26.09.2016 13:02, Jiří Činčura wrote: > Hi *, > > I've had quite a few occurrences of transaction being stuck on Firebird > 2.5.6 in last 7 days. Even trying to close the connection from > monitoring tables was not working. I had to kill the process. > > I was able to take a dump (actually 2, from different processes), but > given it's a production system with real data, I would rather not upload > it publicly. It's 2.5.6.27020, 64bit, Classic. It happened in 2.5.5 as > well. Bellow are the call stacks of threads in fb_inet_server (with PDB > for Firebird loaded). > > I suppose the first thread here is the key. Looks like it's spinning in > "while (owner->own_ast_count)" forever (at least couple of hours before > I found out, because of a lot of deadlock timeouts in SQL). > > Anybody interested in the dumps?
Yes, send it to me. But better is to have reproducible example, or, at least, memory dump before attachments are killed - i.e. when transaction stuck is detected. Regards, Vlad PS Stack shows that EXECUTE STATEMENT is used, probably with autonomous transactions - it could block itself if used not correctly (for example: update record and then update it again in autonomus tx with "wait" option) ------------------------------------------------------------------------------ Firebird-Devel mailing list, web interface at https://lists.sourceforge.net/lists/listinfo/firebird-devel
