|
Shaggar posted:thats part of tomcat and not tc-server specific tc-server's is much nicer than bog standard tomcat
|
# ? Jul 25, 2012 15:10 |
|
|
# ? Jun 3, 2024 21:38 |
|
what does it add?
|
# ? Jul 25, 2012 15:10 |
|
Shaggar posted:what does it add? http://www.springsource.com/developer/tcserver but the part i was referring to was their command line app. deploying a new war is basically selecting a war and hitting enter instead of copying the war to some staging directory on the same disk and doing a mv so you don't accidentally deploy a partial
|
# ? Jul 25, 2012 16:08 |
|
yeah that uses a web "service" thing thats built in to tomcat. you can do the same thing via the management servlet if you have it installed or via a command line tool that u can get from apache. also theres a tomcat deployment pluging for maven that uses the same service i've never actually deployed a war directly to the webapps folder
|
# ? Jul 25, 2012 16:11 |
|
i don't think the tc server version uses the same mechanism as the built-in manager app, or the ant tasks, but i could be mistaken. either way it's nice, i hate the manager app and not all the apps i've deployed were mavenized
|
# ? Jul 25, 2012 16:14 |
|
now I never did any large-scale servlet stuff so forgive my ignorance but are y'all talkin about different tools to use to copy a file?
|
# ? Jul 25, 2012 16:14 |
|
Percona is fine, it's too bad there still isn't an equivalent of pgbouncer for MySQL that's as good though.
|
# ? Jul 25, 2012 16:15 |
|
it looks like theres some cool stuff with monitoring, but alot of the "differences" are things that arent configured by default in tomcat (ex: jmx) or UIs for config files.
|
# ? Jul 25, 2012 16:16 |
|
rotor posted:now I never did any large-scale servlet stuff so forgive my ignorance but are y'all talkin about different tools to use to copy a file? if you hot deploy, tomcat isn't smart enough to notice if a war file is not complete, so if you drop a file into the container in a non-atomic way, tomcat will attempt to explode it even tho it's not complete, so at least in a linux you have to copy the war to a place on the same disk, then `mv` the thing into webapps so that it's an atomic op or you can use one of many other ways, but essentially yes this is all about putting a file in a place
|
# ? Jul 25, 2012 16:18 |
|
rotor posted:now I never did any large-scale servlet stuff so forgive my ignorance but are y'all talkin about different tools to use to copy a file? if you just copy a war up to the webapps folder theres a chance that tomcat will try to unpack the war before you finish copying it which can lead to probs. thats where theres a deployment system build in where you push the war to a url and tomcat waits until its got the whole thing before unpacking. it can also respond with the results of the deployment and startup (ex: it unpacked but couldnt start cause ur missing a jar) so its alot nicer than a dumb file copy.
|
# ? Jul 25, 2012 16:19 |
|
on windows im pretty sure that the copy will maintain a write lock that tomcat cant violate though so ur probably safe doing it on windows. but then you also have to deal w/ certain other lock issues in windows when doing undeploy/redeploys but theres a tomcat param to fix that.
|
# ? Jul 25, 2012 16:20 |
|
Shaggar posted:it looks like theres some cool stuff with monitoring, but alot of the "differences" are things that arent configured by default in tomcat (ex: jmx) or UIs for config files. ya the instrumentation is good, use it with STS (hotrod eclipse) and spring insight and vfabric you can see all sorts of cool poo poo in realtime vfabric is god drat expensive tho
|
# ? Jul 25, 2012 16:20 |
|
GlassFish supremacy you plebs
|
# ? Jul 25, 2012 16:56 |
|
I can't get the tomcat admin app to work on my live server lol. and i don't particularly want to risk extended downtime by loving around with it right now. it's some class path poo poo (needs a special class loader to gain super admin powers and the option that says it enables that doesn't actually seem to enable it) so I just push things using SCP/mv like a big god drat idiot deployment over HTTPS to an admin web service would be p sw8 tho.
|
# ? Jul 25, 2012 16:57 |
|
Currently looking at maybe throwing ZeroMQ into the mix so that I can have a middle tier (i have two HTTP things, wanna add a middle tier to do caching and pub/sub between them). ZeroMQ looks really nice but the ~hilariously zany~ documentation is kind of annoying. "hello sir can i interest you in building your business on top of a middleware package that was written by a gigantic man child?" does anyone have any experience with it?
