在GCE中使用RexRay驱动创建Docker卷失败

Hey there, let's tackle that frustrating error you're hitting when trying to create a RexRay-backed GCEPD volume with Docker. I've worked through similar headaches with GCE and RexRay, so here are the actionable steps to get this sorted:

1. Double-Check RexRay Plugin Configuration #

First up, make sure your RexRay plugin has the correct GCE-specific settings. Even if you installed it via the managed plugin, the default config often misses critical details like your project ID or zone.

• Run this command to inspect the plugin's current settings:

  docker plugin inspect rexray/gcepd
  

  Look for environment variables like GCE_PROJECT and GCE_ZONE—they need to match your GCE instance's actual project ID and zone (e.g., my-gcp-project and us-central1-a).
• If they're wrong or missing, reconfigure the plugin with:

  docker plugin disable rexray/gcepd -f
  docker plugin set rexray/gcepd GCE_PROJECT=<your-project-id> GCE_ZONE=<your-instance-zone>
  docker plugin enable rexray/gcepd
  

2. Validate Service Account Permissions #

Even though you mentioned the service account has Project Editor access, let's confirm it has the specific permissions RexRay needs to interact with GCE disks:

• compute.disks.create
• compute.disks.delete
• compute.disks.get
• compute.disks.list
• compute.zones.get

Project Editor should include these, but sometimes custom IAM roles or organization-wide restrictions can block access. As a quick test, temporarily assign the Compute Admin role to the instance's service account—if the volume creation works, you know it's a permission issue that needs refining.

3. Check Docker-RexRay Compatibility #

Your Docker version (18.03.1-ce) is a bit outdated, and RexRay 0.8 (from the tutorial) can have compatibility quirks with older (or newer) Docker releases. Try one of these:

• Update Docker to a recent stable version (like 20.10.x) since 18.03 is end-of-life and no longer supported.
• Stick strictly to the RexRay 0.8 version from the tutorial: when reinstalling, use rexray/gcepd:0.8 instead of the latest tag, as newer RexRay versions changed how they handle GCEPD.

4. Verify GCE Disk Existence #

The error mentions "querying for existing volume"—RexRay might be struggling to check if a disk named storage1 already exists.

• Head to your GCE Console, go to Disks, and search for storage1. If it exists (even if it's unused), delete it and try creating the volume again.
• Note that GCE disk names are case-sensitive and must be lowercase—make sure your volume name storage1 follows this rule.

5. Enable Debug Logs for RexRay #

To get to the root of the query failure, turn on debug logging for the RexRay plugin:

docker plugin disable rexray/gcepd -f
docker plugin set rexray/gcepd REXRAY_LOGLEVEL=debug
docker plugin enable rexray/gcepd

Then retry the volume creation command, and check the logs with:

docker plugin logs rexray/gcepd

The logs will show exactly where the query is failing—whether it's an authentication issue, zone mismatch, or something else entirely.

6. Reinstall the RexRay Plugin #

Sometimes plugin installations get corrupted. Try removing and reinstalling it cleanly:

docker plugin rm rexray/gcepd -f
docker plugin install rexray/gcepd:0.8 --grant-all-permissions

After reinstalling, reconfigure the project and zone as we did in step 1.

In my experience, this error usually boils down to misconfigured zone/project settings or subtle permission issues that aren't obvious with Project Editor access. Start with steps 1 and 4—they're the quickest to validate and fix most cases.

内容的提问来源于stack exchange，提问作者Avinash