[ckan-dev] Support for updates in the SOLR schema

[David Read]

2011/11/30 Adrià Mercader <adria.mercader at okfn.org>:
> Hi all,
>
> I've put up a small document with all the details I've been gathering
> investigating SOLR. Those include details on how to set up multiple
> cores and a proposal on how to handle updated in the schema used by
> CKAN.
> It would be great if anyone interested could have a look at it and
> give me his/her thoughts.
> In a nutshell, my proposal is to keep different versions of the schema
> in the CKAN source and then handle different versions at the SOLR
> server using different cores.

Adria,

This sounds like a good investigation and certainly a clear explanation.

We're already on the third SOLR schema, and I have no idea of their
compatibility with different versions of CKAN, so I believe this is
essential now.

I'm not convinced it is best to store the lowest valid ckan version in
the schema. It is the CKAN code's responsibility to know which schema
versions it is compatible with, rather than the other way round.

The extra thing I would like is for CKAN to fail if it is hooked up to
a SOLR which is running the wrong schema version. I can just see this
as something we miss on deployment and will cause headaches. So can I
suggest that somewhere on CKAN startup it gets hold of the schema
(e.g. http://localhost:8983/solr/admin/file/?file=schema.xml) and
checks the version matches the version in it's local repo's schema or
a hard-coded list of other compatible schema versions. Maybe we should
also have an environment variable to disable this check, for testing
purposes?.

David

> Unfortunately, I don't think there is a solution to support all
> current sites if we were to update the schema in the SOLR servers, but
> hopefully we can do put a setup in place that will make future updates
> smoother.
>
>
> Cheers,
>
> Adrià
>
> _______________________________________________
> ckan-dev mailing list
> ckan-dev at lists.okfn.org
> http://lists.okfn.org/mailman/listinfo/ckan-dev
>