|
Mr Dog posted:a language that lets you feel like you're doing something very clever without actually accomplishing anything useful in the process. Mr Dog posted:the first like 60 seconds that a backend is unavailable instead of immediately throwing a 503 because i just like how you've spent equal time defending the runtime performance of java and asking for workarounds for your java runtime issues
|
# ? Jun 4, 2014 18:24 |
|
|
# ? Jun 11, 2024 19:33 |
|
Mr Dog posted:One thing I'd appreciate is if Hotstop used mmap and SHM to pool all the decompressed/JITted .class files it loads into memory. That way we wouldn't need these stupid rear end "application containers". That might be a fun patch to contribute upstream actually but I'd imagine that actually implementing this would open up a whole Pandora's Box of poo poo. tomcat is fine.
|
# ? Jun 4, 2014 18:25 |
|
tef posted:i just like how you've spent equal time defending the runtime performance of java and asking for workarounds for your java runtime issues how is using a lovely load balancer a java issue do you have some miracle application server that doesn't need good LB because it boots instantly?
|
# ? Jun 4, 2014 18:32 |
|
Notorious b.s.d. posted:how is using a lovely load balancer a java issue i didn't realise waiting for a minute for something to start was industry best practice. i know rails takes about that long but i thought you java people only did the slow start for desktop apps
|
# ? Jun 4, 2014 18:40 |
|
Shaggar posted:tomcat is fine. jetty is easier to configure and deploy edit: it is especially easier on the ops side. a jar w/ embedded jetty is just a regular unix process that gets started and stopped. there's no separate deployment step. push binaries. restart service Notorious b.s.d. fucked around with this message at 18:48 on Jun 4, 2014 |
# ? Jun 4, 2014 18:45 |
|
tef posted:i didn't realise waiting for a minute for something to start was industry best practice. i assume a full minute was hyperbole but yeah. poo poo takes a long loving time to recycle. it's not just rails and java servlets. django boot. .net app pool recycles. all kinds of poo poo takes time computers get faster so our software can get slower
|
# ? Jun 4, 2014 18:46 |
|
tef posted:i didn't realise waiting for a minute for something to start was industry best practice. the client VM starts up faster, it has Class Data Sharing
|
# ? Jun 4, 2014 18:55 |
|
hepatizon posted:wait what? i don't understand how it could possibly be easier "merge from upstream" button instead of "create a pull request, figure out whether the left or the right drop down should be the original, click merge" i mean i guess it's not as bad as it used to be, you don't have to do it all locally. but it's a p common task to want to merge upstream changes, and having to gently caress around manually creating a pull request and then accepting it is a bit tedious
|
# ? Jun 4, 2014 18:55 |
|
christ are you two being wilfully dense or something e: tef and bsd
|
# ? Jun 4, 2014 18:57 |
|
Mr Dog posted:christ are you two being wilfully dense or something BSD yes, tef no.
|
# ? Jun 4, 2014 18:58 |
|
pseudorandom name posted:the client VM starts up faster, it has Class Data Sharing i thought the client vm was essentially dead e.g. it doesn't even compile on linux/x86_64
|
# ? Jun 4, 2014 19:01 |
|
Mr Dog posted:christ are you two being wilfully dense or something apache is not exactly the world's most sophisticated load balancer. it will reverse proxy, maintain a connection pool, and do some very basic lb stuff it doesn't support health checks, so it has no magical way to know you recycled a node. you can either tune the timeouts, or not.
|
# ? Jun 4, 2014 19:03 |
|
Notorious b.s.d. posted:computers get faster so our software can get slower the java motto
|
# ? Jun 4, 2014 19:04 |
|
yeah but i'm not exactly operating loving Google Search here. it's a side project. the apache frontend handles reverse proxying, static content serving, request logging, and ssl just like a web server ought to, and the backend handles just the web ui stuff just like an application server is supposed to. for my purposes it's the simplest thing that could possibly work (i.e. i don't want to have to deal with logging, ssl termination, and static content poo poo in my web app, i let http's inherent composability take care of that) if this was web scale then I'd use a CDN for the static assets and a real reverse proxy in front of a pool of app tiers with a replicated db of some sort. but i don't, i have like, maybe 200 active users at most, i don't have to care about any of that poo poo. Restarting the back end takes a few seconds, not a minute, but I don't have control over the client. one of the two clients is an http rpc thing that will just retry its request three times rapid-fire and then shrug and just lose data if the third attempt fails.
