JBoss EAP 6 – JBoss for the Enterprise

to the topic of this presentation jboss NEP 6j mostly in twice ? i’m going to talk about basically jboss why historically jobox defeated a big crack in enterprise deployments and they talk about some of the features that will hopefully in jboss dat 607 make it easier to use jboss there a large scale it’s not the massively technical in terms of the slides but if you’ve got questions last week and I’ll do my best to answer if the ever anything is not clear in the slides or anything that I practice and clear please ask questions if I’m talking too fast it’s not because I’m nervous it’s because I haven’t had any lunch earpieces that was really really tempting and convincing my brain if I finish quickly I don’t get Peter anyclip it doesn t work so just ask me to slow down so I’m Matt crazier I’m the head of consulting at CTB too i have 10 years ish experience with jos so from pretty early jboss versions also teenies experience with other application server primarily weblogic but bits and pieces with FC Forge anything to her so if you have questions about other application servers and how or if you have questions about MJ what’s compared to other application servers at all this looks like it’s ripped off from this application server is that what that don’t type questions that lenders are question observances consultants and have been doing consulting alliance application service for a number of years seven years or something out including sometimes of contraction for red hat as a professional services consultant so reasonable amount of understanding of Jaguars and we’ve done a lot of stuff with large customers so that very sperm as some of you guys in here here with customers food to government organizations who have deployments with jboss with many thousands of nodes or thousands of next nope not many thousands of nodes and i’ve been following the bubble to jingles EAP city Ras seven quite closely since it was first announced the early announcements I heard some early podcasts essentially the top features that they wanted in some ways not buying kind of rang bells and Ivan yeah I’ve heard about all these pieces before because they’re all in have questioned several this application typical this application server in some ways was just how are they going to do this I think it’s different to other application servers but a lot of it was well why didn’t you do this years ago like you spent years telling everybody that you didn’t need this and it was an unnecessary feature and then Mary like actually programming all of your feet but I probably well quite personally so I have an idea of some of the reasons design decisions were made I’ve got some very strong opinions on places where I think they’ve still made the wrong decision but only time will tell and CTV to you as most of you know we’re sponsoring running the london jboss user group however i am head photo consulting company and we are independent other the focus of this talk is jboss it’s a JBoss user group it’s not a general sales presentation so feel free to stack up tables as much as you want our questions about it and it’s finally time so kind of things I want to talk about then to the problem space what was wrong with Jay buffs and aversions why did they need to do anything Jay glossy okie cities of a drastic change in previous versions of generous it’s really obvious to anybody who gets hold of it and unzips it i have so i backed up that i can show people later if they’ve not seen the 86 at all the direct research was completely different the file structure two different just a whole architecture is completely different at all see from some of the slides later on so there was obviously they must have realized that something was wrong and that they needed to attack this whole application so everything from a completely different angle and he 86 is the result of that but why what was it that was wrong I’ll talk about some of the color features of EAP success more very I’ll catch will be 86 of me features were talking about how things are done differently and what impact this has on the ability to run large tables deployment and then a bit about the impact of that so what was the problem and what we’ve gone with the ad five well where do I start so let’s start with a history lesson jboss named a loss to time is a pun on a computer

game went on through time on the BBC micro which I have is one of my first computer games are like 284 so it’s probably the significant hit but I found that amusing when I backside Davos was one bob was the first open source EJB container if they built it the guys who built it really they built it because they could buy them because there was a business driver to it because there was a business need fire they’re basically whether this ejb specification was quite good let’s have a go away from their faces over and they did a really good job given up there were a couple of guys who decided to have a go either then built it up into an application server they added existing components like tom cat or jetty if you pick early on which you had a server container their opponents so logging libraries like log4j bits of various other components in hair and laughter which will eventually become part of jealous but which at the time were independent libraries so there’s a lot of stuff to do with a grenade in there a lot of stuff to do with various transaction management libraries and our Gina and so forth it was always amazed about that so this product so it was always very early to market with features that developers liked that made development easy with it but there was always very little take up in enterprise industry in large-scale companies for various reasons so I just wanted to do a quick poll here there’s not and if there’s multiple people from the same company is probably any interesting for one person but how many people have hand the general service of people operation so anyone who’s got around up to five job or servers live for the hands up to a couple up to a ten so five to ten table servers introduction now ten to a hundred Jaguar servers introduction you guys and a hundred plus thousand yeah and anyone more than a thousand now so mostly the biggest concentration is in the kind of up to five or five to attendees kind