Re: Data Loss in Online 5.03 UC5??

J. David Schronce, Sage of sages, spontaneously contributes: > We are experiencing random data loss. Parts of things that were there > for several days/weeks/months are simply gone, leaving the remainder > untouched. It's happening across a broad spectrum of applications (RDS > 4.11) so I tend to reject the "You put something in your code to delete > it" theory. > If anyone has experienced similar problems - and RESOLVED them - or has > knowledge of a known Informix bug that relates I would sincerely > appreciate your feedback. Hmmm...data loss. Gotta love it. I, personally am not aware of OnLine ever losing data, save for perhaps those times we've managed to send the engine into the spiral death spin of eternal recovery and had to have Tech Support get us back up, and then that's more of an issue of a partially committed transaction. I have seen a system go terminally sick when something went stomping on the data pages. Looking at a hex dump of the spaces showed the ugly foot prints of something just blasting everything wantonly. I guess, technically, OnLine COULD have done that damage. It was a raw chunk, and only OnLine really had access to it, but because of the nature of the damage, I'm more inclined to blame cosmic rays. In either case, we knew immediately that it had happened because OnLine choked and died. It's also never happened again. It is fair to suspect OnLine for your woes, because only OnLine can selectivly remove data and leave the rest of the system intact. However, when OnLine does those things, it tends to do them in really arcane circumstances and I would think it wouldn't be wide spread. I'm going to assume that you've run through the gauntlet of assorted tbchecks to verify that OnLine is convinced that everything is hunky dory. "Data? What data? The pages are clean, the b-trees balanced, they didn't need those extra leaf nodes anyway, so that's my story and I'm sticking to it." Of course, the next thing to think about is what may have changed recently? Is it a single occurence, or is it happening consistently? When confronted with "things that can't happen", I tend to add INSERT, UPDATE, and DELETE triggers to watch the traffic and let them sit as silent sentries until someone comes up with "How come this has happened? Nobody changed anything, we didn't do anything, it's not our fault". Then I have this nice history to go back to to help solve the problem. Sometimes it is the app, sometimes it is the people. OnLine has rarely let me down, and it tends to be the last thing I suspect when there is some kind of problem, especially when it is nice and neat like missing data. System problems tend to be random and chaotic, and they rarely have the time to properly update the indexes etc or honor page header boundries and such. I find users to be much more capable at manifesting problems with data. (I include programmers and dbas in that 'users', and I've been guilty more than I'd care to say.) So, anyway, that's just some ideas that you can use. Perhaps there is a known bug. But, I'd seriously consider other options first. And I don't even WORK for Informix :-). Good Luck! Will Hartung (villy@collinscomp.com)