[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [RFC] RFC on Backup tool
From: |
Stefan Hajnoczi |
Subject: |
Re: [Qemu-devel] [RFC] RFC on Backup tool |
Date: |
Fri, 21 Jul 2017 16:30:35 +0100 |
User-agent: |
Mutt/1.8.0 (2017-02-23) |
On Tue, Jul 18, 2017 at 05:29:11PM -0400, John Snow wrote:
>
>
> On 07/17/2017 03:37 PM, Ishani wrote:
> > ----- On Jul 17, 2017, at 12:48 PM, Fam Zheng address@hidden wrote:
> >> On Sun, 07/16 02:13, Ishani Chugh wrote:
>
> [...]
>
> >> Only full backup is implemented in this patch, is the plan to add
> >> incremental
> >> backup on top? I'm curious because you have target file path set during
> >> drive
> >> add, but in the incremental case, it should be possible that each backup
> >> creates
> >> a new target file that chains up with earlier ones, so I think the target
> >> file
> >> should be an option for "backup" command as well.
> >
> > Yes. Incremental backup is to be added. I am still in learning phase with
> > respect to incremental backups. I will modify the arguments and workflow
> > accordingly.
> >
>
> You may consider solidifying the backup target *pattern* during drive
> add as an alternative, such as:
>
> .../path/to/backup/%VM%/%DRIVE%/%yyyy%-%mm%-%dd%.qcow2
>
> Or some such scheme. Simple numerals work well, too:
>
> myvm/sda/incr.0.qcow2
> myvm/sda/incr.1.qcow2
>
> Simple numerals offer the benefit that it is easier to reconstruct the
> chain if you lose your metadata in the python script.
>
> Also consider that even for non-incremental backups, we want full
> backups made subsequently to not, in general, overwrite the previous
> full backup, so the TARGET is more of a "living entity" than a fixed
> thing, even in the simple case.
Patterns would be nice. You may find string.Template() useful:
https://docs.python.org/2.7/library/string.html#template-strings
https://docs.python.org/3/library/string.html#template-strings
signature.asc
Description: PGP signature