JUC West 2015 – Bringing Continuous Delivery to Cloud-Scale with Jenkins, Docker and "Tiger"

Just another WordPress site

JUC West 2015 – Bringing Continuous Delivery to Cloud-Scale with Jenkins, Docker and "Tiger"

thank you for joining me and kosuke for the stock how did he like jeans keynote alright that’s good so we’ll try to keep the same momentum going all pressure on you all right ok so the talk today is it’s called continuous delivery at cloud scale and we’ll talk some bits off the crowd you know of the cloud based product the jenkins platform that we work on we’ll talk about the recent work that’s happened in open source with darker kohsuke talked about this at the keynote and we sort of do demos if all things go ok and the third thing that we will talk about is a project that’s internally named a cloud beezus internally named tiger which is about taking bringing jenkins of the service into your organization’s building a Jenkins of the service to your organization so it’s going to be a fast talk between these three topics so that out of the way ah so there is the the introduction that I do for for cosa gave the keynote and then there’s the real real dirt on keynote on cosas que so you heard me talk about him as this you know CTO program or extraordinaire favorite weapon of choice is an IDE he is coding and all that the scratch all of that what I see him is this is this is the view across the office I see him as the slide boy for marketing and his actual weapon of choice is powerpoint in fact he was telling me yesterday that he wants to certify as a as a PowerPoint a writer so that that’s how it is that so well here i have this maze the we feel for the management of copies to her feet and he job is to ask us to do things i sometimes wonder he has this till i might leak he has a way of making a sworn what we want I forgot very kind of great art and he’s not entirely true so my Jedi mind trick actually is is more kung fu ins I call it the clasping clasping hand technique that employ I employ every day with kousaka is like I need some new features give that to me so so that’s about it that’s that’s us are very quickly getting into cloudbees so if you don’t know us already oh we’re the enterprise jenkins company we do all things Jenkins we have a solution on Jenkins that we call the Jenkins platform that’s helping organization bring CI and CD into the organized into in-house and whatever you heard about you know from gene came we we see ourselves as sort of accelerating that for organizations and as as a p.m I feel fairly very proud to see the list of our customers here you can see there there the who’s who you’ve seen the netflix name can come up earlier there are customers so who’s who off people are trying to bring see I CD and DevOps within the organization so I’ll spend the next maybe five or seven minutes talking about the Jenkins platform and and then I’ll hand off to cosas que to talk about darker so the ways to be the way we actually built the product is is split into what we call the team edition and the Enterprise Edition N and the way the products has evolved is we we often see small teams bringing in Jenkins within the organization as they start sort of I trading and delivering software faster they start hitting into needs like I need a you know integration with get I I want to make sure none of the bad commit commits that I make no bad comments walking to the into the repository I want to have like start employing sort of security in the organization’s there are a couple of plugins around out so so that addition is focus steward smaller organizations and as you sort of buy in to develop cic d what tends to happen is your infrastructure grows fairly fairly quickly and that’s where people come in and take the Enterprise Edition which has things like high availability and a number of other features and as part of the product as

we you know started selling this for years ago the product itself evolved the platform has has something called the Operations Center in it and where it came in is we saw a lot of organizations that brought in CIC d they became the beacon for other people in the organization in terms of look they are delivering stuff so fast how can we do that and you would see like Jenkins popping up all over and at some point of what people wanted to do is have a single pane across all Jenkins within the organization they wanted to share the slaves across groups so they are not going and spending money for machines for each division they can consolidate that and and and just get developers to be on 11 system so that’s where our operation center product came in and that’s sort of available part of the Jenkins platform and as part of this a platform and operation center you end up we live in the last year we’ve added a lot of features around analytics so analytics around build for performance analytics around performance of Jenkins itself and with the release of workflow of what we’ve done is we put a virtual stagevu that’s what we call call it what it helps you do is kind of going to genes mind it helps you sort of visualize how your pipeline is doing how what’s sort of the lead time for your pipeline as things go across the pipeline and where should you focus so that’s that’s part of amongst a lot of features in the product and one other thing that cosa gave me have talked about in a lot of a couple of juices in the last year and I’ll just touch upon it is as organization start becoming bringing see icd what they used to do was they would try to scale one single Jenkins and keep scaling it up and then you had that you know Ops guys sort of