>Actually, haibu is what runs everyone's apps on nodejitsu! We have a few closed source extensions for haibu, but aside from that it's all the same. In fact, I encourage you to use haibu as a starting point for something like raft. Well haibu was the first project that i found to be like mine is. haibu has heavily influenced raft, so has heroku and nodester.
>We have a similar concept, called a drone ( https://github.com/nodejitsu/haibu-carapace). Each haibu server ( https://github.com/nodejitsu/haibu) can spawn n drones running whatever (multiple apps, or multiple copies of the same app). Our balancers can then proxy to these drones however we please. The drones concept is very slimier to what raft has. I just thought there might have been more to it. >I can tell you right now that it's non-trivial. I doubt a single person could realistically afford to run 3000+ apps out-of-pocket, at least without seriously gimping each app's resources. I dont see myself running 3000+ app out of pocket. You make a good point about how much load each app would have. Looking at nodester they seem to limit each app to 25mb of ram and if it goes over then it kills the app and re-spawns it. I dont really want to go that route, how do you guys monitor each app? Do you do it inside the app or from out-side the app? do you ever kill apps or been resouce hogs? >From the second post >They need 3000+ IP addresses which are a limited resource and you have to explain your demand to RIPE. The way raft is setup you only need a few external ip's. These ip's would be pointing at the load-balancer. So 3000+ apps would not be a problem as they would be local ip's >Nodejitsu uses jitsu+http because it allows us to use npm for app bundling. I have followed a very slimier approach to you guys. Do you guys feel that it makes user less likely to use the platform because git is not used? >but imo you should always be able to start a single instance of an app without clustering by running, say, `node app.js`. I agree with you here. I have given it some thought and was thinking of doing something like "raft-cli run" and this would allow the user to run the app locally. The whole node app.js this is far fetched for raft as it runs each file.js in its own context. I need to do a lot more thinking about this problem. On Saturday, August 11, 2012 5:33:21 PM UTC-4, Tim Dickinson wrote: > > Hey all. > > So this is not a ANN but more of an request for advice from nodester / > nodejitsu / haibu and the community in general. What i have been working on > for the past few months is of sort a PaaS. The basic idea behind it is to > create a server to can spawn node apps that are pushed out to it with a cli. > > I'm calling it Raft as in a boat to float apps on. It has gone through a > few iterations since it creation. It started out as a MVC style app > container. The basic app structure was you would have your model's, view's > and controller's, and raft would load all these into the app, kinda so > you didn't have to code and express server or is server or what have you. > As i worked on raft and played around with it more i found that > the MVC style was much less dynamic then i would have liked. So from that > the current version has evolved. > > The current version... OK the current version is now very low leave, in > fact it does not do much other then load the app and its module in a > context with its own process. the only different between the raft context > and plain nodejs context is that you get a global called raft. what the > raft object does is gives you http, tcp, express server and so on. these > servers are just like the native server but for one difference and that > been httpserver.listen, the native httpserver.listen take a port and host, > but what the raft httpserver.listen take is a string that is a domain that > gets routed to the port of that app. > > OK so like i say this is not an announcement but more a request for advice. > > Some of the questions: > > I know that nodester is hosted on AWS's and with that how many apps are > hosted per server? Are the server like a 8GB/RAM 4/core pc running maybe 20 > apps per server, or is is a micro server running just one app? > > Nodejitsu are the developer of haibu, but i dont think that is what they > are using for the PaaS. Now on that is their backend a custom build > of haibu or is it a whole new module in its own? > > To nodester. why did you chose git for the pushing of apps to the backend? > > To nodester / nodejitsu. Have you guys thought of a kind of dynos (heroku > style)? if so how would you guys go about doing that? like you spawn 2 > processes of the same app and just route request to each app like node does > with the cluster module? > > To nodester. On average what are your costs running 3000+ app on AWS's? > > OK so this is what im going to ask for now. I do have more question but i > would like to see if i get any answers for these ones. > > Gota love node! > > The code for now. Please note that this is not a release but a Q&A > https://npmjs.org/package/raft > https://github.com/FLYBYME/Raft > > > Thanks all > Tim > -- Job Board: http://jobs.nodejs.org/ Posting guidelines: https://github.com/joyent/node/wiki/Mailing-List-Posting-Guidelines You received this message because you are subscribed to the Google Groups "nodejs" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/nodejs?hl=en?hl=en