of level and then how many users skip due to the brackets an expert who has less than a hundred users on my system and it would be here between one hundred and a thousand couple more than a million or between a thousand of the million no more than a million No so nobody has public facing systems running on Jay box in which the entire world is your potential users so again they focused fairly low down that scale in terms of texting numbers of users and then how frequently do people pass their occupations so this is your business applications anybody roll out Patch’s weekly no monthly yeah a couple bimonthly yeah every six months or 12 months now and you can see it change so the people who have large numbers of users marjay so there’s no other patches much less frequently that’s fairly understandable because you have a bigger organization needs less tune and your application you get more stability it’s harder to pick a time when you’re not going to impact to users so if you have millions of users from the whole world it’s your potential user base pretty much any time you pick somebody is going to be wanting to use your application are hopefully if you want to use your application as we kind of sort of this summer group 5 10 servers in production probably hundreds of others and thousands of users probably learning update cycles in the range in front of every month of dating our application this can contrasts with some of the larger enterprises so you guys getting to that somewhere hundreds to thousands of servers thousands millions of users of these five hundreds of users updates much less frequently they have like a lotta money which is what my tax is mostly interested but they’re not beyond traditional jaycox feeding so why I know jboss season why whatever I’m using it well those are the benefits of jingles this gives us an idea for what the problems may be or where they need to start working to to tackle this so plug ability the application server is really a container for services you can swap and change services it may require quite a lot of XML co-packing I’ve ever but you’ve got the ability to do that because it’s an open

source product you can see the sauce so if you will change your super container for tomcat to jetty or something similar you can do that you can go ahead make that change more likely people sometimes change during those containers and use something other than jboss messaging are in fact jboss later versions of jboss the AP 5 you can choose between an IQ or tables messing with your implementation out of box so there’s a lot of choice in the implementation of a particular service that you use you can kind of build your own application server after some of the components it’s also free to use if you use a community versions and the enterprise versions are not that expensive compared to the licensing costs of some of their competitors obviously there are different vendors will give you different total cost of ownership figures based on whichever one makes them the best but in general the licensing of tables with pretty cheap compared less effective all their subscriptions to jail also keep compared to the licensing invested technically it’s free to change in terms of editing the code but how many of you here would actually consider editing to Jake what soft-spoken in a future life probably not that many I know the jboss sauce cook very well and I would be very hesitant to do it in anything other than tiny little bow fix that was really obvious just because there’s such a complex system to understand the interactions are they to knock on effects that sort of tiny changing one component can have is really complicated and you can actually see that yourself just looking at the bugs that appear in the explosion to jboss you quite often looking is he okay I’ve found that this boat clears in this person and they didn’t occur with previous versions that he traced through really find that it was passed to some other component in some of the service that was fixing a bug elsewhere as cause got here so there’s a lot of that stuff that goes on in tables or around security this whole swaths of patches that step on top of each other in terms of allowing certain features to work for breaking other features and then those patties get reverted so that the other parties work and they put a little set of fancy dinner break it again and I just kind of flip flops between this set features working on the set of features working depending who happened to work coming changes most recently so it’s also jealousies community-driven especially the community version but there is a lot of take out in the community and a lot of drive from the community to the community makes a lot of noise and say what we really really wanted osgi containers within our application server so that we can do by osgi bundles with our professions then the developers take that on board and it ends up in the application server you can also do stuff yourself so well we’ve never really wouldn’t really consider changing it so scared too much you might develop a module or plugging up something like that for jboss think we don’t submit back to bed having like white chocolate pick up so far so there is a big kind of community driven aspect to it which is really kind of one of the things that people choose to use jboss fall but these are all benefits that appeal to developers rather than benefits to peel two great big companies to the disadvantages of using jboss mmm it has a lot of configuration files anybody who tries too much jboss administration will know that if you want to change the ssl certificate for the tomcat configuration you give them in one place if you want to do some configuration security for jboss messaging and JMS keys that without an XML file and all of these different file is basically may be quite complicated there’s no concept of central configuration not even for a single server so so if it doesn’t have its own like single configuration files that there’s no way that you could say why I’ve got hundred service now onto other weekday configuration from this central location that’s just with tables fight not really going to happen in a consistent way the community our developers and the developers developers so they they tend to put it together as a product burnt out the door and so there you go guys there’s an application server we’re going to go work on the next version now and we’ll add some awesome features for you but if you’ve got folks only five books that good luck with that and we’ll fix it in jail