looking at this with the some amount of trepidation seeing what happens of Jenkins feels all my developers will be impacted so with the operation center we started talking to people about building this architecture which is a more horizontal architecture so instead of having one Jenkins building you end up with multiple Jenkins building and that’s connected with the operation center and so we ended up talking about this reference architecture that you can use with operation center that helps you plan and grow your your CI CD infrastructure in the organization so if you’re facing some of these challenges I would ask you to look up cloud you know Jenkins reference architecture and you find a white paper written by me at some point talking about how you can build a scalable infrastructure and and we’ve sort of we continuously have you know I traded and improved and delivered features on on that so with with our latest release we’ve added some features where we’re when people brought brought in this kind of infrastructure in house they wanted to get to a place where I think I saw this in one talk yesterday which is can I promote my jobs and make sure i’m bringing my jobs and test masters and then i promote them to a production master so there was a feature that we built where you have a test master that’s connected to the operation center and then you can seamlessly promote that to a different master of it in the organization then we started seeing the needs as people started scaling and and this they brought in the reference architecture and had multiple masters and at some point your master has become too heavily loaded what you want to do is split it so you could split the masters and just move jobs from one master and to another master so it kind of helps people sort of easily transition from this monolithic architecture to a more horizontal architecture the top feature is is fairly new i think we’ve just pushed it out in the last couple of sprints and the last feature that we release in the last cycle which was May was this ability to trigger jobs between master so you build this architecture now you want to call you know in your pipeline call a job from from the dev team to a QA master sitting elsewhere so you could do that as well and keep forgetting so I left a couple of slides as as you look through the slides later on that kind of gives you the details of what’s in the platform and so I want to take any more time with that and then the last thing that we’ve done in the

last few months is is released the platform on amazon web services and we are also working towards leasing some things on Azure so the idea here is you have an operation center and you want and you have a lot of teams each with their own master and you want to share slave resourcing across these masters you can use the the amateur amazon pack to to set the configuration up at the operation center and then seamlessly push that across your entire organization so it makes the life of an admin easier and we’ve done something similar on the ashore and the pivotal side of things so that’s that’s who we are and with that out of the way I think let’s get into the Jenkins and darker part of the story right ok so the this part is actually mostly demo doing it so I talked about it yesterday you know the kind of doctor integration features we are building so here I wanted to actually show of most of these plugins in life like the thirties the station that I want you to be happy in your mind is that you’re working for let’s say I come for that’s running out nerdy web services a lot let’s say a micro microservices so you got a one team that’s producing this base image foundation that contains the right words in the Tomcats that’s security hardened and whatnot and then so that’s the that’s time by the platform team that’s the stuff on the left hand side so there’s a source code 3 that hosts those which in Jenkins joke that build a docker image and that gets published as a base image now the there are the other team is building applications they have a source code in their own sauce 3 and then what they do is see there’s a Jenkins job that’s the picking of the sauce tree and then building the application image on top of the base image that the company has produced and then that’s the one that got into the production’s so be so so here what we want to do is see so this is the Magenta’s instance actually let me make sure the landlord mean so here i have the base in this field so this is done by the platform t so if you look at the source code of this guy it contains the you know the doctor file that builds basically build a tomcat and then that that’s this cotton code custom at company customization which is setting the Tomcat configuration file now the actual build if you look at how this job is defined this is a standard a freestyle project and so it’s checking at the source code from this kid people three and then it’s using a darker build and published plugin down here so what this plugin is doing is it’s building up image and then publishing it to the doctor have and by using this credential that specified so even if i have 100 or build slaves every time i do the build it uses the right credential to be able to push it into the river and then that’s your decision there’s nothing more so as a platform team like I don’t really have any knowledge of what teams are actively using my image so on top of it there’s not a separate team that’s building the application so this is actually a defined of the work so what we do is if you look at the source code of this guy actually need to get rid of configuration first