Op 23-02-11 12:41, Maurits van Rees schreef: > CLUSTER does not do anything unless you first tell it which index to use > for clustering a table. Using postgres 8.4 I have done this for the > four tables for which I think this is useful (the other tables had no > content): > > cluster blob_chunk using blob_chunk_pkey; > cluster current_object using current_object_pkey; > cluster object_state using object_state_pkey ; > cluster transaction using transaction_pkey; > > This did not have an effect on the above mentioned test database though. > But that may be because it had already been fully vacuumed. > > Would these be the correct tables and indexes? Some of these tables > have other indexes as well, but these are the primary ones.
I just did this and the database went down from 38 GB to 21 GB, so I'd say this works. > Would it make sense to add these lines to the postgres code in > relstorage that creates the tables? Then users would only need to > perform the 'CLUSTER;' command without any extra parameters and it would > work. -- Maurits van Rees Web App Programmer at Zest Software: http://zestsoftware.nl Personal website: http://maurits.vanrees.org/ _______________________________________________ For more information about ZODB, see the ZODB Wiki: http://www.zope.org/Wikis/ZODB/ ZODB-Dev mailing list - ZODB-Dev@zope.org https://mail.zope.org/mailman/listinfo/zodb-dev