Tom Lane wrote: > Alex Mayrhofer <axelm@xxxxxxxx> writes: >> i have a bunch of indices over rather frequently updated large tables. Those >> indices grow in size with the updates, so i frequently re-index them. > > That usually shouldn't be necessary since PG 7.4 or so. Do you have > some strange pattern of index key usage that is triggering index bloat, > or is this just a habit leftover from pre-7.4? Hmm, my index definition (running 8.1.0) looks as follows: Index "public.l_renderjobs_uri_idx" Column | Type -------------+------------------------ request_uri | character varying(200) btree, for table "public.l_renderjobs" a recent REINDEX reduced that index from about 30000 pages to 18000. The table contains about 1.5 millions URLs, and i load http usage updates on the table every day, which modifies around 40000 of those records daily. So, is this something where bloating should be expected, or am i doing something wrong? thanks, Alex
