Logo TIB AV-Portal Logo TIB AV-Portal

Intro to Kubernetes

Video in TIB AV-Portal: Intro to Kubernetes

Formal Metadata

Intro to Kubernetes
Rethink scalable infrastructure with containers
Title of Series
CC Attribution 4.0 International:
You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor.
Release Date

Content Metadata

Subject Area
Distributing and deploying software inside (Docker-) containers for security, isolation and ease of use is the new big thing. But once you got all your services nicely wrapped - who takes care of all these containers? The open source project Kubernetes, originating from Google, helps you manage containerized applications, as the operating system of your datacenter, treating hundreds of machines as a single resource pool. This talk introduces the core concepts of Kubernetes, its benefits and its huge ecosystem and gives you an idea of how Google controls parts of their gigantic infrastructure.
Keywords System Administration

Related Material

The following resource is accompanying material for the video
Video is cited by the following resource
Point (geometry) Word Computer animation Maxima and minima Scalability
Email Process (computing) Hypermedia Bit Software testing Quicksort
Point (geometry) Overlay-Netz Group action Distribution (mathematics) Open source Projective plane Database Maxima and minima Product (business) Exterior algebra Computer animation Touch typing Automation
Web service Computer animation Bit Cartesian coordinate system Resource allocation Computing platform
Addition Mobile app Server (computing) Multiplication sign Cartesian coordinate system Revision control Uniform resource locator Web service Word Process (computing) Computer animation Right angle Smartphone
Addition Server (computing) Service (economics) Multiplication sign Software developer Virtual machine Mathematical analysis Bit Instance (computer science) Cartesian coordinate system Revision control Web service Data management Uniform resource locator Computer animation Integrated development environment Moving average Utility software Office suite Resource allocation Row (database)
Server (computing) Group action Open source Mereology Number Neuroinformatik Revision control Computer hardware Computer network Cuboid Data structure Codierung <Programmierung> Computing platform Physical system Moment (mathematics) Projective plane Bit Cartesian coordinate system Limit (category theory) Demoscene Cognition Data management Googol Computer animation Software Configuration space Abstraction
Web service Graph (mathematics) Computer animation Personal digital assistant Connectivity (graph theory) Core dump Product (business) Element (mathematics)
Addition Group action Information Namespace Software developer Virtual machine Line (geometry) Cartesian coordinate system Revision control Goodness of fit Process (computing) Computer animation Integrated development environment Commitment scheme Hacker (term) Term (mathematics) Operator (mathematics) Energy level Reading (process)
Process (computing) Computer animation Software Personal digital assistant Plotter Multiplication sign Execution unit Software framework Repetition Cartesian coordinate system Mereology Proxy server
Server (computing) Scaling (geometry) Computer animation Vector space Personal digital assistant Real number Green's function Computer hardware Cuboid Basis <Mathematik> System call Resultant
Area Game controller Scheduling (computing) Server (computing) Connectivity (graph theory) Insertion loss Workload Category of being Web service Data management Plane (geometry) Fluid Computer animation Software Object (grammar) Proxy server
Workload Web service Scaling (geometry) Computer animation Cube Virtual machine Logic gate Local ring
Windows Registry Computer animation State of matter Counting Writing Form (programming)
Data management Game controller Scheduling (computing) Computer animation Cube Virtual machine Parameter (computer programming) Cartesian coordinate system
Revision control Scripting language Stapeldatei Web service Uniform resource locator Process (computing) Computer animation Term (mathematics) Scalar field Office suite Mereology Sinc function
Prime ideal Scaling (geometry) Computer animation Commitment scheme State of matter Virtual machine Right angle Database Cartesian coordinate system
Point (geometry) Server (computing) Scaling (geometry) Process (computing) Computer animation Software State of matter Data storage device Database Bit Cartesian coordinate system
Open source Java applet Uniqueness quantification Multiplication sign Projective plane Data storage device Device driver Cloud computing Volume (thermodynamics) Cartesian coordinate system Template (C++) Mathematics Process (computing) Population density Computer animation Software Integrated development environment Causality Personal digital assistant MiniDisc Physical system Identity management
Computer animation Software Code Operator (mathematics) Plotter Authorization Planning Cartesian coordinate system Mereology
Area Scaling (geometry) Source code Database Volume (thermodynamics) Cartesian coordinate system Mereology Number Process (computing) Computer animation Software Operator (mathematics) Cuboid Right angle
