Re: [Duplicity-talk] Restoring from Glacier to S3

[edgar . soldin]

On 29.07.2015 12:39, Benjamin Henrion wrote:
> On Wed, Jul 29, 2015 at 11:01 AM, Andrew Langhorn
> <address@hidden> wrote:
>> Compared to S3, Glacier will take a lot longer to restore. It's primary
>> purpose is archiving whereas S3 is an object store. The trade-off is speed
>> vs price, since you should rarely need to restore from a Glacier vault.
>>
>> Do you have auto-archiving set on your S3 bucket? You can check in the
>> console. If so, turn that off, and your files should then stay in the S3
> 
> Yes, to 30 days, and the signature file I mentioned earlier is trapped
> in glacier mode.
> 
> I will disable manually the S3 policy of 30 days and use the purge
> --force command to 30 days instead.
> 
> It would be nice if both could be in sync because that would make a
> restore quasi impossible, as the restore from Glacier is not that fast
> and the process will timeout at some point.
> 
>> bucket assuming Duplicity isn't talking to Glacier some other way - which
>> given the s3:// URIs being used seems unlikely, but is possible, I suppose.
>>
>> Do you have the sigfiles in the Duplicity cache? I might be wrong, but I
>> thought that if they were in the cache, it wouldn't download them. It may be
>> that the sigfiles it's downloading are new.
> 
> I tried to restore from another machine which did not had the cache.
> 

Benj,

are you aware of the 
 --file-prefix, --file-prefix-manifest, --file-prefix-archive, 
--file-prefix-signature
switches? they were added by a user not long ago who needed them to create 
backend file names that could be matched by the s3 glacier routine for decision 
to selectively put on ice or not.

..ede/duply.net