Thread: unexpected EOF on client connection errors
Configuration: Postgres 8.0.3 Apache 2.0.52 PHP 4.3.9 ADOdb 4.65 I have a web solution that works extremely well. Yet, the following error appears in the logs a lot. At least one entry every 5 minutes, quite a few more when the site is busy. webuser - xxx.xxx.xxx.xxx(59420) 2005-10-10 11:00:29 EDT> LOG: 08P01: unexpected EOF on client connection webuser - xxx.xxx.xxx.xxx(59420) 2005-10-10 11:00:29 EDT> LOCATION: SocketBackend, postgres.c:287 As far as I can tell, I'm not losing any information when this error occurs. Googling for the problem yields a mixed bag of reasons for the error but didn't give me a clue as to how to begin tracking it down. Can anyone give me an idea of where to begin looking for the cause of this error? Thanks, Charley
"Charley L. Tiggs" <ctiggs@xpressdocs.com> writes: > I have a web solution that works extremely well. Yet, the following > error appears in the logs a lot. At least one entry every 5 minutes, > quite a few more when the site is busy. > webuser - xxx.xxx.xxx.xxx(59420) 2005-10-10 11:00:29 EDT> LOG: > 08P01: unexpected EOF on client connection There used to be some client libraries that would just drop the connection when they were done, instead of sending a polite Terminate message first; which would lead to the above bleat by the backend. I don't think that's done anymore by any current versions, but if you've got one or two applications bound to old libraries then that could account for it. The other possibility that comes to mind is connections being lost due to network issues (eg, firewall timeouts); but if that were the problem then I'd expect you'd be seeing client-side complaints too. If you aren't, it's probably just an impolite client library. Which is not very important, but if you want to get rid of the log messages... > Can anyone give me an idea of where to begin looking for the > cause of this error? It might help to turn on log_connections and see if you can correlate the connections that get "unexpected EOF" with a particular client app. regards, tom lane
Hi, On 10/10/05, Charley L. Tiggs <ctiggs@xpressdocs.com> wrote: > I have a web solution that works extremely well. Yet, the following > error appears in the logs a lot. At least one entry every 5 minutes, > quite a few more when the site is busy. > > webuser - xxx.xxx.xxx.xxx(59420) 2005-10-10 11:00:29 EDT> LOG: > 08P01: unexpected EOF on client connection > webuser - xxx.xxx.xxx.xxx(59420) 2005-10-10 11:00:29 EDT> LOCATION: > SocketBackend, postgres.c:287 AFAIC, above messages was caused by the 0 return of recv() call, which makes me think that your PHP scripts's execution maybe killed for some reason instead of a clean exit. (In practice, PHP closes all of its previous database connections (unless it's a persistent one) at the exit stage of the script. Namely, pg_close() doesn't matter on a non-persistent database connection.) Maybe execution takes to much time and the script is killed by PHP or any other external proccess before the exit stage. H.T.H.