Group action Server (computing) Service (economics) State of matter Multiplication sign Mereology IP address Direct numerical simulation Web service Roundness (object) Different (Kate Ryan album) Term (mathematics) Maschinelle Übersetzung Proxy server Address space Form (programming) Debugger Data storage device Virtualization Variable (mathematics) Cartesian coordinate system Uniform resource locator Word Kernel (computing) Computer animation Software Integrated development environment Lastteilung Right angle Whiteboard
Revision control Computer animation Multiplication sign Single-precision floating-point format Real number Neuroinformatik Social class
Sine Game controller Plane (geometry) Computer animation Set (mathematics) Right angle Cartesian coordinate system Mereology Error message Computing platform Stability theory
Area Stapeldatei Scaling (geometry) State of matter Kälteerzeugung Multiplication sign Image resolution Projective plane Data storage device Gene cluster Set (mathematics) Bit Volume (thermodynamics) Cartesian coordinate system Web service Computer animation Software Different (Kate Ryan album) File system Right angle Local ring Condition number
Computer animation Software Open source Internet service provider Data storage device Hard disk drive Virtual machine MiniDisc Cartesian coordinate system Datei-Server
Installation art Slide rule Greatest element Projective plane Mathematical analysis Cartesian coordinate system Stack (abstract data type) Formal language Compiler Word Computer animation Oval Internet service provider Computer network Right angle Compilation album
Server (computing) Game controller Scheduling (computing) Open source Connectivity (graph theory) Virtual machine 1 (number) Solid geometry Scattering Product (business) Green's function Energy level Computer-assisted translation Proxy server Booting Social class Addition Default (computer science) Consistency Projective plane High availability Cartesian coordinate system Degree (graph theory) Word Bootstrap aggregating Computer animation Cube Metric system
Slide rule Server (computing) Scheduling (computing) Open source Multiplication sign Workstation <Musikinstrument> Motion capture Solid geometry Client (computing) Mereology Dimensional analysis Computer programming Product (business) Revision control Medical imaging Web service Coding theory Forest Software testing Macro (computer science) Physical system Form (programming) Installation art Scripting language Electric generator Software developer Data storage device Counting Database Instance (computer science) Cartesian coordinate system Process (computing) Computer animation Software Software repository Cube Order (biology) Moment <Mathematik> Right angle Figurate number Asynchronous Transfer Mode
Computer animation
so hello everybody our next talk some introduction to Cuban the Cuban users and itself by max so please welcome our speaker few there'd Maritornes Samuel the point is that it is right introduction to command this to today will run everything scalable infrastructure which contain let's see how many buzz words with
implement this and a
max I'm a test engineer chorus I'm working with the Prometheus team and their commands thing and helping them giving the quality of and we have an entire hours so I don't like to have this as interactive as possible so if you have questions feel free jobs during the during the talk offers talk but I'm also here a little bit after the conference so that you can ask me afterward sort of reach out to me over any and social media or via e-mail has appreciated as well OK and why somebody from
Chorus standing up here and what the scores have to do with uh containers and communities and so choruses accompanying sentences group and that is also based in New York and Berlin uh and um what we do is we secure simplify and automate container infrastructure and now that that might not tell you a lot so just go over some of the stuff we do so
where we happen to rest products and I'll not go into detail at the free and open-source Conference on our and the 1st products but I wanna touch briefly
our open source projects LB container Linux and as a Linux distribution very minimal it's distribution that you just from your containers and on then as a container engine you can think of it as an alternative to Dr. and flannel for example as an overlay network or at city is a database have which is now pretty much the brain of couldn't as and these are our open-source projects we also contribute to a bunch that we Prometheus think notice of you guys are a different point in your account reappointed both uh so I know we meet his income others and and so we're very involved in upstream projects there OK and that's that's for chorus so what
is communities that's going to be that the title and the topic of the talk today if I want you to take anything with with this topic that would be at the platform for running the applications that nothing very complicated it's so whenever you have a chit-chat you can just she out that sentence but but we wanna dive a little bit deeper into given the and that so let's 1st look at what kind
of problem is is actually trying to solve the and I wanna do that with little
example here so let's just imagine we are a start up and we have a really really great idea how we have our own and we do in augmented reality at so we walk around the streets and then we find creatures all over through our phone we can see them and then we throw balls at them until we catch them and uh we no swap and swap them and fight with on me which is the that's very popular so we got allocation service at the very beginning so that just returns true or false when given GPS coordinates that's very high sophisticated so that and
