Re: repository id for shared datastore - how to manage a cloned repo

[Amit Jain <am...@ieee.org>]

Hi,

[Adding the correct Oak users list]

No its not ok to have the same repository id for the cloned instances.
But we need a reset of the repository id for the cloned instance. This
piece of documentation is missing from the Oak docs but there's an oak-run
command 'resetclusterid' which is to be executed to reset that [1].
It's actually covered in the public documentation for AEM [2] though.

Thanks
Amit

[1] https://github.com/apache/jackrabbit-oak/tree/1.8/oak-run
[2]
https://helpx.adobe.com/experience-manager/6-4/sites/deploying/using/data-store-config.html#DataStoreConfigurations

On Thu, Aug 9, 2018 at 4:10 PM Stephan Matschinsky <ma...@adobe.com>
wrote:

> Hello oak devs,
>
>
> I’ve been referring to the documentation here:
>
>
> https://jackrabbit.apache.org/oak/docs/plugins/blobstore.html#check-shared-datastore-gc
>
> but have not been able to find the answers to my questions about the
> repositoryid and how it is used in the case of a shared datastore.
>
>
>
> Each repository has its own ID which is set at creation and this is
> recorded in the datastore (slightly different location if file datastore or
> S3 or Azure).  This allows for the garbage collection on the shared
> datastore with the mark and sweep methods.
>
>
>
> If I’m adding a new repository to the shared datastore, this is no problem
> because the new repo will have a new repository id.  The potential problem
> I see is when cloning a repo (for example when creating a new Publish
> instance to deal with added load on a website).  When cloning an AEM
> publish instance, the repositoryid is also cloned.
>
>
>
> *QUESTIONS*
>
>    - Is it OK for two publish instances to share the same repositoryid?
>       - I would think this is not a good idea because, even if each
>       publish instance should be identical, each repo could create specific
>       binaries.
>       - With two AEM 6.4 (oak 1.8.3) publish instances using the same
>       repositoryid I have not run into any issues so far but have only done
>       limited tests...
>    - How to ‘reset’ a repositoryid?
>       - Or, another option would be to clone the repo but get a new
>       repository ID.  In other words, how to clone a repository without copying
>       the repositoryid, so that when it connects to a shared datastore it is
>       registered correctly with a unique repositoryid?
>
>
>
> Thank you,
>
>
>
> Stephan Matschinsky
>
>
>