all it’s really doing is specifying where the source code repository is I hear a github upholstery and then they run in a walk so from there the other interesting thing about it is you know when do you want to execute this job so obviously whenever the source code changes in the github but also if you wonder the base image changes and I want to be able to rerun the bill so imagine a security vulnerability like on like the heartbleed so we did eyes a portion of the Tomcat or the ssl library has issues so the top home team needs to produce a new image when that happens all the app also needs to be built so that’s what this is trying to do and the beauty about it is we don’t need it you need to tell Jing is exactly what images it means to look for the changes because the artwork fell and then the the doctor support talks to each other we can automatically figure out what images to build Eliza and that’s the right thing now so this is the actual app the app itself is and

defined as a maven build so there’s a doctor file that turns this into a container it starts from the base image that you know the platform team has produced and then you put the app that we you know going to build inside this container and then that’s the end of it so that’s pretty straightforward the pipeline definition is indistinct in smile and then this one is an easy enough hopefully but still probably is worth explaining so it has three stages the line till nine and 12 and then so we have a build which uses Maybin to build and then the package and end up with created over images and avoid the stuff on the build is a little bit hairy so what’s going on is this is actually use also using docker as a build environment so imagine your product developers you want to make sure that the abutilon violent has the right version of energy JDK or the specific personal Maybin with certain configuration files or the find box is the rightful filtering duel already pre-installed so what this does is it in it so that’s this image so this is the build environment image that contains all that by tool and you create them and prepare them somewhere else now as a build start what dentist is going to do is to launch that chillin container and then run these three following steps inside that container so we check out the source code and then we run the Maybelline install which is you know beauty java web applications and then finally we kind of pull out the river war file which is the binary that’s built as a result of this bill and then you know unlike line seven we close the scope meaning this is where the building container is gone so all the intermediate files any leftover it’s all cleaned up nicely and then in the line thing we’re going to do the dock or build which build the image by adding this war file into the container as we saw and then finally the result gets pushed and the deployment step is a little bit involving for this demo but that is basically it’s like a pushing this container up and then running it in the in its own location so what I’m going to do is the you know simulate be I guess the simulate how the simulator how how the whole thing is put together so imagine now we discover the security born every day so i’m going to i am going to sort of change as a platform guy i’m going to make a change to the server configuration file so here font little smaller so let’s say I discovered that Oh CC doing it in small screen is making me salad with interesting challenge but so imagine like the vulnerability that oh I’m exposing the details of the tomcat more than I’m running it is not good so I’m going to change this feed be another meaningless Polly’s right so this is the actual body the server header that the HTTP response is going to contain seminars but let’s say fix the security vulnerability and then commit this change straightening so if I go back to the Jenkins instance so it already Jenkins already got the notification from github and then so started building the base image and then this is where I’m cinelli I think it’s already finished building and publishing image up into the docker hub and then what’s supposed to happen next if this push go through number of layers is kind of amazing here oh so it’s successfully pushed the image so it went up to the docker hub and now we are waiting for the don’t perhaps to send us back this notification heiser the new version do an image and then what i was talking Jenkins has figured out that oh the app depends on it so let’s make sure that it gets me also now the workflow had started building so there’s these all these stages are getting executed right now so you see this this is part of the ax Travis Jenkins flatform feature building a new image so the standard maven build is going on the packaging is so nice the malady container is getting deployed so so this is a log file let’s see let me make the hot little smaller not that is really important here but it’s getting pushed and then finally when it’s done any deployment it okay so this is the app that actually running so to prove that it has actually finished building debug it’s for both so if you look at the server header that it has been correctly updated so it has to be really fix now the other neat things