we run that on and on and on an almost over right we permit that to study in on it and do not get installed the dependencies we're good to go we got our 1st as service running and then in addition our started slowly evolving at people 1 more than just locations so we develop the user service and we develop in something at x service I wasn't creative enough and and
we put those on service as well to so now and we have 3 applications each on a separate server just managing the dependencies via the app gets so it's really easy but not be too complicated it whenever 1 or lot new version would take out take down the old 1 and to start the next 1 down time is not really a problem at the beginning now no this is getting really popular people really like it and a lot of words you running around was a smart phones in their hands and 0 in on the job
and so we
probably have to scale this so we don't only need 1 an instance of all of our applications but the probably need multiple as
instances of all of our applications and are slowly it's getting a little bit now complicated so 1st of all dependency management how do you do that on all the servers will you have to keep all the reasons that under the same and whenever you wanna roll out a new version of the Office's agent every single server take down the application put on a new applications now you be solely worry about the down time you might want reproducible environment so you want your uh developers not to say anymore it works on my machine um bending annotation and you break 1 networking in between the services you want uh the location service maybe to know where the user service actually is so do you configure it's hard coded every time you get a new server and but on the In addition you might wanna monitoring around all this so you want to be woken up and then either something is going very very wrong the and and occur plenty of other stuff you wanna do you so it's actually a lot of work 3 servers crap might not be a problem at the beginning but more and more you get it's getting more and more complicated and maybe that is all developed under different technologies that like 1 is written in go once written in Python and so on and suddenly you can't run the the patient service on the user's server and the user application under the cations server because dependencies are simply not there and now it's all incomes utilization I want to use as much of my hardware as possible so maybe the user's service that some analysis in the nite but really does need any rows resources during the day so that could be taken by the location server but then again our dependencies are a problem so we can't really run allocation service under the user server of a bunch of problems have the following year for the solution so this suggestion I am making here
is why we have specific servers when we can abstract in between them so we just have servers that all configure the same and we have application on top we need something in between that the abstraction us extracting from hardware expressing from network and obstructing away from processes and as you might guess that will be tremendous today and and today i wanna explain what this boxes and so that you can explain to others as well and maybe even use it set it up yourself a little bit about communities
this week already have our number 1 sentence it's a platform for running applications it's a platform for running application abstracting away infrastructure that's the basic thing you have to think about when the the name for the net is concerned limit of history I would estimate that is all about where there's a kid actually come from but comes from cool and it turns out who will actually has a lot of experience running containers if done that says since a lot of years and actually a lot of the main technologies that we today using linear scramble to run containers management containers IID contributed by tool for example see groups the concept of the groups so who has a lot of experience with that and in 2014 their open source their way how they manage the infrastructure but they don't do that by open-sourcing encode but they'll do that by open-sourcing ideas in learning how and that is picked up by the community and fighting structures of Google on get up and now it's an open project developed by a lot of companies altogether it's very influenced by pork as assets is the learning and ideas from google this is their internal monitoring system at the moment that thing as well and um all this open source project evolved and in 2015 version 1 was released and with that version 1 it also joined the scene CFD project so now it's not part of Google anymore but it's actually part of the scientific cognitive computing Fundación cooler is still heavily invested in it as their decay xt running on this so it's very important so for them OK let's
dive into a beloved of details on and thing was 1 element graph like what's your experience with it so here who has ever that they've read about in yeah correct that's that's legal debate is ever run kabineta so have intact with chips or something like that Hong OK cool and who's actually using communities in production right now the 1 2 3 end of the cool again how after this talk all of you just changing everything and that is going to core concepts so it core components so you 1st and stand all this
