Re: [HACKERS] Hashjoin status report - Mailing list pgsql-hackers

From Michael Contzen
Subject Re: [HACKERS] Hashjoin status report
Date
Msg-id 01be9871$fa35aa60$4dbe7fc2@wcontzen.dohle.com
Whole thread Raw
Responses Re: [HACKERS] Hashjoin status report  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
Hello,
 
(using snapshot of May, 5th)
 
because we have the need to have a workaround to this hash problem, I looked into the hashing code (well, without having the background).
 
For reducing the probability of an overflow I increased
#define FUDGE_FAC 3
witch was originally 1.5. I think it´s a data dependend constant (correct?) and for my data it works...
 
It does the job, but certainly that this is not the solution.
 
Increasing -B 256 doesn´t work:
NOTICE: Buffer Leak: [248] (freeNext=0, freePrev=0, relname=, blockNum=0, flags=0x0, refcount=0 25453)
pq_flush: send() failed, errno 88
pq_recvbuf: recv() failed, errno=88
 
Kind regards,
 
Michael Contzen
Dohle Systemberatung, Germany

 
Kind regards,
 
Michael Contzen
Dohle Systemberatung, Germany

 

pgsql-hackers by date:

Previous
From: ZEUGSWETTER Andreas IZ5
Date:
Subject: AW: [HACKERS] cvs 6.5 regression tests on Linux x86
Next
From: "Tim Perdue"
Date:
Subject: Weird Errors in 6.4.2 - Indexes/Vacuuming