about this water locker support is that be we have at well the traceability support here so now if you look at the bill that that’s what just happened you know we we know that well the Jenkins work we know that these are the images that we have used so here’s the one that I guess the the Optimates that was built this is the one that’s the bill and box are used as a tool in container and it keeps this is also the base image and it keeps track of exactly where the best image came from so that came from base image field 34 so we can look at this changes of the co it contains the security vulnerability fix we needed to see so that’s the kind of liability so that you can see where in the applications this year fix is being deployed all rights I native and spit up here so now i’m going to change the daughter so this demo so far is for people for using container or the unit of the opponent all right and then it’s probably you know i think a lot of you here or not building web application maybe you’re feeling the embedded applications already in appliances or stuff like that so but those people are still benefiting from I mean they can still benefit from Jake from Jenkins and dr. integration so this is the accession of the one talk about how you can use docker to know the two runs innings so the so the first station i’m going to talk about is if you already jenkins administrator like you have the tens of build slaves you’re responsible for creating is coherent the homogeneous build environment for your team now if you can use docker to define the environment and massively deploy them then that makes it your kind of easy so drink instead she do that by using a taco flogging which is historically the first rock quarry dated integration that came into being in the community so what you do here is you know the you as a system administrator define what’s called the cloud so here you specify where your dog or back end is running and a few information like that but more critically so you specify the docker image that is the slave image and then you associate back is available so now I call it standard so imagine like this my company provides the quote-unquote standard build environment and then this particular container has that so is that what I could do my be so that’s all that the administrator have to do and as a user I can configure job on my Jenkins and then all I have to say is to make sure that the build is tied to this standard level and now if I do a build Jenkins kind of makes this mental connection that 02 now this stove wants to run on the standard build environment which is this container so I’m going to talk to the doctor back in provision the new build slave lime cilantro container send the Jenkinses ain’t over start the build which is clearly not oily ok here it came for a moment I worried but now it’s starting tiniest new build sleep and the night started executing a beaut and then after a while since in this case there’s nothing else going on this builds they will disappear so again you get this like a constant clean isolated build environment for every build so that was the kind of second case another third case is if your Jenkins the user and then you’re dealing with unresponsive Jenkins administrator imagine like you know the administrator I have to deal with hundreds of teams they don’t have enough time to make sure that the tool you need is available so you can also take a matter in your hands by defining the the build environment inside your job diffusion as well so this is what we call the doctor custom build environment so in this case the guy who is in control is the other guy if it is configuring the job so here I’m checking this build inside a docker container in the specifying the building image again if I do the build the end result is more or less the same you know you do w and then the builded start and any hope in this case it should have supposed to have sort the inside function tonight so that’s new maybe so so the end result is it’s going to start a new container and it’s it run the building side and then if I need to finish the container will also goes again you always have the screen isolated build environment but it you know it guarantees that them well it allows you the product authority beauticontrol so that was the kind of this demo other quickly so what you saw was first and i showed the Jenkins workflow as the means of creating continuously pipeline to use a worthy

docker images is the unit of the foreman and the other one that I talked about is by using the docker build content docker containers the build environment and this can be freezin for combined with something like Krabby sinkings operation center that he talked about a year to create mini vaca in the homogeneous building violent across masters and then third we did something similar using a container as a separate like I isolated the building behind but we gave the developer the control like a people who are configuring job skip to decide exactly which images they want to use so it caters to be different use cases and then this stock of work is very important for us so we started taking its own to the other area so we’ve been developing the Cuban in this plugin I believe there’s going to be talked about that either yesterday or today and then the work though is also big for a bit as we saw so alright so there is that I think we are doing ok we are coming back to the addy Tyler so the or more officially called crappy stinking stuff own private SAS edition so i’ll have the half-breeds we are to deploys for the talk that’s that’s the new metric that i’m starting work yeah that’s the new metric that i’m going to measure from now how many deploys can we do in a talk so the reason why we talked about the Jenkins platform in in the first part was to sort of set context of how we see sort of adoption happening within within organizations and what we what we saw you know was this level of automation coming down to departments but increasingly a the use case is becoming I would say