Bug #12084 Updated: pg_close cause a lot of sockets in TIME_WAIT state

ID: 12084 Updated by: thies Reported By: gianpaolo@preciso.net Status: Open Bug Type: PostgreSQL related Operating System: Debian 2.2 PHP Version: 4.0.6 New Comment: this is not a bug in php. here's the reply from Jan Wieck (PostgreSQL core guy) also read http://globecom.net/ietf/draft/draft-faber-time-wait-avoidance-00.html AFAIK TIME_WAIT is the status TCP/IP connections enter after successful close to hold the port number for some time just in case there are retransmit packets and the like still on the wire, so that no other process right now can bind to this port and gets confused when these orphaned packets arrive. Well, on 127.0.0.1 this doesn't really make sense. And I think it can be avoided by using SO_REUSE, but I'm not 100% sure on that. OTOH it's only a real problem if you open/close many, many, many (and I really mean many) connections in a short time. These TIME_WAIT's disappear after a while and they don't eat bread in the meantime. So if you don't run out of unprivileged ports, why worry about them? Previous Comments: ------------------------------------------------------------------------ [2001-07-12 02:42:29] gianpaolo@preciso.net php-4.0.6 postgres 7.1.2 I noticed a lot of sockets leaved in TIME_WAIT state in our application. so I worte a simple script <pre> for ($i = 0 $i < 100 ; $i++){ $db = pg_open ("some string"); pg_close($db); } <pre> After that I noticed around 200 sockets in TIME _WAIT state. If I omit the pg_close statement averything works fine (and is considerably faster). The same thing happens if I do a query before reopen a connection. This bug is not reproducible with postgres 7.0.x Thanks and goodbye ------------------------------------------------------------------------ Edit this bug report at http://bugs.php.net/?id=12084&edit=1