[ckan-dev] Search: uniqueKey when indexing

Tue Nov 8 16:55:29 UTC 2011

Hi all,

As you may know, with some kind help from Rufus and Friedrich we
solved the issues with the IATI SOLR server [1] (A misconfiguration
issue made that both cores shared the same data dir).
While investigating the issue, I changed the indexing code and the
SOLR schema to use a generated hash as unique key instead of the
package id [2]. This turned out to be not relevant to the IATI issue,
but I think it might be worth keeping it and merging to default at
some point (not for 1.5!). Think e.g in a CKAN instance that harvests
records from another one and both share the same SOLR server.
The only problem is that schemas will need to be updated and indexes
rebuilt, so there will be some migration work to do.

What do you guys think?

Adrià

[1] http://trac.ckan.org/ticket/1430
[2] https://bitbucket.org/okfn/ckan/changeset/855f5a452f60