On 10/29/07, Keisuke MORI <[EMAIL PROTECTED]> wrote: > "Andrew Beekhof" <[EMAIL PROTECTED]> writes: > >> I think that the old behavior is preferable because running > >> "a part of the group" is pointless from the service > >> availability's point of view and confusing to users. > > > > no. > > just because items later in the group fail doesn't mean the rest of > > the group should be stopped. > > In the HA database cluster, the database service is typically provided > by the group like: > Filesystem + MySQL + IP > > If any of the resources failed then the database service is no > longer available. Running only Filesystem does not mean anything to > "the service availability."
Sure it does - just not at the level you're talking about. Would you also stop everything if the IP couldn't run? Because thats the behavior what you're asking for. But more importantly... what, apart from your personal taste, is the problem with having it running? If its truly not being used at all then there is hardly likely to be a performance impact (especially for a filesystem) and stopping it doesn't reduce the probability of cluster-induced corruption. So what's the point? Seriously. > > consider: > > IP + Filesystem + Apache + MySQL > > > > Just because MySQL fails doesn't mean Apache, the Filesystem nor the > > IP should be stopped. > > I can understand that, but in that case, > I think it would be more straightforward to have > two separate groups; one for the database server and the other > for the web server, because they can run independently, right? Not if part of the website needs the database. Probably the better example is if you swap the database and web server. You'd not want the database to shut down just because the webserver wasn't available. > We usually group up resources because they need to run together to > provide "the service" (database, web server, or whatever) as a whole, > therefore running a part of the group does not make sense. > > > > > >> 2) If it would not be possible, then would you tell me > >> what is the "correct" configuration to achieve the same > >> result as 2.1.2 in the new version? > >> (with "correct" I mean "by design" and "unlikely change in > >> the near future") > >> > >> I'm also wondering how anybody else configures about this behavior. > > > > Let me instead ask what you believe you gain by stopping the first resource. > > Because it is just simple and intuitive for users. I don't buy this at all. > And I believe that the most of commercial HA software would also > behave like this (at least in the typical usage). Why? You've still not produced a reason other than "I'd prefer it that way". Which is frankly not enough motivation for me to spend time on it. The design goal of a high availability cluster is to keep as much running as possible. What you're asking for is to stop a resource that has no need to stop. > Our costomers are considering to migrate from a commercial HA > software to heartbeat, and all of them expect to behave like > this so far. Does the fact that Ctrl-S saves a file in MSWord and OpenOffice make Emacs' decision to use Ctrl-x Ctrl-s a bug? Clearly not. Different software can and will behave differently - this is not unexpected. > At least it would be nice if it's able to be customized, I would think. If you write a patch, I'll consider it. _______________________________________________________ Linux-HA-Dev: [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev Home Page: http://linux-ha.org/
