[Framework-Team] Re: The new Plone 4.0, was Re: Plone 3.5
l at lrowe.co.uk
Tue May 5 22:23:16 UTC 2009
Ross Patterson wrote:
> Andreas Zeidler <az at zitc.de> writes:
>> On May 5, 2009, at 10:05 PM, Ross Patterson wrote:
>>> BLOBs: Has the backups/repozo story been sufficiently worked out?
>> this will need a good backup story, but it won't be via repozo.
>> repozo was meant to backup a single data.fs, but not your entire
>> zodb. the blob storage will tend to be big and might live on some
>> media with other backup strategies (think SAN or S3). there should be
>> some recipe or something that provides a single script to backup both
>> for the standard use-case of having the blob storage live on the same
>> filesystem, but that shouldn't be repozo.
> I should clarify my question here. Is there an issue with making sure
> that the backed up BLOB directory is consistent with a particular backed
> up state of the Data.fs via repozo. IOW, can we say something like "so
> long as you restore your BLOB directory to a state as it was in the same
> moment or after the repozo process started then it is guataneed to be
> consistent"? I'm not saying that the above statement is correct cause I
> don't know. :) I'm just saying we'd need to be able to make some
> promise about repozo backups of Data.fs and backups of BLOB directory
> being consistent.
No problem here, you just take the copy of your BLOB directory after you
take the copy of your Data.fs. The dangling blobs in the backup (those
created since your backup of the Data.fs) are not an issue.
Creating a consistent backup of two filestorages (e.g. Catalog.fs and
Data.fs) is more tricky.
More information about the Framework-Team