linux 37 (Workstation Edition)
 OSType: linux
 Architecture: x86_64
 CPUs: 8
 Total Memory: 31.06GiB
 Name: fedora
 ID: JJM4:AEWX:KYZS:3X54:B3GH:3ACD:JK7S:AK4R:CDT2:MT4C:BVHT:CNYK
 Docker Root Dir: /var/lib/docker
 Debug Mode: false
 Registry: https://index.docker.io/v1/
 Labels:
 Experimental: false
 Insecure Registries:
  127.0.0.0/8
 Registry Mirrors:
  https://registry.e.securiton.int/
 Live Restore Enabled: true

答案1

因此，实际上问题在于使用docker compose build，继承的 NOFILE 限制太大（无穷大），这导致旧的 jdk-8 尝试保留过多的内存（参见https://stackoverflow.com/a/56895801/1737363）。

不幸的docker compose build是不支持该--ulimit参数，所以需要编辑docker systemd单元

cat /etc/systemd/system/docker.service.d/10-limits.conf
[Service]
LimitNOFILE=65536

然后重新加载并重新启动 systemd 单元

systemctl daemon-reload
systemctl restart docker

如果你想检查 docker build 期间的 ulimits，你可以使用这个 Dockerfile

FROM alpine

RUN ulimit -a

和这些命令

$ docker build . --no-cache --progress plain

$ docker-compose build --no-cache --progress plain

$ docker compose build --no-cache --progress plain

$ DOCKER_BUILDKIT=0 docker compose build --no-cache --progress plain

相关内容