|
# ? Jun 4, 2014 19:12 |
|
Otto Skorzeny posted:the java motto mgmt pays attention to how long it takes to shove stories out the door. last i checked "make the startup time acceptable to some guy on the internet" wasn't in the backlog Mr Dog posted:yeah but i'm not exactly operating loving Google Search here. it's a side project. the apache frontend handles reverse proxying, static content serving, request logging, and ssl just like a web server ought to, and the backend handles just the web ui stuff just like an application server is supposed to. for my purposes it's the simplest thing that could possibly work (i.e. i don't want to have to deal with logging, ssl termination, and static content poo poo in my web app, i let http's inherent composability take care of that) can you make apache fall back to a backend that just logs the request, then replay the log once the app server is restarted
|
# ? Jun 4, 2014 19:16 |
|
Mr Dog posted:apache frontend handles reverse proxying, static content serving, request logging, and ssl just like a web server ought to, and the backend handles just the web ui stuff just like an application server is supposed to. for my purposes it's the simplest thing that could possibly work (i.e. i don't want to have to deal with logging, ssl termination, and static content poo poo in my web app, i let http's inherent composability take care of that) once again mr dog I understand some of the words you are saying is there like an oreilly pocket guide for modern web architectures or something? I dont web but it's interesting
|
# ? Jun 4, 2014 19:18 |
|
Mr Dog posted:Restarting the back end takes a few seconds, not a minute, but I don't have control over the client. one of the two clients is an http rpc thing that will just retry its request three times rapid-fire and then shrug and just lose data if the third attempt fails. 1. set the timeouts super short 2. use a connection pool 3. run two copies of the server 4. recycle them one at a time
|
# ? Jun 4, 2014 19:21 |
|
Suspicious Dish posted:BSD yes, tef no. you pop up so often with nothing to say that i am starting to think you just stalk my posts
|
# ? Jun 4, 2014 19:21 |
|
Mr Dog posted:Restarting the back end takes a few seconds, not a minute, but I don't have control over the client. oh, ok. that's a bit annoying.
|
# ? Jun 4, 2014 19:22 |
|
Notorious b.s.d. posted:you pop up so often with nothing to say that i am starting to think you just stalk my posts I stalk Mr Dog's posts actually.
|
# ? Jun 4, 2014 19:24 |
|
Internaut! posted:once again mr dog I understand some of the words you are saying don't ask me i'm like 10 years out of date with this poo poo, although most of the new stuff revolves around new and exciting ways to dildo your own rear end using javascript on the client side (and server side via node.js if ur dumb) A reverse proxy is an HTTP server that forwards requests from the open internet to another HTTP server (usually one embedded inside a web application process), typically one inside your firewall or on localhost. They often perform some sort of load balancing. In my example I use one to handle all the gunk that isn't related to actually turning a request into a set of database queries and response HTML: logging the request, handling the SSL, that sort of thing. A CDN is a content distribution network. it is a paid service that hosts static files for you via HTTP. Their bandwidth and storage costs tend to be much smaller than yours due to economies of scale. Akamai is a popular one, and Amazon S3 can also be used as a CDN. Suspicious Dish posted:I stalk Mr Dog's posts actually. Sapozhnik fucked around with this message at 21:08 on Jun 4, 2014 |
# ? Jun 4, 2014 21:01 |
|
Kevin Mitnick P.E. posted:mgmt pays attention to how long it takes to shove stories out the door. last i checked "make the startup time acceptable to some guy on the internet" wasn't in the backlog Yeah, gently caress they guys who actually run it. They can all stay late for a 20 hour shift to bounce the cluster right?
