Обсуждение: unexpected EOF on client connection vs 9.0.3
Hello there, I'm getting regular Postgres log entries with the following error: 2011-03-07 01:00:01 CET LOG: could not receive data from client: No connection could be made because the target machine actively refused it. 2011-03-07 01:00:01 CET LOG: unexpected EOF on client connection They always appear together. The server is PostgreSQL 9.0.3, compiled by Visual C++ build 1500, 64-bit running on Windows 2008 R2 SP1 x64, the client is a ASP.NET application running npgsql. I've checked and verified that all connections are closed within the code, what's more, the problem has appeared just as I've moved server from Fedora Linux x86_64 running Postgres 8.4.2 to the Windows and 9.0.3 (details above) thus I conclude this is not a client problem indeed (the failure didn't occure on Linux). The connections are closed from time to time immediatelly as I get ASP.NET errors. Thanks for any help in advance. Piotr Czekalski P.S. The connection string is: ...Server=127.0.0.1;Port=5432;User Id=<user here>;Password=<password here>;Database=<database here>;CommandTimeout=360;" providerName="Npgsql"... -- -------------------------------------------------------------- "TECHBAZA.PL" Sp. z o.o. Technologie WEB, eDB& eCommerce tel. (+4832) 7186081 fax. (+4832) 7003289 email: biuro@techbaza.pl
On 08/03/11 02:49, Piotr Czekalski wrote: > I've checked and verified that all connections are closed within the > code, what's more, the problem has appeared just as I've moved server > from Fedora Linux x86_64 running Postgres 8.4.2 to the Windows and 9.0.3 > (details above) thus I conclude this is not a client problem indeed (the > failure didn't occure on Linux). Windows firewall? You can also see these error reports when the connections are closed uncleanly, without a proper backend close message. Perhaps you have client processes crashing? Or doing hard shutdowns where the client code doesn't get a chance to run any cleanup/dtors/etc? -- Craig Ringer
On Tue, 08 Mar 2011 11:30:10 +0800, Craig Ringer wrote: > On 08/03/11 02:49, Piotr Czekalski wrote: > >> I've checked and verified that all connections are closed within the >> code, what's more, the problem has appeared just as I've moved >> server >> from Fedora Linux x86_64 running Postgres 8.4.2 to the Windows and >> 9.0.3 >> (details above) thus I conclude this is not a client problem indeed >> (the >> failure didn't occure on Linux). > > Windows firewall? > > You can also see these error reports when the connections are closed > uncleanly, without a proper backend close message. Perhaps you have > client processes crashing? Or doing hard shutdowns where the client > code > doesn't get a chance to run any cleanup/dtors/etc? > > -- > Craig Ringer I think in 8.x releases is no need to make any spacial close operation, or at least You do not get notice, about it. Closing socket is enough. E.g. JDBC driver closes connection, by closing socket. In 9 this changed and I see many of such notices in log. Probably Your driver, as many others, is written to close socket without backanad message. Regards, Radek
Indeed those messages appear when the client closes the socket connection only. Npgsql has connection pooling and if the app domain is unloaded and there are connections in the pool, those connections don't get a chance to have the Close message sent to backend server. So, even when you are correctly closing your connection, you still will see those messages. You can prevent that from appearing by disabling the connection pooling. Which I advise against. Also, you may want to check if your domain app is being unloaded for some reason. I never see any harm being done because the connection exited without sending the close message. So I think you could (safely?) ignore this message on this case. I hope it helps. On Tue, Mar 8, 2011 at 05:50, rsmogura <rsmogura@softperience.eu> wrote: > On Tue, 08 Mar 2011 11:30:10 +0800, Craig Ringer wrote: >> >> On 08/03/11 02:49, Piotr Czekalski wrote: >> >>> I've checked and verified that all connections are closed within the >>> code, what's more, the problem has appeared just as I've moved server >>> from Fedora Linux x86_64 running Postgres 8.4.2 to the Windows and 9.0.3 >>> (details above) thus I conclude this is not a client problem indeed (the >>> failure didn't occure on Linux). >> >> Windows firewall? >> >> You can also see these error reports when the connections are closed >> uncleanly, without a proper backend close message. Perhaps you have >> client processes crashing? Or doing hard shutdowns where the client code >> doesn't get a chance to run any cleanup/dtors/etc? >> >> -- >> Craig Ringer > > I think in 8.x releases is no need to make any spacial close operation, or > at least You do not get notice, about it. Closing socket is enough. E.g. > JDBC driver closes connection, by closing socket. In 9 this changed and I > see many of such notices in log. Probably Your driver, as many others, is > written to close socket without backanad message. > > Regards, > Radek > > -- > Sent via pgsql-general mailing list (pgsql-general@postgresql.org) > To make changes to your subscription: > http://www.postgresql.org/mailpref/pgsql-general > -- Regards, Francisco Figueiredo Jr. Npgsql Lead Developer http://www.npgsql.org http://fxjr.blogspot.com http://twitter.com/franciscojunior