|
# ? Jul 27, 2012 12:22 |
|
why use zero and not rabbit? rabbit is good and erlang and poo poo
|
# ? Jul 27, 2012 15:20 |
|
trex eaterofcadrs posted:erlang and poo poo Q.E.D.
|
# ? Jul 27, 2012 23:18 |
|
is head first c any good? i don't really plan on learning c any time soon, but it's on sale and it might be a nice distraction from the semi-colon hell of javascript
|
# ? Jul 28, 2012 12:42 |
|
fidel sarcastro posted:is head first c any good? i don't really plan on learning c any time soon, but it's on sale and it might be a nice distraction from the semi-colon hell of javascript if you're itching to spend money and learn C get K&R http://www.amazon.com/C-Programming-Language-2nd-Edition/dp/0131103628/ref=sr_1_1?ie=UTF8&qid=1343489787&sr=8-1&keywords=k+r+c+programming
|
# ? Jul 28, 2012 16:36 |
|
trex eaterofcadrs posted:if you're itching to spend money and learn C get K&R even if you never use C in your life you will still learn useful things from this book
|
# ? Jul 28, 2012 18:08 |
|
5.3.0 Prior versions treated arrays as the string Array, thus returning a string length of 5 and emitting an E_NOTICE level error.
|
# ? Jul 28, 2012 21:34 |
|
CaptainMeatpants posted:5.3.0 Prior versions treated arrays as the string Array, thus returning a string length of 5 and emitting an E_NOTICE level error. haha
|
# ? Jul 28, 2012 21:44 |
|
learning programming languages is cool and awesome
|
# ? Jul 28, 2012 21:45 |
|
Mr Dog posted:Currently looking at maybe throwing ZeroMQ into the mix so that I can have a middle tier (i have two HTTP things, wanna add a middle tier to do caching and pub/sub between them). In the end I just went with ROUTER and DEALER sockets (I have no use for multicast or a pub/sub model in my app) and completely ignored the rest (I had no need for PUB/SUB), REQ/REP are nice in theory but the rules that govern what can talk to what (http://zguide.zeromq.org/page:all#Core-Messaging-Patterns) can seem constricting if you try to build a complex app using every bloody socket type possible. Further the behavior of ROUTER and DEALER with regard to how how they manipulate frames in a message to perform message routing are super simple. It's nice that ZeroMQ handles connection, reconnection on dropout, does it in an asynchronous manner and has internal buffer support. If you want guaranteed message delivery you have to bake that into your own protocol. They have a whole bunch of recipes provided for guaranteeing delivery, synchronisation between workers etc but once you get a feel for how ROUTER/DEALER work building a solution will come naturally (just don't feel the need to mix in the umpteen other socket variants just because you can). One bugbear I had (and I don't know if they are addressing this in v3, I think so?) is that you don't get any notification of whether a ZeroMQ socket is up. In a lot of cases you won't get any notification on raw tcp since an endpoint can turn into a black hole, but some introspection of ZeroMQ's internal state would be nice (whether a ROUTER/DEALER is connected [it's not as simple as that since both can have multiple connections]). What you end up doing to get around that is baking your own heartbeat functionality into your app. So yeah, I haven't worked with RabbitMQ. ZeroMQ was ok if you avoid all the extra duplicate poo poo it has to offer (ie REQ/REP are implemented in terms of ROUTER/DEALER, but due to weird msg frame fuckery only req can talk to dealer and only rep can talk to router there's a good reason for it but really who cares). Also, ZeroMQ was forked into Crossroads I/O by some core devs (it's still zeromq but with all the functions renamed) because they disliked the community development model, I don't know which of these projects will win out, ZeroMQ has a much higher mailing list volume.
