disk_setup destroys data despite being told not to.

[James Masson]

This is the config used.

###
disk_setup:
    /dev/xvdf:
        layout: false
        overwrite: false
        table_type: gpt
fs_setup:
- device: /dev/xvdf
    filesystem: ext4
    label: c6dev-fraud-solr
    partition: none
    overwrite: false
###

1. If partition is set to "none", as above.

the code will skip to here - https://git.launchpad.net/~usd-import-team/ubuntu/+source/cloud-init/tree/cloudinit/config/cc_disk_setup.py?h=applied/debian/sid#n905

... bypassing all of the checks on whether the filesystem exists or not.

This will result in mkfs being executed every time, destroying data, despite "overwrite: false"

2. If partition is set to "any/auto" ....

... to trigger the smart partition matching in https://git.launchpad.net/~usd-import-team/ubuntu/+source/cloud-init/tree/cloudinit/config/cc_disk_setup.py?h=applied/debian/sid#n879

update_fs_setup_devices - https://git.launchpad.net/~usd-import-team/ubuntu/+source/cloud-init/tree/cloudinit/config/cc_disk_setup.py?h=applied/debian/sid#n156

... will detect the lack of partition and silently change the specified partition "any" or "auto" to None - triggering the problem again.

Summary:

No matter what you set for "partition" or "overwrite" in fs_setup - if you are using the raw device, then disk_setup/fs_setup will destroy your data every time.

Bug identified with cloud-init-0.7.5-10.el7.centos.1.x86_64 - from the looks of upstream git-master, it still exists there.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.