From: JT S. <jt...@pl...> - 2004-07-11 17:38:04
|
The reason I can do this from an application environment is 2 fold. 1) The interval that the slaves use to update from the master is determined largely by the configuration of the slaves to their master. 2) As a developer, you control whether your app is reading directly from the master or from the slave, and presumably you know what your app needs to do to work properly. From a content viewing point of view, yes the data can be out of date for a little bit, but that's the whole purpose of caching in a content environment. Let the data go out of date for a little bit in order to not stress out the machine. If you're in an environment where cachable reads need to not be cached, then instead of buying 2 small database servers and doing replication, buy one motherhog database server and serve all requests from there. I'm not trying to be a prick, shun anyone's ideas, or anything else dickheadish. It's just that: 1) This isn't an easy problem to solve in any way that doesn't cause other problems. and 2) I don't think that the amount of time required to solve this particular problem is a good investment given that it's not even a problem yet. On Sun, 11 Jul 2004 12:35:24 -0400 "Jay R. Ashworth" <jr...@ba...> wrote: >On Sat, Jul 10, 2004 at 11:05:51PM -0500, JT Smith wrote: >> >I hate to sound like a pissant, but that doesn't help precisely the >> >audience who care: those people who might, in fact, need to use it. >> > >> >So I guess the question is, how much impact will it have on them? >> >> >> 1) I have no idea what you're even asking. >> >> 2) You sound angry, and if you are, why? > >No, I wasn't. > >The issue, as I understand it, is this: if I have multiple WebGUI >servers, running from replicated databases, one of which is defined as >the 'master' instance, how do we handle reads of data which is -- >because we're in admin mode -- *required* to be current. > >My recommendation originally was to allow the read to be to whatever >instance was local, and force a cache flush at that point. > >You, if I understood you correctly (and vice versa), said that wasn't >efficient enough, and that it was better to short circuit the read in >the code directly to the master. > >Now, as I recap things, I wonder how you *could*, since they might be >on different machines with no cross-connects, but perhaps I've >misunderstood the issue completely. > >But in any event, if you *don't* force cache flushes (by which, in this >case, I actually mean forcing an update between the replicated >databases -- but from 30,000 feet, you can treat those replicated >databases as a 'local cache', I think), then the slaves will be giving >out bad data... > >and since your stated target for WebGUI is as an applications >development environment -- notwithstanding that the only application >being developed by a large part of it's userbase is "my website" :-) -- >then it seems to me that you can't just brush off the question he >asked... which is what it seemed to *me* like you were doing. > >But I might be wrong. :-) > >Cheers, >-- jra >-- >Jay R. Ashworth jr...@ba... >Designer Baylink RFC 2100 >Ashworth & Associates The Things I Think '87 e24 >St Petersburg FL USA http://baylink.pitas.com +1 727 647 1274 > > "You know: I'm a fan of photosynthesis as much as the next guy, > but if God merely wanted us to smell the flowers, he wouldn't > have invented a 3GHz microprocessor and a 3D graphics board." > -- Luke Girardi > > >------------------------------------------------------- >This SF.Net email sponsored by Black Hat Briefings & Training. >Attend Black Hat Briefings & Training, Las Vegas July 24-29 - >digital self defense, top technical experts, no vendor pitches, >unmatched networking opportunities. Visit www.blackhat.com >_______________________________________________ >Pbwebgui-development mailing list >Pbw...@li... >https://lists.sourceforge.net/lists/listinfo/pbwebgui-development JT ~ Plain Black Create like a god, command like a king, work like a slave. |