Re: Forcing lock release

Jonathan Leffler wrote: > > >From: trubisz@cs.sunysb.edu (Joe Trubisz) > >Date: 21 Jul 1997 14:42:56 GMT > >X-Informix-List-Id: <news.40674> > > > >Hi... > > > >INteresting problem: we had some ESQL/C code written which > >was not written 'friendly', and when it crashed, it left in > >the syslocks table a bunch of locks, which prevented subsequent > >users from doing work. > > > >Is there a way for the sysadm to issue a command for force > >the lock records to be deleted? > > The SysLocks table is in the SysMaster database (part of OnLine). > Consequently, the only way that locks can be left in the table is if the > OnLine system has not yet recognised that the client has died. > > I think that a shared memory connection is more vulnerable to a long delay > between application dying and OnLine detecting this than a network > connection is, but someone else who knows better may say otherwise. It is not often I would disagree with JL but it I have always found that network connections take significant longer to tidy up than SHM connections. I the app dies then generally it is not a problem, but if the client machine dies, the router goes down etc, then the network connections goes to either FIN_WAIT and/or FIN_WAIT2 (netstat -a). It can then take up to 15 minutes for the connection to time out. > > I don't know of a direct way of removing the locks. An indirect way might > be to use 'onmode -z <sid>' to kill a session ID (process ID if you're > using version 5.0x OnLine, though that can't be relevant since SysMaster > was not available until version 6.0x). > > Yours, > Jonathan Leffler (johnl@informix.com) #include <disclaimer.h> > PS: Warning I do not reply to messages with anti-spam in the return path. -- Paul Watson WF Software Ltd. Tel: +44 121 440 3179 Fax: +44 121 440 3806 Email: paulw@wfsoftware.co.uk