there is no way to get a proper StackTrace from an OoM because of a memory leak, which is obviously the case, after I've read your " But then it happened for much smaller files too (~2MB) and I started to suspect it was a case of a)." comment, Jake.
Just to have asked: You are not using LS2J, because this is very prone to Memory Leaks, when used in a server agent.
Its a good idea recycle EmbeddedObject and MIMEEnties. Maybe also Documents. If you recycle something which is supposed to be recycled by the Java code that calls the Domino Agent internally you get an error message in log.nsf.
I've done a lot with Webservice-client kind of scenarios with Domino Agents some years ago and never run into memory leaks, though I paranoically expected them to pop up.
Memory leaks are really one of the nastiest errors that exist, mostly because you don't get an immediate feedback. In a league with weird JSF/JSP lifecycle madness, threads that block each other or concurrent write-lock stuff with rdbms.
Gents,
there is no way to get a proper StackTrace from an OoM because of a memory leak, which is obviously the case, after I've read your " But then it happened for much smaller files too (~2MB) and I started to suspect it was a case of a)." comment, Jake.
Just to have asked: You are not using LS2J, because this is very prone to Memory Leaks, when used in a server agent.
Its a good idea recycle EmbeddedObject and MIMEEnties. Maybe also Documents. If you recycle something which is supposed to be recycled by the Java code that calls the Domino Agent internally you get an error message in log.nsf.
I've done a lot with Webservice-client kind of scenarios with Domino Agents some years ago and never run into memory leaks, though I paranoically expected them to pop up.
Memory leaks are really one of the nastiest errors that exist, mostly because you don't get an immediate feedback. In a league with weird JSF/JSP lifecycle madness, threads that block each other or concurrent write-lock stuff with rdbms.
wish you all the best
Axel