There was a bug in the connectionpooling code in 1.6.x (which was fixed back then). The problem was that prepared statements where closed, but NOT the ResultSet. This should be ok according to the jdk api-doc, but with postgres, this resulted in a resource leak.
Maybe, somewhere else in the code, resultsets are not closed, resulting in this kind of errors. Edwin -----Oorspronkelijk bericht----- Van: Michiel Meeuwissen [mailto:[EMAIL PROTECTED] Verzonden: vrijdag 11 juni 2004 10:09 Aan: [EMAIL PROTECTED] Onderwerp: Re: PostgreSQL runs out of connections Ronald Wildenberg <[EMAIL PROTECTED]> wrote: > Hi, > > We have a problem with our PostgreSQL database in combination with MMBase. > Since we switched to MMBase 1.7 the database repeatedly reports: 'Sorry, > too many clients already'. This happens about once every three or four > days. > > This problem did not occur when we were using MMBase 1.6.3. I vaguely > remember a similar problem with MMBase 1.6.0 and PostgeSQL, but I'm not > really sure what caused the problem back then. > > Are there ideas on the cause of this problem? Are there any other database related messages in the log? E.g. that queries are killed because they took too long (in such cases a new connection must be acquired, and I imagine that the mentioned message would occured then if in the mean time postgresql has not connections to give left). Michiel -- Michiel Meeuwissen Mediacentrum 140 H'sum +31 (0)35 6772979 nl_NL eo_XX en_US mihxil' [] ()
