Image is being modified in import-snapshot

[landgraf]

Describe the bug

Even if this process is used the image is modified which causes boot/mount issues afterwards.

Regression Issue

• Select this option if this issue appears to be a regression.

Expected Behavior

/etc/fstab remains unmodified:

LABEL=visopersist /data   ext4    defaults        0       0
LABEL=visodata    /viso   ext4    defaults        0       0

Current Behavior

sh-5.2# cat /etc/fstab 
LABEL=visopersist /data   ext4    defaults        0       0
LABEL=visodata    /viso   ext4    defaults        0       0
# This was generated by AWS
# and broke the boot because sda is not available (only /dev/nvme0n1p1 is)
/dev/sda1       /boot   vfat    defaults        0       0
/dev/sda4       /data   ext4    defaults        0       0
/dev/sda5       /viso   ext4    defaults        0       0

Reproduction Steps

The image has been imported:

  aws ec2 import-snapshot   --region eu-central-1   --disk-container 'file://image-import.json'
 aws ec2 register-image --name="VisoOS-testAMI" --cli-input-json="file://register-ami.json"

with:

aws-cli/1.41.9 Python/3.13.5 Linux/6.15.3-arch1-1 botocore/1.39.9

Possible Solution

No response

Additional Information/Context

No response

CLI version used

aws-cli/1.42.2 Python/3.13.5 Linux/6.15.3-arch1-1 botocore/1.40.2

Environment details (OS name and version, etc.)

both pip and system (aws-cli/1.41.9 Python/3.13.5 Linux/6.15.3-arch1-1 botocore/1.39.9) produce same results

[adev-code]

Hi @landgraf, thanks for reaching out. I have tried to reproduce the issue but I did not see any modifications on /etc/fstab. I tried using the most up to date CLI on V1 and V2 and also aws-cli/1.42.2. For investigation,

1. Could you tell more about how to repro it?
2. What was the imported VM? The VM that was used?
3. To share full --debug on the commands? And also if possible for full --debug on where there is no modification on /etc/fstab, just to compare results.
   Thanks.

[landgraf]

Hi @landgraf, thanks for reaching out. I have tried to reproduce the issue but I did not see any modifications on /etc/fstab. I tried using the most up to date CLI on V1 and V2 and also aws-cli/1.42.2. For investigation,
Hi @adev-code
Thank you for looking into this

1. Could you tell more about how to repro it?
   I'm not sure what exactly. I've built OS image locally, uploaded it to s3 (tried both raw as well as vhd formats) and tried to import the snapshot with this import.json:

{
    "Description": "ewaol docker image",
    "Format": "vhd",
    "UserBucket": {
        "S3Bucket": "${BUCKET_NAME}",
        "S3Key": "os-images/"
    }
}

and registered AMI with ` aws ec2 register-image --name TestAMI --cli-input-json="file://register-image.json" :

{
    "Architecture": "x86_64",
    "BlockDeviceMappings": [
        {
            "DeviceName": "/dev/sda1",
            "Ebs": {
                "DeleteOnTermination": true,
                "SnapshotId": "snap-01120dc9cd2bf17c5",
                "VolumeSize": 60,
                "VolumeType": "gp2"
            }
        }
    ],
    "Description": "MyAMI",
    "RootDeviceName": "/dev/sda1",
    "BootMode": "uefi",
    "VirtualizationType": "hvm",
    "EnaSupport": true
}

According to the documentation the image should not be modified at all, however I see modifications of fstab which broke the image boot sequence. AWS' pipeline adds /dev/sda to my fstab, maybe it has something to do with register-image flow and not import-snapshot because it has root device mapping in the config?

All steps have been taken from the meta-aws layer: https://github.com/aws4embeddedlinux/meta-aws/blob/master/scripts/ec2-ami/create-ec2-ami.sh

2. What was the imported VM? The VM that was used?

The imported VM was raw/vhd image I've built locally using Yocto Project build system. Nothing really special about it (I hope), you can see fstab in the issue discription. The original fstab didn't container any record for / or /boot though, not sure if this fact triggers AWS smart logic to "fix" the boot process.

4. To share full --debug on the commands? And also if possible for full --debug on where there is no modification on /etc/fstab, just to compare results.> Thanks.

I'm attaching log file from the import-snapshot command. with auth token/buckets names etc deleted.

import.log