feature request: quick disk erase

[Ryan Beisner]

Please consider adding a "quick erase" feature, such as writing 0s to the first and last N number of blocks.

The current disk erase feature can take many hours to complete and release the node, which is not realistically usable in test automation where many deploys are exercised every day.

I think this feature would also have value for normal users who want to ensure that the partition table (specifically GPT) is adequately ruined, but aren't concerned about a full or secure wipe.

As an example: we run into particular issues with ceph, where a previous deployment can affect a new deployment, because GPT survives on /dev/sdb.

Thank you!

[Christian Reis (kiko)]

How long does hdparm --secure-erase take? On an SSD I believe it just swaps out the encryption key, so it should be instantaneous.

[Ryan Beisner (1chb1n)]

Linking for reference with regard to gpt:
http://blog.gnub.net/2009/03/die-gpt-die.html
http://askubuntu.com/questions/211477/how-to-remove-gpt-from-hdd

[Christian Reis (kiko)]

See also bug 1507724

[Blake Rouse (blake-rouse)]

This is already being done in MAAS 1.9. We a node is deployed all disks get there superblock wiped and the start of every partition also gets wiped, just in case its the exact same partition offset as a previous deployment.