tonight something like that so it’s not motivation it’s not about now and the obviously the enterprise platform versions are much better food you can raise support but certainly with community versions you’re not really going to get a patch in the version that you’re using you getting at Apache the next version coming the developers are mostly ghostly interested in delivering a patch if it’s to a new feature if it’s a pass to a picture that’s been around for ages that nobody other they don’t find particularly sexy they’re not going to be that interested in looking at it and administration features a lot have never be for your eyes I had a conversation with with the ex CTO sum over J box many years ago probably five years ago two tables practice on it where yes for questioning some partners and I said what are you going to do about the administration because you have no decent administration tooling and he didn’t really have an answer we talked a bit about Davis operations network which was released at the time but has never been pushed as a product that was never being a really big drive to make people aware that it exists educate users in t2 using it as the primary method of managing tables and even beyond that the developers have never develops they live writing tools for facing other developers will find exciting they’re not so interested in watching tools so they think administration staff operations staff again finally attacking so you just need to look at the day box admin consoles the web admin console through time and they’ve slowly added more buttons but very still very feature pull so James console which does pretty much anything you want it to do but it’s almost impossible to find the right button to click and then they added the web console which was just an applet downside of the jmx console to put things in and slightly more logical tree structure then the jboss five puntos which was based on the RFQ tooling and it was kind of okay but there were a couple of funny features like the ability that there was a restart button next to a server when you take out these little chips down and never comes back up again because it was placed on our HQ and expected to have an agent that will be able to ship a server down and then start it up again and without an 80 win against everything you just did we start and shut itself down and then obviously it’s dead constructors how public game so the most funny things in there and I’m never a really good everything and again the diversion so the console that came with the with juggle squad could only manage a single jayhawks instance and that’s because there is no central configuration so there is no design there’s no architecture that allows to have a central configuration repository to push the changes out where the servers and because there were so many configuration files it was even too difficult for the developers of the admin console to come up with a reliable way of allowing each make changes and persisting back to configuration files so they basically gave up on that and most of the stuff that you could do in the constables we’d only just because they couldn’t guarantee that if they try to write it back to the files they would interrupt your generous instance fight those are the disdain effects are the problems materials and these are mostly things that impact operations teams so the CJ box has benefits for developers and it has problems for operations teams as we kind of been through summarizing game so the similar configuration files if you’ve got junior administration team that’s responsible for my magic Jayhawks are probably happy citizen that kind of thing there were some databases but the 2nd line support team is pretty much impossible to say here are some instructions for doing some stuff with jboss if things failed you need to have some kind developer or backgrounds and 14 the commanding it’s really complicated to get consistency across sets of servers and he can’t really manage large numbers of service so most people come up with some shell script that will ssh into all 40 jboss servers and start them all ssh into the mall that stop them all that kind of thing but it is left to the developer to do and there again there’s no kind of solid way of doing it in their best practices around it so in those organizations the operations teams often get popular saying 10 is it going to be used because those are the guys that are going to support that system for the next ten years of it with a lot of cycles that application is so the developers come along and I spend a year or two voting it but the support team traditionally has not a massive budget but they have the budget to take time and they also have the ability to refuse quite often to take on system the

day go see is supportable which pushes a lot of power their way in terms of saying what they considered me to support a book or technologies get used in that type of thing tradition is the 14 forth under some kind of IT services directivity technology kind of leadership structure and development teams under a similar structure but offered as a project or program type system whether the lives of money is a bit more complicated so do sucesso to be an effective operations teams in large companies operations team to a bit more power and money than in small companies and the large country to the ones with the money so turning the problem upside down see the solution is to maintain or proper to make money for red hat they have to make gbox attractive to operations teams so the goals then for J cross the AP 6 on 07 the biggest goals that they talked about reading or solving this big manageability problem because it was beginning to hurt them as well as their customers never starting to find that actually they humbly devote a component of the jboss bikini messaging implementation are some new transaction management implementation and final day of conflicting with some component with someone else is using which is using a digital version of it services library or a different version of