|
# ? Jun 5, 2014 13:15 |
|
p much yeah if they don't like it they can find a new job i'm paying those slackers too much anyway rule 36 bitch
|
# ? Jun 5, 2014 14:36 |
|
Zombywuf posted:Yeah, gently caress they guys who actually run it. They can all stay late for a 20 hour shift to bounce the cluster right? well, yes. if it's a BFD then they can talk to my boss and it'll get done pretty loving fast. ops comes to us with requests often enough that we have a specific story type for them, and our turnaround on ops asks is 2-4 weeks aka 1 release cycle my understanding is that it's not a high priority for ops because the NOC handles stuff like that and those guys are 24/7 already
|
# ? Jun 5, 2014 14:44 |
|
Kevin Mitnick P.E. posted:well, yes. if it's a BFD then they can talk to my boss and it'll get done pretty loving fast. ops comes to us with requests often enough that we have a specific story type for them, and our turnaround on ops asks is 2-4 weeks aka 1 release cycle So what you're saying is your lovely software^Wstories need an army to run?
|
# ? Jun 5, 2014 14:49 |
|
so has anyone else played with Nimrod yet? (http://nimrod-lang.org/) it gets compiled to C, has optional garbage collection (and two different implementations), and extremely powerful macro functionality that operates on the AST at compile time. i've been having fun with it
|
# ? Jun 5, 2014 15:18 |
|
Kevin Mitnick P.E. posted:mgmt pays attention to how long it takes to shove stories out the door. Kevin Mitnick P.E. posted:ops comes to us with requests often enough that we have a specific story type for them I'm not familiar with this usage of the word "story" is this from some hot new development/PM methodology I should be aware of
|
# ? Jun 5, 2014 16:01 |
|
Internaut! posted:I'm not familiar with this usage of the word "story" story as used in "user story". probably from agile or scrum. welcome to 2004, when this poo poo swept the industry the first time. i'm only half being snarky, you should really look this poo poo up. at least browse wikipedia. it is mostly snake oil but people expect you to know this and look at you like a space alien if you don't understand the kanban tracking the 6-month sprint for their agile waterfall
|
# ? Jun 5, 2014 16:29 |
|
Internaut! posted:I'm not familiar with this usage of the word "story" here it means work item p. much
|
# ? Jun 5, 2014 17:08 |
|
Notorious b.s.d. posted:it is mostly snake oil but people expect you to know this and look at you like a space alien if you don't understand the kanban tracking the 6-month sprint for their agile waterfall eh probably not in my industry but yeah I'll check it out
|
# ? Jun 5, 2014 18:47 |
|
Shinku ABOOKEN posted:i am employed therefore i use windows i haven't had to deploy to windows in over six years, eh, it's a living. quote:if i want to use the psycopg2 binary installer i need a vanilla python installation. if i use a vanilla python i need to build the qt4 applications and packages myself. gently caress me welcome to programming, also welcome to using windows and not using the microsoft ecosystem fwiw you could probably run two installs of python that talk to each other BUT THEN YOU'D HAVE TWO PROBLEMS HA HA HA HA HA HA
|
# ? Jun 6, 2014 00:44 |
|
if you're smart enough to use windows you're probably smart enough to not use python
|
# ? Jun 6, 2014 02:17 |
|
ironpython lol
|
# ? Jun 6, 2014 02:54 |
i use scala
|
|
# ? Jun 6, 2014 02:55 |
|
use jython, get access to all the great java libraries from the plang you love. qt is bad: write your GUI in swing.
|
# ? Jun 6, 2014 08:34 |
|
Internaut! posted:eh probably not in my industry but yeah I'll check it out a story: a feature request written like software fanfic. kanban: we have a number of post-it notes and we move them from left to right agile: we have no long term plans and you need to keep up with our improvisation (compare: waterfall, we only have to think everything out in advance before doing it) sprint (theory): like a tiny release cycle, where you start with planning, code for a week or so, and then maybe a breakdown at the end burndown chart: a chart that goes down in number to show progress sprint (practice): whatever management has said to do today, we're agile you know scrum™ - we paid for an agile consultant and someone on our team has a certificate too continuous integration - we have a build server continuous deployment - the cron job to build things does make && make install
|
# ? Jun 6, 2014 09:39 |
|
now do personas
|
# ? Jun 6, 2014 11:06 |
|
tef posted:a story: a feature request written like software fanfic.
|
# ? Jun 6, 2014 12:49 |
|
tef posted:i haven't had to deploy to windows in over six years, eh, it's a living. I read your post in ridley's voice.
|
# ? Jun 6, 2014 14:22 |
|
|
# ? Jun 11, 2024 19:33 |
|
tef posted:a story: a feature request written like software fanfic.
|
# ? Jun 6, 2014 14:25 |