[Pgcluster-general] PGLB question

[Zoltan Boszormenyi]

Hi,

I noticed that when I try to dump a largish database via pglb
pg_dump stalls while receiving COPY data, while trying
to dump connecting directly to the node doesn't show this behaviour.
Using "pglb -n -v" I noticed many similar messages:

2008-07-28 12:17:11 [1143] DEBUG:read kind from backend d
2008-07-28 12:17:11 [1143] DEBUG:pool_process_query: kind from backend: d
2008-07-28 12:17:11 [1143] DEBUG:read kind from backend pending data d
len: 1010 po: 13
2008-07-28 12:17:11 [1143] DEBUG:pool_process_query: kind from backend: d
2008-07-28 12:17:11 [1143] DEBUG:read kind from backend pending data d
len: 997 po: 26
2008-07-28 12:17:11 [1143] DEBUG:pool_process_query: kind from backend: d
2008-07-28 12:17:11 [1143] DEBUG:read kind from backend pending data d
len: 984 po: 39
2008-07-28 12:17:11 [1143] DEBUG:pool_process_query: kind from backend: d
2008-07-28 12:17:11 [1143] DEBUG:read kind from backend pending data d
len: 971 po: 52

Does it mean that pglb collects all data before passing it to the client?
It's not a good behaviour, pglb may OOM in case I want to dump
a VERY large table.

Best regards,
Zoltán Böszörményi

-- 
----------------------------------
Zoltán Böszörményi
Cybertec Schönig & Schönig GmbH
http://www.postgresql.at/