spring or something like that they got themselves in such a muddle from having all of these separate projects tight together with a couple of xml files the labor began to feel the pain himself so in addressing app they had to essentially address the whole thing so they wanted the ability to managing configure multiple jboss service centrally if you’re familiar with the other application servers you’ll be familiar with that concept you have a sensual configuration repository other instances startup go to that central repository and say I need a copy of the configuration this is my identity or whatever and they get copied the bit that they need this is new to anybody that is a basically a Bedouin codon so unless your goal in life is to run a single jboss incidents with a web app your local church or purple whatever it is he likes you in the evenings then it’s a good idea for you because it simplifies a lot of things and the main question question but the thing that I’ll give us and give white with a tape of this one so the biggest changes the biggest features first one is to take off now has a single central configuration file all the configuration of the jail is in a single file that’s either standalone XML if you’re reading a single standalone stable systems oh its domain XML if you’re running a domain which is this feature where you have multiple servers with a set tool configuration they’ve introduced this concept of domains as an overarching configuration concept really architectures the word not so fast factors were able to have been concept is a better way so domains allow you to configure manage deploy to and control multiple genders servers of single central point so this domain XML contains configuration for every jboss server in that domain and when Anthony starts up and goes to the domain controller and says I am an instance of my app server one please give me development configuration and get set back because they’ve done the change is necessary to have this single configuration file man comes quite simple it also means they can do other things with administration tools later on like now that they’ve got a sensible schema they can actually have an administration tool that allows you to change any configuration anywhere an application and save it and have that system so you can kind of see this massive chunk of XML this is a standalone XML file a bit slow zooming in step on this because it’s quite large so I’m just going to share a few Bourbons two at the top it defines a bunch of modules that get loaded to jboss EAP 68 / 7 is taking a modular approach and all of the components are defined modules and it’s module has a descriptor that defines what jar files make up that module so this extension module or jboss hey ESPO Joe means the tables will go off and looking at modules directory and go down the directory of South table / chairs that’s podium and it will look for a JBoss modules xml file telling what is due to load that module we have modules for the various things web application thrilling security promoting

osgi navy jmax he’s every three women scanning all of the kind of services you would expect in an application server you’ll also see in this file we have this source definitions so previously j ross used to have d s xml files that define the extent yourselves have non now gone down the approach of having all of your data sources defined it is one the xml file which to me is a little interesting because the other application service attended have recently got in the other dimension of having all the configuration acceptable data sources and je m’excuse in a single configuration file and then each JMS chaotic sauce is defined in a separate file but this is your folks that do is build for now so we’ve got an example data source they’re chasing using hypersonic 282 is over very database i’m not sure at any effective our words but excellent was pretty awful so their source configuration configuration mak it to use in all the configuration in 14 in one single fire fire Robinson splitter exactly again so for the data sources we’ve got some debugging information native people of Texas people of HTTP Xbox osei amazing poverty class stocky body for Bob on 80 81 I think I was made back around bubbly there is amazing some of your manic phases to 990 though 94 for free so we had a separate management interface to its external web interface so you can expose your management interface essentially over only your Lyle Waggoner so subnet and then have the Xzibit pots of bootable information so you could buy this for different address i bided different pods are you swallow all to it so when you go and do any administration says whether that’s command line or where the admin console if you’re using these management courts not the public-facing fox here in this configuration file is all of your deployment archives so every application who gets to pull up its net and entry in here so this sample app got one the world has a Content unique content ID and there’s a mapping in this file as well so it’s which service at which content ideas met so you can have this is list all of your deployments but not but not all totally to observers so these IDs either open a file to say this server hands the content item 9 and 6 7 ZF but above deployed to it there’s quite a heavy fire the standalone XML it’s gone and there’s at every piece of configuration thing that made in this there were no other configuration files essentially in jboss rather not any serious ones so how do they get to this point and potentially lead rigid major parts an application something applications are starting with the coal they have completely new architecture and you server architecture deployment architecture that essentially presented a boss is a single application server product with a single configuration file single administration console and much more sensible directory structure but multiple processes are involved so those of you familiar with weblogic which has this concept of a node manager as an external process words that can start and stop weblogic instances absolutely team a little bit more complicated currently reads the titles of these boxes we have a domain controller which has preity zinta main.xml file it leaves them there the server manager process of process manager persons to these are the process manager is a very lightweight process and server manager is also kind of a lightweight process the server manager of process manager starts processes to conserve servers and style Service manages the server manager communicates to the domain controller to load the configuration for particular server instance and install images just your instance of jungles on the disk so you have these multiple processes that communicating a fairly complex way it’s not too necessary to understand it other than we’re figuring course you just need

