juju create-backup does not safeguard against using up all of the disk space on controller units

[Barry Price]

In an HA Juju 2.1.3 environment, I ran:

$ juju create-backup --filename mybackup.tgz

The size of Juju's data store (/var/lib/juju/db) in that environment was 25GB, but none of the three HA controllers had more than 15GB available disk space.

Fortunately I noticed before we hit ENOSPACE, and managed to kill it with / at 99% full.

The create-backup routine should make a conservative guess at how much disk space will be required, and refuse to run unless the amount of free space on the target partition is greater than that number.

[John A Meinel (jameinel)]

we should decide on backup-restore or restore-backup but not mix them.

On Jan 4, 2018 00:45, "Heather Lanigan" <email address hidden> wrote:

> ** Tags added: backup-restore
>
> --
> You received this bug notification because you are subscribed to juju.
> Matching subscriptions: juju bugs
> https://bugs.launchpad.net/bugs/1693780
>
> Title:
> juju create-backup does not safeguard against using up all of the disk
> space on controller units
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/juju/+bug/1693780/+subscriptions
>

[John A Meinel (jameinel)]

I think it is sane to have us check that the target directory for the backup has at least as much free space as the size of the Mongo Database that we are backing up (ignoring the backups database itself).
We can allow passing a flag like "--ignore-free-space" to 'juju create-backup' in case we get that estimate wrong and the user needs to do it anyway. But by default, we can do some quick checks to make sure that we have a strong expectation of not filling our whole disk and breaking the backup in the first place.

[Ian Booth (wallyworld)]

https://github.com/juju/juju/pull/13489