Why docker daemon fails to start when docker root is on a EMC powerpath device in Red Hat Enterprise

本文解决了一个特定问题：在Red Hat Enterprise Linux 7中，当Docker根目录位于EMC PowerPath设备上时，Docker守护进程启动失败。通过在systemd中添加依赖项，确保在底层逻辑卷挂载后Docker服务才开始启动，从而解决了此问题。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

https://access.redhat.com/solutions/3643862

环境

Red Hat Enterprise Linux 7
docker-1.13.1-75.git8633870.el7_5.x86_64
/var/lib/docker is a LV and the underlying PV is a EMC powerpath device
OpenShift Container Platform 3.9

问题

Why docker daemon fails to start when docker root directory is on a EMC powerpath device in Red Hat Enterprise Linux 7 ?
From the log messages:

Raw

Oct 08 17:35:27 test dockerd-current[3409]: Error starting daemon: error initializing graphdriver: devmapper: Unable to take ownership of thin-pool (vg_docker-docker--pool) that already has used data blocks

决议

Add a dependency so that docker service wont start until the underlying LV is mounted.

Create a file varlibdocker.conf under /etc/systemd/system/docker.service.d/ directory with the following contents

Raw

[Unit]
RequiresMountsFor=/var/lib/docker

After making the change, reload systemd using the command systemctl daemon-reload

根源

docker daemon was starting early even before the EMC powerpath devices were initialized. Since /var/lib/docker was not mounted, docker daemon will detect an empty /var/lib/docker and tries to initialize as if it was starting for the first time on the system. Due to this, it will not be able to use the existing docker thin pool and fails with the error messagedevmapper: Unable to take ownership of thin-pool (vg_docker-docker--pool) that already has used data blocks