more granular and the way this starts off is as people come in and listen to guys like Jim as he said gene he said CEOs go back and ask their organizations on how can I deliver software faster because delivering software faster is now a competitive advantage as as gene made the case and typically uh if you kind of goes down and it comes down to our oral evidence organization that’s responsible for bringing CI and CD across the organization so that’s that’s the use case that the private SAS edition or Tiger is targeting and when that requirement comes in and I’m sure a number of you have sort of been on that route in the last year is it translates into how do we provision masters for teams within the organization how can I actually provision not even for Team put for projects how can I provision stuff for slaves faster for teams that’s where so the sharing of slaves across masters comes into place and then there is always the case of building an administrative front end to the guys who are launching the CIC d infrastructure within the organization and that typically involves you know single sign-on and so on and the last part here is is sort of resilience to infrastructure failures so these are sort of the core requirements that I’ve seen in the last two or three years as we’ve seen people throw rollout Jenkins in mass and and this was personified by a company in the bay area that I can’t name but they have like 3,000 people they wanted to roll out CD it was beyond Jenkins it was get a go and get it and it took them about a year and a half to two roll that out a year and a half to two to roll that out it cost them a few million dollars they’re really happy about it however sort of what you could clearly see is what’s what was happening was internally they build this customized snowflake that now a few heroes that sort of led the effort are solely responsible for maintaining so they had like a huge ops team to just manage that infrastructure because the guys who built it for building managed services and had not built something about that sort of is cloud scale so for example we have a Jenkins as a service called cloudbees David cloud where we run a few thousand masters that’s run by a very lean ops team and if you’re trying to sort of roll this internally you don’t have that expertise to bring it at that scale scale the cost of stabilization was hidden there was you know I think it took him a few months to stabilize this and so overall about a year and a half from the time they wanted to roll out the CD as a service to getting getting it out of the door and getting it to the first few customers and that’s where I internally we we started looking at what we’ve done with garbage link in the service devid cloud and trying to bring

that to a wider audience and that effort was called tiger say what we wanted to do was bring the same cloud scale implementation into organizations are you can have that running in your in-house it was supposed to be it’s supposed to be battle-tested because we run this for about three or four years and the idea is you bring it in and you have your own internal cloud and we are launching this with on amazon AWS and on OpenStack by Red Hat and the key idea here is again you don’t need to pay like a year and a half two years to get a turnkey solution to kick off a ACI CD as a service we want to cut down the time to market and the infrastructure should be just resilient from get go and a of the other benefits here was there are no snowflakes because this is built on Jenkins it’s on Jenkins LDS you’re not on a custom branch of Jenkins that you are maintaining from here on its it’s fairly supportable and then we can bring in some of the features that we’ve had from the Jenkins platform in terms of you know analytics metrics and and you can bring that up in to get insights into these these are gonna are all these masters running in and the key sort of paradigm shift is you’re no longer thinking in terms of like one master protein but it’s so lightweight that you can actually just say I have a new project that’s coming on board let me go and provision something for you and underneath it goes and creates a master for you that’s sort of the idea here so you end up with the because we’ve run this will be a bit cloudy you end up with a sort of a leaner ops team to manage this infrastructure so that’s that’s sort of the context does it make sense right with that out of the way I like cosas que to talk to the specific features and we can demo right okay so yeah so that’s the kind of idea right well it takes what does it take what do the people run junkie such a massive scale so you know what it needs to be basically a service so we the club is is giving you a software so in that sense it is a packaged software but I’m you know you in turn then operate that other service the rest of the company and I know as harpreet said you know we also came to be time to deliver agree getting the result very quickly which means we want to come in not instead of having like us a 48-point checklist for setting up we actually have got the script that you only need to you only need to edit a few configuration parameters and you get the entire a tiger environment launched in the in the order of the tens of minutes and then it’s building on top of all the existing features so they can’t be sinking splat form if you think about the fact that Arabs theme that’s run in the background only have hundred people I can can’t be in single hand when they managed a dental services they did I on a lot from the features of a building from the CEO see and so so that’s what we are building