This doesn't change the fact that external monitoring is still going to be better.
On Wed, Jun 6, 2012 at 1:27 AM, Jeff Willden <jeff.will...@gmail.com> wrote: > Some readers are not getting the point. When running node in a > production environment, it is assumed that the app has no blatant > infinite loops. Yes, that would block the event loop and prevent > anything from running, but if you've done that, you should take Intro > to Node 101, or maybe the remedial class... > > In a production environment, we're concerned about long running events > that might be blocking the event loop. If your code is doing database > calls, then it's yielding to the event loop. No problem there. But > then if you run a bunch of processing on that using synchronous code, > and that processing takes longer than whatever threshold you give in > the code I posted, then it will tell you. We're not concerned as much > about overall request time, because during the time it's processing > the request, it's probably yielding. For newbies, yielding means that > you perform some action and pass it a callback, and when that action > is done, it calls the callback. Users see this all over in node code. > That doesn't necessarily mean it yields though... > > Case scenario: we had a bunch of db calls to load data that would be > processed (sorted, organized, transformed, etc) and converted to json > and then fed to the client. That db calls yield each time a db call is > made. No problem. The processing was all linear programming. Problem: > for large data sets it was taking over our threshold. When you have > more concurrent requests than the number of instances of node (we're > using cluster) then every instance is busy and your server goes > unresponsive. BAD news. We had to change some calls to accept a > callback, and then it would call the callback inside a > process.nextTick(). That made the code async, and the problem > disappeared. We could ramp up with hundreds of users and the users > with only a little data got served fast, the users with lots of data > not quite as fast, but at least they weren't blocking other users. > > People often misunderstand how to build server-side apps with Node. > With async coding in node, you can easily outmatch performance for > lots of other environments. With synchronous code in node, you can > easily have a dead server. Node is great, but it's oh-so-easy to shoot > yourself in the foot. Gotta be really careful about excessive runs of > synchronous code. > > -- > 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 nodejs@googlegroups.com > To unsubscribe from this group, send email to > nodejs+unsubscr...@googlegroups.com > For more options, visit this group at > http://groups.google.com/group/nodejs?hl=en?hl=en > -- 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 nodejs@googlegroups.com To unsubscribe from this group, send email to nodejs+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/nodejs?hl=en?hl=en
