Re: Closing OnLine session (2)

}From: rferdy@kuwait.net (Rudy Fernandes) }Date: Wed, 14 May 97 10:46:54 GMT }X-Informix-List-Id: <news.37888> } }This is a follow up on the original thread related to closing }dormant OL connections. I've attempted using some of the }suggestions posted, but they don't work - I think this is }because we only have RDS (no ESQL/C or C-compiler and thus, }no access to the SQLEXIT() function). If you don't have a C compiler, then you do have problems. I'd suggest GCC, but you will need to get a binary of it to get started (you can't compile GCC from source if you don't have a C compiler to start with!) If you get over that, then even without an ESQL/C licence, you could build a custom runner and add an entry to call the i4gl_sqlexit() function I sent previously. The sqlexit() function is built into the runner (or, at least, the libraries provided to allow you to build a custom runner), so you can call it OK. }I am trying another approach which would be ridiculous as a }generalised solution. However, it could just possibly be worth thinking }about in sites which have, like mine, recently }upgraded to v6 tools and v7 OL. That is, they still have the v4 tools }and the v5 SE versions available. } }The logic : }Under SE, the database connnection runs as a unix process (sqlexec). }If it lies dormant, it gets swapped out and its memory utilization }goes to 0. This applies to any program which is waiting for a long time for an event to occur, such as its child process to terminate. }The Plan : }My menu programs and database (which are the dormant processes that }I have targeted) will be under SE, while the }'action' programs which are activated (RUN 'fglgo ...') from }the menu program will run under OL. Maybe you'd be better off using the 7.x OnLine; you'd then have 1 dormant program (the menu program) and one dormant thread in the engine, instead of two dormant programs (the menu program and the sqlexec). This should reduce the total memory requirement on the system -- all other things being equal, which they aren't. }I've have commenced implementing this starting today and I }will be monitoring the daily statistics that we collect to }evaluate the effect of this. I will get back to the group }with the conclusions I draw. } }However, I'd be interested to hear comments on the theory }behind this approach. Yours, Jonathan Leffler (johnl@informix.com) #include <disclaimer.h>