to know everything to in the magic box and the main control box you need to tell so good store more instances you would just because essentially take this integral and copier and we need to update this would have divided into the local servers in there and you need to tell it the location of the domain controller sorry knows where to go to to the configuration I’m not too sure about the slip between the process manager the server manager why couldn’t just be one process but I’ve had email conversation good work actually on the developer list never really got much of an hour my personal view is that it comes down to them believing they were about web project node manager being too heavy weight on it reliable but there may be valid reasons why they’ve split it that way I just they have tried to tell me what they are and i just it doesn’t go in and doesn’t make any sense to me yeah multiple courses and Zimmerman’s so you have per box you have 15 secs manager at one server manager but potentially multiple server processes and then i’ll live on the whole summer right that’s like thee and then the logic is the admin sir exactly yes I just in this condo in controllers work yeah so the host controller has only his specific configuration on it so you would tell any tell the host controller on this box I want to start an instance of silver type 1 and in two instances of server 22 I want to buy them to these parts please it also tell it the domain controller that you go to you to get the configuration from these is to touch address and we have got the talks to them versus controller to a stock purchases you divide work process the domain controller has all of the profiles for the whole domain it reads all of this from the domain xml and it also runs the administration console which is something slightly obscure that takes a lot for you five so when you want to go to the admin console we go connect to a jo things to keep connected to the domain controller which is I guess just come back to this goal in there so I hear the thing about this whether this is lightweight i love this is actually running a web container or at least this domain controller is a very wonderful day when a domain controller the server manager are the same essentially saying this Java code perceptive one of them has a configuration file one of them doesn’t and configuration colleges has to be I pierre-joseph somebody else who’s got a configuration file so this is actually a fairly large processing that it does contain a web container to this running a web application it’s an admin panel you can end too can you connect to each container you friendly with domain owner today they are other ones sales yeah it is really only for anything so in the sterile run was out of the net floor yeah yes seriously moment after making todavia they can’t they don’t know if they can start and stop without the domain controller via really interesting experiment it might be that it won’t start and you can’t ilysm but because it wouldn’t have anywhere to go I don’t know whether a server control it keeps a cached copy of the most recently valid profiles or not let’s follow roughly on a lot of bourbon that’s what that’s what they do but whether they put up that’s on them yet we never tried it so two main mode kind of talk about attitudes to share a single configuration the poetry main controller basically owns that configuration the cross between they request the configuration from the domain controller when they start up so the key benefit of all this approach is that you then have a centralized configuration of all of your servers so either it demain XML on standalone eczema and you can start to stop them as ammonia so you can go to the domain controller and say I would like to stop all of my servers and it gives you talk to the processing I do a lot of the boxes that stops them all that starts tomorrow so you have a single bite to start stop my servers you can also deploy to all of your service so you can say I want to upload this lever and you go through the browser to play that to the six servers go and it probably 12 Sixers and spouts it so the key thing is to kind of notice about this then super configuration based on profiles domain XML contains configuration for all of the servers profiles in the

domain so if you’ve got different types of service and of which have some web applications deployed others have others maybe sometimes JMS services and that was dead then you have multiple sections in the domain in tomorrow so the main interface configuration followed with silver falls reading between the lines in one day families the 7-11 xml contains only one configuration for profile so standalone ex-mo just as one server profile in it so this is the sender learn xml you can see it probably get to my far we zoom out this is a domain X enough you can see it’s roughly eight nine nine five times as as long as the standalone XML I think this one has three server profile a bunch of stuff about the domain as well so literally everything is rolling that one almost everything wrong and that wonderful fit into main mode the admin console of those are the domain controller lot of service if you want to connect to it then you connect to the domain controller and confuse people bit because it’s different to have labels have done it before so the administration several tables with just a web app running in gables now it’s essentially wedding with different process one of the questions that comes up its hands over this domain speculative clustering and domains and clusters are completely orthogonal concepts so if domain is a administration configuration unit of information of configuration and a cluster is a runtime replication of state so you