this feature for so another big sand custer operations and so it’s all available and then if you have existing build environment the build slaves that needs to be running on some existing card setup let’s say you have some specific game I the two team is supposed to be using for building them then you can use our hybrid car crossing so you can connect your entire Tyga deployment all the Masters inside it and I have them talk to this sucker Club as a build stuff to create this elastic billed amount now the keyboard of Tiger is this idea of the multi-talented master Asian slaves so the idea is you have a relatively static set of big honking boxes and we are software will multi-tenant them basically put each different masters and slaves inside its docker containers and then end up way we can derive a lot of bodies so the one of the obvious benefit is a deeply digital obstacles you’re packing a lot more things inside zynga boxes but it also allow dusted in a very quickly allocate post masters and slaves so for example have an API call that let you spin up a new master and then you guys come up really quickly and also even more soulful slaves where the time to provision is more pronounced because that’s the time that the developers wait for the bill to go so that is the dissimilar key benefit last point I’ll come back to it later now in the world of failure you are no longer thinking about individual masters but only thinking about them in aggregate right we are talking about the hundred or even add on two masters so analytics is becoming incredibly important in this town because you kind of the only way for you to keep track of them is by statistically you are aggregating them so for example babies the club is drinking some of it is you can get all the bills result i drove individual

master centrally aggregated into the aspect search and you can run queries and graphing and so on so it’s a great means to keep track of the progress of continuous diwali across teams now we the the other analytics we collect is around the health of the adjunctive masters and build slaves so again you don’t want to individually monitor these jenkins master that your users are using so what we did to do is the automatically collect is information and then you can spot the outliers without spending in time from the again the individual machines and finally you know you want to be able to show well which team is using hi Marty resources because you know you in a way is on the hook for running all these the big clusters of tens or hundreds of machines and that money that that needs needed to run them from has to come from somewhere so now we are trying to help you show exactly how many build is going on in this master now how many total combined minutes are consumed and so on so you can use that if the distributor costs among different customers that uses the tiger now also in a world of tiger if you are the operating tiger that’s a very different role from people who are running individuals in its masters so we did we recognize this role separation between the tiger administrator who is more interested in keeping the entire cluster is up and running a separate from the client master administrator so it disguise the second role is probably more embedded inside each team he is in charge of installing and selecting the right version the plugin helping the team of the goal processing the team to set up the dog that kind of things so it’s already qualified from the get-go so that he can you can start using tiger quite productively now from the resource point of view so this is how it’s going to look like so again like I said you hang in the middle you have a big clusters of like a computing resources the big honking boxes now this is where I come back to the one of the other point about the multi-tenancy so even though you’re providing this incredible elasticity between in boston masters and your slaves the total number of capacity doesn’t have to change all that frequently so in a son for example it takes like in the matter of weeks or even months to get the new virtual machine provision from the ipos that’s Heidi that’s the kind of pace awfully large company would operating but you can use that kind of relatively slow ID and the thrown around still provide this fast elasticity today to the users of Jenkins’s a service which is great um and because you don’t need it care that where the master has revealed slaves are running there’s a HTTP reverse proxy sitting in front so that’s part of Tiger and when the request comes in it be able to distribute the load to the right back pain and finally the behind-the-scene we need a big storage to be able to stall the Jenkins home as you know Jenkins needs a persisted file storage to be able to keep on so that’s the major components of tiger so imagine right Sophie you have the award codomain see I told example.com all assigned and then in this Tiger environment we have two big honking box is running three masters alpha Bravo and Charlie there’s also see Jo see running but that’s as acting at the brain so this is at the point in time I be certain and you know with the users it commented let’s say alpha dot C added example.com the front-end web tier will make sure that the right master gets the workload now we the build slaves are also multi-talented from the same pool of resources so if one of the master realized that though I need to do a new build I need build slaves to do it and then it talks to the brain of tiger which figure out one of the available knows to start the field slave on and it’ll do that and then it connect them to the junkies masters for the