On Fri, 05 Jun 2009 06:18:45 -0400, sebb <[email protected]> wrote:
The funny thing is that 2.3.2 worked, but not
exactly as I expected. I have 2 threads (in a connection pool with max
2
connections) running 2 iterations, and in the postgres syslog output I
saw 4
sequences of parse/bind/execute. I was expecting to see the "parse"
step
omitted on the 2nd and subsequent iterations for each thread, e.g.
parse/bind/execute/bind/execute.
Using a connection pool count > 0 is not recommended, as the
connections aren't tied to particular threads. See if the same happens
with max=0, which ensures a single connection per thread.
Sadly the same happens with max=0. To simplify I set threads=1 and loop
count=2; the JDBC connection has max connections=0. And in the postgres
log I see two rounds of parse/bind/execute (below). In my own program
which uses libpq I see the parse step when I prepare the statement, and
then only bind/execute for every time I execute the prepared statement.
So while this is working in 2.3.2, it doesn't appear to be working right.
If I get the chance I will inspect the JDBC code.
Jun 5 09:07:10 kc5 postgres[10529]: [2-1] LOG: duration: 0.313 ms parse
<unnamed>: select $1
Jun 5 09:07:10 kc5 postgres[10529]: [3-1] LOG: duration: 0.097 ms bind
<unnamed>: select $1
Jun 5 09:07:10 kc5 postgres[10529]: [3-2] DETAIL: parameters: $1 = '1'
Jun 5 09:07:10 kc5 postgres[10529]: [4-1] LOG: duration: 0.016 ms
execute <unnamed>: select $1
Jun 5 09:07:10 kc5 postgres[10529]: [4-2] DETAIL: parameters: $1 = '1'
Jun 5 09:07:10 kc5 postgres[10529]: [5-1] LOG: duration: 0.069 ms parse
<unnamed>: select $1
Jun 5 09:07:10 kc5 postgres[10529]: [6-1] LOG: duration: 0.030 ms bind
<unnamed>: select $1
Jun 5 09:07:10 kc5 postgres[10529]: [6-2] DETAIL: parameters: $1 = '1'
Jun 5 09:07:10 kc5 postgres[10529]: [7-1] LOG: duration: 0.010 ms
execute <unnamed>: select $1
Jun 5 09:07:10 kc5 postgres[10529]: [7-2] DETAIL: parameters: $1 = '1'
Ken
--
Using Opera's revolutionary e-mail client: http://www.opera.com/mail/
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