can have a set of stand-alone nodes each with its own standalone XML that contains information that says join a cluster using these multicast details and they will join a cluster and run the cluster I would advise doing act because then you’re back to the whole day was five system of what happens if somebody logged on to another doctors and accidentally changes configuration in one load but not another load and they stopped seeing each other but it is possible to have addressed at sound of their domain lendemain can contain zero or more clusters so you don’t have to cluster server just because they’re in a domain and it’s not a one-to-one mapping between service in this domain oh it is a cluster you could say Civil Code far one defines it cluster at all service little tikes ever quit far one former cluster so the profile to doesn’t have any clustering and he served as a little types if you’re up to David cluster so the profile three might have another cluster setup from 0.1 so can you have the same server in multiple positive so 2 to 1 702 Plusle probably not so we might think in the ticket essentially just joined the clusters together because anything about getting to Ecuador okay to the other servers and you mentioned multi most beatiful union else it doesn’t yeah so just to follow Jesus called unicast I haven’t charged configuring haven’t recently tried configuring a unicast cluster with diego 7 so one of my concerns about Jay box 70 e 86 depending what you want to call it is we find that big xml phone again so everything is defined in this file and there’s a pretty solid schema that says what you can put in this file and then it being fairly good at making sure that there are ways past things through so if there’s a property that they haven’t exposed in the schema of most things there’s a way that you can pass it through so if you’re using some obscure Tomcat setting that they haven’t said here there’s usually a way to pass time flew but I’m fairly sure that we places where there aren’t whether miss Delf ways to send configuration and that they don’t see there being a you spaceport would actually turn a customer somewhere wants to set this flag and just quickly the general developers too concerned and interested in and schema suddenly cancer this package mark so I think it will be teething troubles around there and I can and that’s maybe one of the areas where you might see that it is in some of the clustering stuff so administration tools new admin console it’s looking better than it did in early business 07 it allows you to administer the whole domain and the changes that you make

with it a persistent so you can go in and you can change the back way back for setting those apologetic oddball that was me playing around with the admin console and saying what happens if I creatively plug back in court manqabat fog and click Save and we’re giving the xml file of i could look it’s actually a that’s not so you can make system changes because of this single configuration file single syntax there’s also command line interface they have their own management API away and management syntax pretty complicated a lot of people so saying what if you just stick with geri metz there is more powerful but it’s not really a standard so this kind of complicated but it allows you to use one vital to go in browse through the tree of configuration things view configuration elements it’s kind of like the command line you think that people are like gem expert if it’s not using Jennings but you navigate through a tree is part of our core components that configuration attack student and you can do anything that you expect to be able to do with administration tools that change the figuration deploy stuff and we start so it wasn’t good as Stephen yes the more humbling so typical st is based on joyful noise it’s kind of that’s almost a complete programming language where this is more of like a shell where you can go in and CD in two locations and change properties and things like that rather than a full program language it’s the administration console secured while booking it is instituted by a login I’ve never tried changing it to anything more secure than if default so far default there was a text file essentially that scores good looking information tip there is identity stolen today but it and there is no default user so when the first go to the admin console they say there is a fusion October just modify one then you win that come out pretty easily won’t have that muhammad bello integrated with like elders with each other so i would guess you can get up not track pleasure still with the little a nipple step together the lifecycle management over the future does it give me the way lots of operational please that’s a good question i would guess not at the moment but i’m writing their something that they might look at daily if it’s going to if its life what other that better to do you’ll probably find their domain can contain multiple versions but only versions that are older than the main control so it’s got the domain controller has to know about all of the skin problems that might exist and I’ll 18 only know about past things with a commended area we change this female menace to this in this version become in the future we’re going to change something else so so in the de- is so essential so the dog there are individual log files bank which is interesting turn this bit so that’s basically it so yeah let the Thunder start in terms of now it’s out there now when we’re going to start seeing real bugs come out as really just start hitting any during actually I wanted service tomcat flag that you don’t support some of the future so it’s going to leave though sometimes the development list is quite fairly quiet at the moment the biggest discussion they’re having it at the moment is right in the next version whether the default profile for standalone XML should contain the osgi container or not and it’s pretty much a 50-50 split to even as to whether OST I is a cold components on that I’d imagine it will end up in there but that’s kind of thing they’re talking about at the moment you’re moving here at addresses it would depend what your tongue scales are for deployment I think he’s probably the answer so Yankee six is a lot better than the universal Table seven if it’s an application that is not completely mission-critical then it’s probably reasonable thing to start looking at if its mission face black person I’d give it a while the year or so please recording star you’re deciding it cheaper efes okay yeah no sir it seems like then between the developers you’ve made your changes said I think they’re probably in between where they fight over box that they are

