Re: Logical log backups to non-rewind tape device

On Fri, 13 Jun 1997, David Williams wrote: > In article <Pine.BSF.3.91.970611092019.18030D-100000@kilimanjaro.dallas. > herald.net>, Richard Stanford <richards@herald.net.REMOVE> writes [ comment: Gee, that's a long 'writes' line ] > >The one (1) time we had a catastrophic hardware failure, we ran into so > >many Informix bugs restoring log tapes it wasn't even funny. Combine > >that with the pletora of problems involved in running continuous-logging > >with non-technical operators (for the most part), and I'm beginning to > >consider recommending dropping continuous logging from our sites. Any > >comments out there? Has anyone else reached the same conclusion? > I have sucessfully rolled forward several instance after using > continuous logging. I would use it anyway as without it you have > nothing. With version of online are you running? Either 5.05 or 7.2[23]. HP and IBM mainly, with a few scattered others (Sun, DG, NCR, etc). > Make sure it is 5.08,7.14 or 7.22/3 which even if the latest for > your platform. Early 7.x releases are very buggy. We've found the "known bug" in 5.05 that can corrupt archives. Why Informix won't send out alerts on things like this... Aargh. But that's another rant entirely. Saying "run the latest version" is all well and good, but doesn't work for us in many situations. Let me explain a little: We (the company I'm with (Dallas Systems)) develop and maintain warehouse management systems, generally requiring 24x7 availability. Our customer base, with a few exceptions (Heya, Bruce) is non-technical, at least in the RDBMS/UNIX arena. Our product undergoes several months of internal testing, followed by a long period of customer testing, before installation. This certification process occurrs on a specific tool/engine/code release. - - - If we upgrade an engine, even if a tool upgrade is not needed, we will re-certify our application and procedures. This is no small endevour, and we'll only do it if we're pretty certain its needed. If Informix would document /every/ behaviour change in the release notes, no matter how inconsequential, in easy-to-read language things might be different. Even then, we'd probably still re-certify, but without that level of documentation its not even an issue. Once we've re-certified, our customer will probably want it set up in an independant instance/machine for them to test. This happens with any change, tools/engine/code. After all, if the system goes down, they're dead in the water (and losing money continuously). Then, and only then, will they deploy. This is oversimplified, and there are occasional exceptions, but its accurate for the most part. You may think that a full test of all procedures isn't warranted for a 5.05 -> 5.08 engine change, but think: We were on 5.01. Upgraded to 5.05 to correct the negative npused glitch (primarily, there were other reasons). The main 5.x archival bug was introduced in 5.03. Had we been able to test this properly the way we do now (ie: restore /after/ a full simulated day of activity is in the logs, instead of just restoring a level-0) we'd have caught this problem. There are other issues, of course, but in most cases both us and our customers (remember, non-technical) have to be convinced that there is a serious production issue in order to upgrade an engine. The archive issue will probably be such a one (although we seem to be having other problems with 5.08), but these are rare. Sorry to rant a little, its just that the "Oh, just upgrade" answer that seems to be post-it-note'd to every Informix tech's monitor gets old after a while, and I guess you pressed a button. Please don't take it personally :) -Richard