duration of the build and then it disappears again into the ether so that’s the kind of experience you from are no longer subtly configuring your slaves it happens all elastically now the great thing about this kind of location transparency the fact that you don’t need the care but where stuff is running is that the 50 something bad happens to wonder the master so let’s say you know the maybe the Ziploc happens or the memory went over or something like that and then how your notice is that the wonder the master is lost and so it will automatically ensure that the master of you reschedule the master problem maybe somewhere else depending on the availability of the ed cluster and then it’ll make sure that the right straight is attached to the right and get that going so in this way I guess as a a tiger administrator you don’t need it again have to worry about the individual well-being and a master because the cluster you automatically ensure that the stuff is running you only have to worry about this whole

computer environment so that is the premise that so what I’m going to do is actually I’m trying to be even more brave and try to set up the the tiger demo dive so so this is like I where we start these we provide is the tiger is in the single command line tool so what i can do is the tiger in it project let’s see workspace so this creates all the necessaries staff under this workspace which is pretty simple and then i’m going to edit this file which is the basically configuration file that specify the topology of the network so that you can say how many workers you need and so on and this includes the AWS credential so I prepared this file up front so I’m just going to copy the same here oh so it’s copied in here and then then I’m going to then I’m going to just run tiger in it in here so what this is fat masters doing is it skating it’s slanting the entire cluster studying or however many nodes necessarily front and all the backend and everything what’s that a maximum number of epc has finished I wasn’t expecting that alright so that was a failure on my phone but as you can see the kids this is going to start taking its own time so because this kind it takes a little more than now I figure that you probably don’t want to scroll like a watch show this text scroll by so I have a pre-launch the environment all ready to do the rest of the demo that that one was launching that on he WS right yes that’s the creating all the instances if maybe i can stop using bpc I since maybe I’ll be brave enough to deviate from the prepare the demo scenario under maybe i can stop using with this oh we can go for a kid danger all right maybe I shouldn’t do that okay you didn’t see that party so I’m going to close the third deployment haha yeah alright I need to be set back because i think it’s giving the coded so this is the danger of going deviating art from the stairs but anyway so this is a lance the environment that i created up front so after that can attend minute or so it gets to this point so this is the master brain behind it and so what i can do it for example is i can create a new master so here i’m going to say that’s the golf and then i specified am creating a new master so let’s say you know your new project is coming on board so you’re going to start the new master for it so that’s I just have to specify well I didn’t actually have to do anything in this case I bumped up the CPU slice so that this gets little this gets the run little faster and the behind-the-scene what’s going on is that the inner door brain is figuring out which of the available node we can put this new master on and then I it’s at that container and then that’s allocate the right storage I in this case I believe you using the NFS like a big server that’s running FS but we can we are working on supporting the EBS packed storage and then it will come up and they’d all do the proper negotiation with the new club is C 0 C and then prepare itself for the build so I hope it’s gonna start soon enough let’s say I’ll help it by deleting this old masters so I can also delete the old one and a needle or disappear let’s see so that’s so the this is tia the underlying cluster she could be as big as you need in this case this demo environment only has a small number of available machines but I imagine that but you’ll be running tens or even hundreds of them and then now it’s kind of taking longer than I hoped for so I don’t know okay so now it’s got nice so now it started and got connected and got this new URL basically so there’s a longest name here but it’s basically the point is and it is the Gulf thought something something so if I go there it’s a spa it’s hopefully starting a new Jenkins master here you know we be as a thai administrator you can specify the board and lodging industry run in this case it’s finding a specific portion of the

