Docker无法启动（服务器空间不足）：如何手动删除日志与悬空镜像？

Got it, let's walk through this step by step—since Docker isn't running, we can't rely on the built-in docker system prune commands, but we can directly target the files that are eating up your server space.

一、清理Docker日志文件 #

Docker stores container logs by default in the /var/lib/docker/containers/ directory. Each container has its own subfolder, with log files typically named <container-id>-json.log (these are usually the biggest space hogs).

1. Identify large log files
   Run this command to see which container directories are taking up the most space:

   du -sh /var/lib/docker/containers/*
   

   This will show you the total size of each container's log and data files.

2. Clear or delete logs

   • If you want to completely remove the log files (safe since Docker isn't running):

     rm -rf /var/lib/docker/containers/<your-container-id>/*.log
     

     Replace <your-container-id> with the actual ID from the previous step.
   • If you prefer to keep the log files but empty their content (avoids potential issues if Docker recreates the container later):

     truncate -s 0 /var/lib/docker/containers/<your-container-id>/*.log
     

二、清理Docker镜像与悬空镜像 #

Docker images and their layered data are stored in /var/lib/docker/images/ (metadata) and /var/lib/docker/overlay2/ (actual file system layers—most modern setups use the overlay2 driver).

1. Check image and layer space usage
   First, get a clear sense of where the space is being consumed:

   du -sh /var/lib/docker/images/*
   du -sh /var/lib/docker/overlay2/*
   

2. Delete dangling images
   Dangling images are untagged images that aren't referenced by any container or other image. To remove them:

   • First, list all image IDs stored on your system:

     ls /var/lib/docker/images/*/imagedb/content/sha256/
     

   • Next, extract image IDs that are actually used by containers. Each container directory in /var/lib/docker/containers/ has a config.v2.json file referencing its image ID:

     grep -r "Image" /var/lib/docker/containers/*/config.v2.json | awk -F '"' '{print $4}' | cut -c7- | sort -u
     

   • Now, delete any image ID from the first list that doesn't appear in the second list (these are your dangling images):

     rm -rf /var/lib/docker/images/*/imagedb/content/sha256/<dangling-image-id>
     rm -rf /var/lib/docker/overlay2/<layer-id-associated-with-image>
     

     Note: Overlay2 layer IDs link to image layers—if you're unsure, you can safely delete overlay2 directories that don't have a corresponding entry in /var/lib/docker/image/overlay2/layerdb/sha256/.

3. Delete unused volumes (optional)
   If you have volumes you no longer need, you can delete them from:

   rm -rf /var/lib/docker/volumes/<volume-id>
   

   List all volumes with ls /var/lib/docker/volumes/ to identify which ones to remove.

Important Notes #

• Always back up any critical data before deleting files manually.
• After cleaning, restart Docker with sudo systemctl start docker—it should now have enough space to run.
• Once Docker is running, set up log rotation (e.g., via Docker's log-driver configuration) to prevent this issue from recurring.

内容的提问来源于stack exchange，提问作者Ayo K