Hi alexander, thanks for your response. We have not yet enabled a connection timeout so the reconnect might help us. we will play around with that.
The problem with orion is that there are settings for pooling configuration but they don't work reliably (checked that by wrapping tracing code around the jdbc driver to see which connection is used for how long) and the product is essentially dead since it was bought by oracle. so as long as it takes for us to migrate to another appserver, we have to solve the problem either at the database or jdbc driver level. I have sent knldiag for several crashes to the list (or urls where they can be downloaded together with our linux kernel parameters) but got no response to recent mails. I'm happy about anyone who looks into them. e.g. http://listserv.sap.com/pipermail/sapdb.general/2002-October/014761.html if you (or anyone else looking into that) require(s) more information I'll provide it. we still have crashes about every 3 days on our production system and no real clue what's going on. thanks in advance, robert On Tuesday 29 October 2002 16:13, Schroeder, Alexander wrote: > Hello Robert, > > indeed, the JDBC driver does such a thing *by default*, but > only if you are not inside a transaction (as here a > re-connect would result in weird things, as the timeout at > the database does an implicit rollback). > > Because you ask of this feature, I assume that automatism > did not work, so you are possibly out of luck with that idea, > cause your application does some transactional stuff. > > You may look into the Orion appserver itself, whether there > is some mechanism for keeping a connection only a limited > time - most application servers have some option like this. > For the use with application servers, it is generally not > the usual practice to use a connection that times out at > the database. > > Coming back to your problem - perhaps you can send more details > on the crash itself, such as the contents of knldiag after the > crash, such that it is perhaps possible to check what is happening. > > Regards > > Alexander Schr�der > SAP Labs Berlin > > > -----Original Message----- > > From: Robert Kr�ger [mailto:krueger@;signal7.de] > > Sent: Tuesday, October 29, 2002 3:51 PM > > To: [EMAIL PROTECTED] > > Subject: JDBC driver reconnect question > > > > > > > > Hi, > > > > we still have a problem with the database crashing with > > malloc failures > > (seemingly runnig out of memory) after running for a few > > days. We have > > investigated some more (after having exchanged operating > > system version and > > some hardware components) and noticed that the connection > > pooling algorithms > > of our application server (orion) lead to connections being > > used for a long > > time. We somehow suspect a memory leak in SAPDB related to > > connections being > > used for a long time (BTW. is there a way to check that?). > > > > Does the JDBC driver support transparent reconnects of timed > > out connections? > > If so, we could simply set a very low connection timeout and > > rely on the JDBC > > driver to reconnect? Is this possible? > > > > Thanks in advance, > > > > Robert > > > > > > _______________________________________________ > > sapdb.general mailing list > > [EMAIL PROTECTED] > > http://listserv.sap.com/mailman/listinfo/sapdb.general _______________________________________________ sapdb.general mailing list [EMAIL PROTECTED] http://listserv.sap.com/mailman/listinfo/sapdb.general