jenkins master but you can you know we should be able to let you choose our with reversions and this master that we stirred it doesn’t really have any build slave attached if you see it’s completely empty so it has no capacity run the build on its own because it’s supposed to get all the bill resources from the cloud I mean I’m sorry from the I from tiger so what I’m going to show you is let’s create a new job just so that we can see that the build is going on and so I want to do is just quickly make sure that this bill does happen somewhere else and I’m going to run the canonical like a sleep 10 as a build can we come to Southern deployment yeah I guess we deployed a new master so that is that what’s account a sweetie yeah oh no I did start a new version of thing so that’s like I guess I started in your tire so that’s you before all right as we’re talking about it’s already like I trying to bring any new slave on board and so it’s got the new field going here and then that’s you know from the from then known it’s basically like a regular game in so it’s doing the build and then it will keep the records so I’m going to do the fate / experiment here so what I can be is your unusually brave today yeah yeah well I’ve already failed once so what can go wrong so turning fear into boredom that’s that the idea here so what i can do is um so so i’m going to simulate the failure of the junkies master and you know drinking soda Davos I can’t really kill it from the UI so i’m going to go back into the app the back end and then try to sort of really more force process kill so the trick here is to find me they’ll find the worker node because you know i don’t need a catchy detail where this golf the new master that i created is running so I have to I have to I got three work knows i need to see where it is so so this is running nope this is not it so if it’s not running on the first worker node so they may see maybe it’s running on a second poker node so the only trick for me is to look at the timestamp oh yeah so this guy this master has started four minutes ago sigh I guess I would bet that this is it so I’m going to just like you know basically force kill this master by killing the process alright so if I got lucky then I should be able to reload this page and yeah so now you see that bit below this hanging here because the master event time so what’s happening is Tiger is noticing a lot but its master is gone now so you got to do something and then so it’s figuring out somewhere else to run it and then deploy the container and then doing all its magics to move the storage around to make sure that them to make sure that the service gets restored so let’s see i can keep reloading it to see if it comes back up and the jenkins does take a little bit of time to sort so i guess now it’s you know now it’s ready to work we can see if it’s i noticed early tends to move around ok so i guess it came back on the same node so it looks like this is the least loaded nodes so it’s it happened to be scheduled on the same box but you can the same thing exact same thing work if you lose the entire book so say you have a 10 the big honking box is running a dozen masters and then somebody pulled the power plug on that box then the whole thing is lost but no llores go sleep again the tiger is distributing that workload in pee everywhere else so again the neat thing is the handle this story is automatically so when it comes back hopefully let’s see turning fear into boredom yeah this part I’m pretty I’m pretty sure oh no they’re waiting for the docks to go yeah so the amount thinking can be so the yeah so you know even if you have not it doesn’t matter lost all at the same time provided that we have enough redundant capacity in the system you be able to basically operate the loss of this master and then cut them last so as a tiger administrator you don’t need you have to again worry about the health of individual masters right each team might see a brief disruption like this but again in at the end of the day they come back up when you see all the building codes that we just need to build all all these things are intact so the company organization of the whole has minimal disruption to so that’s the the the

agenda so the part of the demo here so yes I’m very happy to announce that the abbey no discovery Jenkins phone privates a solution is alpha hit the Alpha do this today now the team has been really working hard and as I’m very happy up and in the you know if you’re interested in talking to more about talking to us more about it and you can either catch me or have read or Patrick grandi the club in space and then they’ll be able to show you you know they play with you a little hopefully and I’m explaining your Heights put together and we are also trying to not to talk to a small number of serious prospects who are interested in a driving this product forward with asti no work with us to determine what it takes decoding vga so if you’re interested in signing up for that kind of beta program then we’d love to talk to you so you know also can buy are boosting and drop us a note so yes and then we are hiring cuz you know that my baton harp or collectibles is really sick greeting us on the hiring so he is looking for the more product managers you’re looking for the boundaries of all for sale so basically everything that any so we need a lot of different people this company is growing fast and associate interested in working sighs no so you should also do a person oh it’s the lamest sell on hiding i’ve ever seen I’m sorry yeah you get to work as a technology evangelist with Kosuke that’s all right now i can say that myself all right should I be what was it yeah so what I’m so what we show you in conclusion what we are trying to show here um maybe I think the doctor in Jenkins is a great combination if you’re building this end-to-end continually pipeline and also at the club is is not taking this crappy thinking splat form into two favors the team in addition for smaller organizations Enterprise Edition for even larger organizations so that’s where our current offerings are and we are working on this exciting new addition to your portfolio which is the Travis Jenkins platform private last edition AKA tiger that will allow you to operate rankings as a service in massive scale inside your company so i think that’s it so thank you very much and then talk to you thank you you