• Article Type: General
• Product: Aleph
• Product Version: 18.01

Description:
Is it possible to do full reindexing of a database on a separate server to minimize impact on the production machine?

Resolution:
We've tested and actually done a production reindexing of our XXX01 bib database on Aleph using an alternative parallel indexing method.

To accomplish this we used the transportable tablespace feature of Oracle. By using this functionality we were able to do our indexing on a separate machine from our production database. We then exported the tablespaces that contained the newly built indexes from our testing/development machine and imported the tablespaces to our production environment with minimal downtime. We then used the Z07H (util/e/21-22 functionality) table to reindex the delta documents that were updated while reindexing.

Attached is a white paper that describes in more detail how to accomplish this.

Another site doing something very similar adds: We found that the "traditional" parallel indexing didn't work with our central, shared LCSH/NAF authority file which runs in its own Aleph instance. The headings linking didn't work right.

We use the Oracle Recovery Manager (RMAN) to move our data around, and aside from the fact that our remote authority databases being a problem with "traditional" parallel indexing, we have multiple copies of our databases (TEST, Production, and Reporting/failover/standby). Maintaining all of the "parallel indexing disk" required by the traditional method on all three copies of each of our databases seemed like a very expensive proposition for an activity that would happen only on occasion.

A third site adds: Another option, to conserve the disk space, or be able to get that space back afterward is to put the parallel indexing tables into their own regular tablespaces (modelled after the TS4D and TS4X that they're normally in). After moving the new indexes into the production library, I can just drop the entire indexing tablespace until next time. Our indexes use roughly half of the entire database size.

Thanks to Alan Rykhus of the PALS Consortium, Michele Newberry and Lydia Motyka of FCLA and Christine Moulen of MIT for posting this information to the ELUNA-ALEPH-SYSADMIN-IG-L list.

• Article last edited: 10/8/2013