incremental backups

[venizio krups]

On Thu, Jan 21, 2010 at 2:34 AM, Igor Lautar <igor.lautar at gmail.com> wrote:
> On Thu, Jan 21, 2010 at 1:52 AM, venizio krups <0000.vk at gmail.com> wrote:
>> On Wed, Jan 20, 2010 at 5:45 PM,  <Deepak.Kumar1 at rbs.com> wrote:
>>> Is there a way to take incremental backups of the repository and then
>>> reconstitute them to a separate location as required ?
>>> Similar to a sql database where the logs can be separately backed up and
>>> then replayed
>>>
>>>
>>> thanks
>>> kumar
>>
>> check out hg bundles --base. write a little script/extension that
>> keeps track of the last hash id it was run on, and then pass that to
>> --base, update the hashid,  and so on ....
>>
>> we've done this for more than a year, and on one instance restored
>> from the incremental-backup-bundle store. works very well.
>
> Isn't this the same to doing pulls to a clone?

the incremental backups have to be encrypted (we encrypt the bundles
after creation), and there is no hg in the medium that hosts the
bundles. i know it's crazy.

> Anyway, for me, it seams problem could also be not only data inside
> repository, but repository conifiguraton as well (hgrc & other files
> that are not _part_ of repo).
> But this is only an issue if you use centra repo approach and actually
> _need_ different hgrcs there.
> To elegantly solve this, we use separate repo for these files and
> symlink them to their location. You can then just clone config repo
> and be done with it (and config repo hgrc also points to file inside
> this repo).
> There is also post-commit hook to update config repo local working dir
> so we can push new config - no need for login etc.

do you symlink .hg itslef?