Re: multi-column btree index for real values
От | Martin Weinberg |
---|---|
Тема | Re: multi-column btree index for real values |
Дата | |
Msg-id | 200210052305.g95N5avH002186@osprey.astro.umass.edu обсуждение исходный текст |
Ответ на | Re: multi-column btree index for real values (Bruce Momjian <pgman@candle.pha.pa.us>) |
Список | pgsql-general |
Ok. I don't have the disk space to migrate to 7.3 right now. Our database is about 500 million tuples with about 100 fields. However, I checked out the beta source and noticed the following new lines near line 934 in src/backend/access/rtree/rtree.c in 7.2.2: + pfree(DatumGetPointer(union_dl)); + pfree(DatumGetPointer(union_dr)); Looks like a leak fix to me. Patching these in, I found that the leak is much reduced; still looks like a tiny leak remains. We will try 7.3beta asap. Thanks. Bruce Momjian wrote on Sat, 05 Oct 2002 17:39:45 EDT >Martin Weinberg wrote: >> Thanks Bruce. Some simple tests on a 10 million tuple data base shows >> that r-tree works well for this. (It took me a while to realize >> that I had to sort boxes of zero area rather than points). >> >> However, it seems that the rtree index has a serious memory leak for >> 7.2.2. Is that known? > >Uh, I am not aware of that, but you can try 7.3beta to see if it is >better, and if not, report back. > >-- > Bruce Momjian | http://candle.pha.pa.us > pgman@candle.pha.pa.us | (610) 359-1001 > + If your life is a hard drive, | 13 Roberts Road > + Christ can be your backup. | Newtown Square, Pennsylvania 19073 > >---------------------------(end of broadcast)--------------------------- >TIP 4: Don't 'kill -9' the postmaster >
В списке pgsql-general по дате отправления: