CentOS 6.7虚拟磁盘导入gcloud后无法启动VM实例求助

Let's break down the core issues here and walk through step-by-step fixes to get your CentOS 6.7 instance running smoothly on GCP:

1. Critical Mistake: Mismatched OS Parameter During Import #

Looking at your import command, you used --os centos-9 for a CentOS 6.7 image—this is a major red flag. GCP's image import tool uses the --os flag to apply OS-specific boot optimizations, driver configurations, and compatibility tweaks. Mismatching the OS version will almost certainly break the boot process.

Fix: Use the Correct OS Flag #

Always specify centos-6 for any CentOS 6.x image:

gcloud compute images import centos67-custom-image --source-file gs://your_gcs_bucket/centos67.raw --os centos-6

2. Prepare Your Source OpenStack VM for GCP Compatibility #

CentOS 6.7 is an older release, so it needs targeted tweaks to work with GCP's virtio-based virtualization stack. Before exporting the image from OpenStack, apply these changes to your source VM:

• Enable Virtio Drivers: GCP uses virtio storage/network devices, which CentOS 6 may not load by default.

  1. Add virtio modules to your initramfs to ensure they load on boot:

     mkinitrd --with virtio_blk --with virtio_net /boot/initramfs-$(uname -r).img $(uname -r)
     

  2. Update GRUB to include virtio kernel parameters and enable serial console output (critical for debugging):
     Edit /etc/grub.conf and add console=ttyS0 virtio_blk.ignore_index=1 to the kernel line.

• Fix /etc/fstab: Avoid using UUIDs for the root mount—GCP's disk device names will differ from OpenStack. Use /dev/sda1 or a filesystem label instead:

  /dev/sda1 / ext4 defaults 1 1
  

• Replace NetworkManager with Legacy Network Service: GCP relies on the network service for network configuration, not NetworkManager.

  service NetworkManager stop
  chkconfig NetworkManager off
  chkconfig network on
  service network start
  

• Adjust SELinux: Set SELinux to permissive mode to avoid boot-time permission conflicts:
  Edit /etc/selinux/config and set SELINUX=permissive, then reboot the source VM before exporting.

• Clean Up OpenStack-Specific Data: Remove temporary metadata and network configs that could conflict with GCP:

  rm -rf /var/lib/cloud
  rm -f /etc/sysconfig/network-scripts/ifcfg-*
  touch /etc/sysconfig/network-scripts/ifcfg-lo
  

3. Resolve the import_precheck Tool Error #

The "must be run as root" error despite using root is likely due to binary compatibility—GCP's precheck tool may be built for newer OS versions that aren't compatible with CentOS 6.7. Instead of relying on the tool, manually verify all the checks above—they cover every critical requirement for GCP compatibility.

4. Re-Import and Debug the Image #

After applying the fixes to your source VM, export it as a raw or qcow2 file (raw is preferred for GCP), upload it to your GCS bucket, and re-run the import command with the correct --os centos-6 flag.

If the VM still fails to boot, use the serial console to diagnose specific errors:

# Enable serial console
gcloud compute instances add-metadata your-centos67-vm --metadata=serial-port-enable=1
# View serial output
gcloud compute instances get-serial-port-output your-centos67-vm

Common errors to watch for:

• "No root device found": Virtio drivers are missing or /etc/fstab is misconfigured.
• "Network initialization failed": NetworkManager is still enabled or virtio_net drivers aren't loaded.
• SELinux denial messages: Confirm SELinux is set to permissive.

5. Alternative: Use GCP's Official CentOS 6 Image #

If manual tweaks are too time-consuming, launch a VM with GCP's pre-built CentOS 6 image and migrate your data from the OpenStack VM:

gcloud compute instances create centos67-public --image-family centos-6 --image-project centos-cloud

Transfer your data via SCP or rsync once the instance is running.

内容的提问来源于stack exchange，提问作者Deepak Bansal