Thread

1. [PATCH] HOT on tables with oid indexes broken

   Andres Freund <andres@2ndquadrant.com> — 2013-01-31T12:37:35Z

   Hi,
   
   The fklocks patch moved HeapSatisfiesHOTandKeyUpdate (or rather
   HeapSatisfiesHOTUpdate back then) to be called way earlier in
   heap_update as its needed to know which lock level is
   required. Unfortunately the oid of the new tuple isn't yet setup at that
   point.
   
   Due to this everytime there's an index on an oid - like in most catalog
   tables - no HOT updates will be performed as the old tuples oid will be
   compared to the new ones which isn't setup yet (so either InvalidOid or
   uninitialized).
   
   There's also a related bug which seems to go further back but has far
   fewer implications, namely that the tableOid of the old tuple isn't
   necessarily setup before determining HOTability. Now that only matters
   if there's an index on 'tableoid' which doesn't seem to be an all that
   frequent thing to do.
   
   Greetings,
   
   Andres Freund
   
   -- 
    Andres Freund	                   http://www.2ndQuadrant.com/
    PostgreSQL Development, 24x7 Support, Training & Services
   

2. Re: [PATCH] HOT on tables with oid indexes broken

   Alvaro Herrera <alvherre@2ndquadrant.com> — 2013-02-01T14:07:23Z

   Andres Freund wrote:
   > 
   > Hi,
   > 
   > The fklocks patch moved HeapSatisfiesHOTandKeyUpdate (or rather
   > HeapSatisfiesHOTUpdate back then) to be called way earlier in
   > heap_update as its needed to know which lock level is
   > required. Unfortunately the oid of the new tuple isn't yet setup at that
   > point.
   
   Applied, thanks.
   
   -- 
   Álvaro Herrera                http://www.2ndQuadrant.com/
   PostgreSQL Development, 24x7 Support, Training & Services