well 1st of all and let's go back to cases service I wrote that very high sophisticated case servicing go of course and I and now as
a good hype Driven Development start out by rapid contain around it and that's hopefully I think I don't know only read Hacker News but actually but some spots into this so what I do is I take my application and put everything that application needs inside that container that would for example be the goal line environment and and
from these solver containers what what is that
actually I talk about the groups earlier and where namespaces the lens galaxy has no clue what containers are this just a higher level concept that you build from smaller concepts and being the namespaces what can a process see and what kind of process actually used I'll not go deep into containers today that's entirely topic and I'm very happy to point you to to what's more information now in this and meta-level transition is done by for of darker and rocket and our container runtimes that we can use what does that help us this container idea with being doing Soffer ever since Why do we need this saddening both 1st of all probability I said I talked about works on my machine problem and you want your operator you developers to probably run in the same environment then i it is in the and deployed so now your developers can develop in the same doctor container and have the same versions around In addition in terms of operations it's very important to have everything isolated we were able to do that on machine level was virtual machines before but now we can drill a lot deeper into that on process level and lastly will 1 resource accounting we want to see who is eating how many researchers and we want to be able to restrict the spot so all of this and all these things now combined now in this uh height of containers the commitment is the stop
here clean is not only wraps a
container around our application but it's reps another concept around it node be
pod you might be asking why another concept command we already have enough uh upon is it that this plot
this smallest deployable units that you can possibly have in communities and the idea behind it is sometimes you don't only 1 to have 1 process so for example in case of our the cases server we might also be put in into network proxy in front of it or some loving framework into the front of it so something that has to be deployed with that part I with that container every single time so why don't we all put it together in 1 plot and then we can schedule a pot somewhere so that's just a simple concept divided predict and
remember this vector after jumble of it and they will come back to the the OK so we wanna run
offer what do we need for that of course hardware that this result by a bunch of servers that's for real of fun and let's call this 1 the masters and a mastering through that this it's just a server it can be very metal or it can be the and or you can get really creative and and it needs linear strong that's the main idea we need a lens scale as a basis and that could for example be container Linux but that could also be relat daddy and and so on now I am on the master we deploy 2 minutes and as you say you could this just think of that's just that 1 black box or in this case the green box and that's
delivered into detail so that thing is called a control plane and their commanders control plane and it has the API server and has the control Mandarin area the schedule and the property of go below that into detail of each of those components but just to wrap it up real quickly the API service the thing you talk to the controller manager is actually taking care of all the objects inside a cluster the scheduler schedules the workload on each node and the q proxy for example takes care of the network will die fluid into detail
running 1 losses prior boring so we buy more service we biased worker and have the work again is
just a bare metal machine any the Linux and that is sexy well workload was running on you can still run the workload on the master node but you probably wanna run when you run an analysis and on a scale of more than just to service gate here again we
deploy a little given this that is called the
cube let it listens to the recommendations and whenever the bickering as as they please deploy something here then the local sense OK I'll I'll do that and it starts up on your OK uh 1 more and boring
it's there have couple ones and that we have our infrastructure cool it to go that's deploy some implications on this so
given that has has this very declarative style of you interacting with the cluster you don't really tell a to what to do or where to get there you simply describe the state you want to get to because command is knows a lot better how to get there then you because commanders knows entirely what its current status and how to get to the next 1 so what you do is you write deployment yeah most you can write deployment Jason's as well it's just the form in the end and and what you do is you write this deployment yellow you give it a name you give it a replica count so photon now 1 of our run 3 replicas and you give it a container the container you just push to cut a registry like for example Doctor of and I'm sorry I skip that love fast uh we
give that deployment now to kabineta and namely the argument as it and the community's