|
# ? Jul 29, 2012 09:44 |
|
homercles posted:The docs certainly are rather zany, I dont have any experience with v3 only worked with the v2 API, the parts where they breaks down REQ/REP into asinine moma/papa sockets along with the associated ridiculous analogies gets old very, very quickly. this sounds like zookeeper with less features
|
# ? Jul 29, 2012 11:30 |
|
homercles posted:the parts where they breaks down REQ/REP into asinine moma/papa sockets along with the associated ridiculous analogies gets old very, very quickly. “The use of anthropomorphic terminology when dealing with computing systems is a symptom of professional immaturity.” quote:In the end I just went with ROUTER and DEALER sockets, I have no use for multicast or a pub/sub model in my app. So what would be wrong with tcp or udp? quote:It's nice that ZeroMQ handles connection, reconnection on dropout, [... but] you don't get any notification of whether a ZeroMQ socket is up. It handles reconnection except when it doesn't? quote:If you want guaranteed message delivery you have to bake that into your own protocol. [...]What you end up doing to get around that is baking your own heartbeat functionality into your app. The feeling I get is that 0mq solves problems you don't have in return for making the problems you do have, harder. A framework in search of a protocol.
|
# ? Jul 29, 2012 12:14 |
|
i've been poking at 0mq trying to work out why it smells weird to me, and I think I know why. 0mq doesn't smell like a messaging library, It smells like a selection of magic unix pipes. 0mq feels more about data flow than IPC, because message behaviour is defined by the properties of the pipe, rather than by properties of the messages. 0mq messages are opaque blobs. i can imagine making something like a bash script run hog wild over 0mq. for message passing/ipc, something like 0mq doesn't really help you more than bare tcp. you still have to define your message serialisation, error handling, and middleware.
|
# ? Jul 29, 2012 13:33 |
|
tef posted:So what would be wrong with tcp or udp?
Put more simply, it takes a lot of poo poo work you have to do that's different for each language/architecture and instead gives you a common lovely interface. tef posted:It handles reconnection except when it doesn't? For the situations in tcp where you're getting icmp messages it would be nice to have that immediate "this endpoint is hosed" notification in 0MQ, but ultimately in both your custom framework or 0MQ you're going to have to bake in your own health checks for the times when the endpoint turns into a black hole. tef posted:The feeling I get is that 0mq solves problems you don't have in return for making the problems you do have, harder. A framework in search of a protocol. [e] my written prose is just the worst homercles fucked around with this message at 07:47 on Jul 30, 2012 |
# ? Jul 30, 2012 01:30 |
|
homercles posted:In app design you have to deal with some or all of the problems below:
|
# ? Jul 30, 2012 16:03 |
|
pub-sub doesn't really map well to rest
|
# ? Jul 30, 2012 16:43 |
|
Mr Dog posted:pub-sub doesn't really map well to rest Pubsubhubbub Or just do REST over SMTP, there's nothing in REST that says you have to use HTTP.
|
# ? Jul 30, 2012 17:09 |
|
Mr Dog posted:pub-sub doesn't really map well to rest
|
# ? Jul 30, 2012 17:31 |
|
just use soap
|
# ? Jul 30, 2012 17:47 |
|
Enough about your personal hygiene, we're talking about grown up stuff.
|
# ? Jul 30, 2012 17:56 |
|
homercles posted:In app design you have to deal with some or all of the problems below:[list][*] designing a wire-level protocol for encoding your messages (this is the simplest of your problems) thing is, if you're doing rpc/ipc you might as well use an rpc/ipc library. They actually handle that problem. you can roll your own in 0mq, sure, but really, there are other options out there. HTTP makes a decent request-response protocol, and there is a poo poo ton of library support for it. if you use a good one, you can take advantage of http caches and load balancers. also I have no idea how 0mq makes things 'scalable'. scaling an application is m often an operations problem, more than an implementation problem. Mr Dog posted:pub-sub doesn't really map well to rest here's a nickel kid, get yourself a real architecture
|
# ? Jul 30, 2012 18:15 |
|
just use http. put what you want to send back in a json. done.
|
# ? Jul 30, 2012 18:16 |
|
Jonny 290 posted:seems like there are a few dece scraping gigs that i can underbid a bit but that'd just be for beer money. idk if you find an interesting open source project let me know jonny i think we are in similar boats
|
# ? Jul 30, 2012 18:36 |
|
i doubt it. your boat is sinking
|
# ? Jul 30, 2012 19:01 |
|
|
# ? Jul 30, 2012 19:30 |
|
|
# ? Jun 3, 2024 21:38 |
|
I was complaining about someone who had done this (returning a pointer to a function's stack) and someone in irc linked me this http://www.functionx.com/cpp/examples/returnpointer.htm fukken shameful.
|
# ? Jul 30, 2012 22:22 |