next OG&E baubles this is only stubo seven and i don’t think there was that much uptake of it because it is a massive and also the as I can’t talk about the drivers been using jboss seven enough diapers the people who were playing the community worship that leading edge community version are going to have so no one is going to like nuns going to sit down at home ago yeah i like using latest version of j box and i really want to try it out for my 15 a cluster with a hundred thousand users to kiss the people who travel it especially tickles don’t have the e versus so i think you’d only now that the AP six is actually come out that we’re going to start seeing Papa cases of the things you will do in common in classroom yes it’s horrible yes and that was one of there has been so much so it’s not so much now but because I think people have settled down to it but it was the biggest complaint when the detail versions and these characters have set them around but it’s so massively different it lots of things don’t work I know wanted to be I think the first feature release no even beat Alex it was 700 for the first GA release you couldn’t get his austere swing didn’t work too couldn’t load they say epocrates ball out of the jar file on xml as agile as a speed we just couldn’t find there’s lots of problems in translating the most of them are probably even though workarounds for them if you post about various forums now but it’s massively different so the there are a lot of things about migrating an application to be a p6 that i haven’t even touched office is just bit we’re kind of the main structure in terms of migration application i’d also be like see I have more experience of people actually doing it if most of the seven stuff absolutely I patients rather than my great further so people are starting from scratch in there going okay so it doesn’t find a class that if I do it this way so if i do it that way if itís classes are not just carry on that way rather than taking a big application dropping it on there I go I’ve passed on family searching everywhere yeah the tazo thing is I’m not a massive fan of it the cars and my opinion they’re just getting further and further away from the standard je class whether the webspam weblogic and everybody else in the world using that hierarchy the reason they’ve done what they’re doing anything 7 is because of this thing I kind of really touched on where they keep ending up eating different versions of services or spring or that character the thing that killed jboss community sex was the fan that they used a really new duties part that broke pretty much everything else including skirts so if you had a scrubs application she couldn’t learn it on community six because skirts wanted to use a different version exerted and there was no way to manage that kind of dependency so they’ve gone for a much more isolated clutch lever hierarchy all right by one complaint jboss to use one versions earth is another version of music another veteran services letting me do deploy your application and it goes other than operations overseas one therefore from and click remember one Chewbacca what sort of zero problem are they going to know how they’re planning to progressive some there for years I haven’t seen it if I my lifestyle I don’t know whether it’s an internal or a partner document public document there will be one button after them anything else this piece of ice caused a home’s no lunch they are not osgi they’re not fully OST I they’re very light osgi but the life cycle is love simplified they support OST I module so you see there is a module that allows our CEO modules to belated but they’ve taken a simplified version of essentially a life cycle and what’s involved in xtr knees back step is just one water you can see anything still there was interoperability between the team landed there to accent it goes because open not really done much with msgr I was not it’s not over the features that I kind of kick and see how it works when I was back counter I know there is integration between them because i was reading some posts on the development list last week for somebody who had an application with his part

osgi and Part A to E and he was having problems with plastic between the two that it works in general 50 at least osgi container I think that they need to use it with tables one problem was deployment ordering at the gym jeopardy ejbs that the OTO components start up a specific order but they also expect djb to set up I specific item they didn’t they immediately started up in a random order and it’s not the same each time you start tables so that was the problem I’m having is good even if they don’t we did there were two young components to start up in a different order the next time I staff at wgbh stuffing in different models and fortunately the process manager supervisor is that two separate process which one all the server he has an ace stop and killed processes and it it’s a start and stop receiving how does it stop they sometimes within the same jvm they’re two separate processes like if you running in standalone mode they were all in the same jvm to their different threads essentially related thereto separately in a domain thing that you simply think of a it started out that a script that starts I think the process I think it’s the script up process manager which starts the 7th rounder your dad’s a book service and the students dog suzay you start service you’re not sure you can start the start service the managed service I don’t know a little scar I don’t need to start something I think their life cycle is dependent on having a a process manager but that first if magic can be same JTA what should I think I found the way backwards but then understand one server can’t just invoke the service using for strategy yeah there is a long way if you could go to the Commandant of momentum yeah come on scripting to a thing