API 0 picks it up and say that uh the controller manager exert up and goes over it doesn't work then that the scheduler sees all there are 3 replicas but they're nowhere deployed to fix it up and then it's schedules those on the worker machines then the cube let's see how something that's scheduled on me and so let's start that on on me and that's that's that's now you have the applications up and running just by giving that Gamal to you it's over now I talked earlier about rolling
deployments or how you can now roll your versions in your cluster which is very difficult if you just have a bunch of batch scripts doing this and if you want a scalar so in terms of tremendous it's very declarative you just change the version of the documents and don't forget to that registry and
uh give that bananas and couldn't and it knows what to do from here and will do rolling release so slowly move all the parts over to maneuvers and that's it that's all the deployment process now you might be saying OK the location service come on you give it on latitude longitude and it returns true false that's really not difficult with office since years that's nothing new at all so everybody can do
status and everybody can do state of applications and that's what Prime Ekman is really shines that right because it's not that difficult but statelessness easy
stateful is actually hard and I don't wanna just talk about status application wanna go evident into stateful applications and how you could possibly run those and commitments so the 1st look at the
problem status applications we don't really care if they're dying in the start of a new 1 that's stateful applications for example young my scale database it for example next to the local loop uh this and race that stuff there and if this fails all your data is gone or you maybe restart the machine but you definitely of downtime once on the so
I suggest a different approach and some of might look mad at me here and I see propose having network storage and putting that idea of replicating and so on into your network storage and having just mount points into your server and you my scale database attaching without so now whenever my name my server that's a dies we can just move the destiny of server and start the process that again and we're happy we add a
little bit of downtown but we can optimize that happened really quickly that's the main idea how to run stateful applications and governance the um would as here introduces as
different concepts that's all persistent volume we Indian state for as say somewhere so we need a system volumes LP network storage it can either be statically or dynamically provisioned so the idea is you know you get to the is a bunch of this or Europe for Java running on a cloud provider and then as already knows how to spin of new disks on the US it and these drivers are for a T C I think AWS there'll bunch of open source projects that you can integrate with so this issue to community around the and of course we have a new concept not deploy and gambles but now we have faithfully animals um and these are just forehead stateful applications and their key takeaways here is all stateful and unique network identities of going to detail how that works and persistent storage of course now the application and stateless environment you don't really care what application it is and where it weightless than a stable environment we actually care really care about so how would this look like again in the
this idea all you give that the stateful young and you can look at this we have the change here is that we have a volume claim template and thereby you distracted that what does my application need so in this case we need storage cause of anything that could be as a steel spinning this but it was not very specific here and let's say 1 and and then the this budget myself with humble somewhere starts the process and amounts this going there and you get to go is go through the example this thing dies and then you can just start of a new 1 on the same note that say the entire no dies or on how you cut a cable or something then you can just start this process on a different worker and mount the same disk in and now I talked about network identities network density moves from the other 1 down to the lower 1 see applications really notice the down time but they don't notice that the entire data is gone as it's not go go
and this is for a basic staple applications is also the idea of operators I went through that delivered to my previous talk the ideas that we have all that knowledge around how to operate stateful applications which is really difficult and then people but that knowledge into code and road operators and now for example be at the at the operator on the
Prometheus of we can do without any any questions so for also cover the network part on so the author you know the yeah this work was a to plan was yeah going to lot yeah I to do this this thank you some as to how do we applied the concept of plots here in
this scenario and and here in the same area pretty much containers in parts of the same and we just care about whole process right at the beginning it's just 1 MindScope process you don't have any Cyprus next and here that contain would just be in a pot and they would just be 1 contain another 1 called and then you would amount of volume into the container itself and let's imagine for example we want more logging around this so for down we want a lock whenever there was a request on my scale database then we might not want to change my scale source code but we wanna deployable application next would just place it in the same pot inside a new container and then we would have to containers in 1 bond so that doesn't explain why you hesitated so all of that so that was not not then you don't put in the same box only if that number is always match that if you have an application that always list with and so with this the other application you always deployed in one-pot if their numbers are not equal you don't forget it the right and yeah hello move onto networking and now we can go on all questions again and I'm now going more expert details because this OK we got the the
operators covered right and now on
my talk about networking and that is actually a huge pain and it's not only storage but also the networking in total and and the
promise of pods can really move around in our architecture over and over and over again they can die and come back and we don't really know about we don't want to know about it because human as manages a lot better in a lot faster than we do but the problem is how do we tell 1 to communicated to another part if they really don't know where the other part is if that is if it's such a fast-moving infrastructure so here we introduce
the idea of services you can think of services as pretty much just a proxy in front of it and it's basically just groups the parts so example of application part and i'd apply 3 of those facts lead occasion parts that but this location service in front of that and whenever another application needs to talk to my location service I just pointed to that service and I don't really care which part that is specific applications talking to I just wanted to talk to the type location happened these services the idea is that you create a service and the service will keep the static the inside a cluster at all times and will you can obtain that he as a different application by environment variables DNS you pray wanna start out with environment variables I which are automatically mounted in your part but diaper everyone go over to the innocent so how this this work and how can 1 IP address to be in the entire cluster and how can I call that purest form everywhere and still talk to different parts that's a little bit strange crazy we don't do any board matching here in communities because every container has its own IP Indian so let's say we want have affirmed in part and that talks to all that back and and and somehow the front end departed so it goes over the network and just as the requested it directs the IP address in the environment variables inside the container and now does that there are no the 2 that I here so what actually happens here this service the front and service really doesn't know anything about what happens in it doesn't need to know about anything so it things just talks that the address in something is answering from here but what is actually happening underneath this we build on a virtual network on top of our normal network and I think you proxy well the traffic goes through and talks to the API server and gets the part that piece of that service I the and right said into like a coffee fruits and now they're linux-kernel thus a load balancing of translating a server-side into part at least so the of front and doesn't know anything that just stocks that service and then there's kernel automatic translates to all the other piece and now this would look like that and they would just for example they the first one of review round robin I would just after the 1st word in the next time it might even talk to the next 1 so that is very random instead in terms of state classifications OK
and how can you get started on this in fact benedicite here and well 1st of all I think to to really tried out in a quick way you can just start of need you need you've is really nice and it's just a VM on your computer and it's a single node cluster you up and running right away and you too can deploy stuff and of course you bound by the requirements of our you resources by of the lot of the next this course tectonic which I want to talk a little bit about in a minute of extra time and of course the hosted versions like GKT or Hubei ATM that you can use to spin up cabinets classes OK that is still
real quick recap and I'll go over questions and then if there still people I can tell you about more than 1 ways how to and when 2 men is in a different way
so our sins again given as the platform for running applications are you have
your application in Europe that in a container you can grab multiple containers in 1 part of it they're still always present with each other that is your infrastructure you have a
master yet the control plane on the master you have tubelets on workers you build that infrastructure you write deployment Gamilus and give
that to command is and you write stable sets young house and give those 2 minutes In the end
command as is just affecting way infrastructure
right for the pool and yeah although questions now and if you still have some and yes please feel free to call the yes and yes all the by scaling that's because right yes so repeat the question and an error because a right of in the go little more into detail to the right is all this
long right for you what forget so the question is do we have what happens we have race conditions if we have 2 containers mounting the same volume and then again that is in writing of us offers Alterman as I think cannot solve everything that's pervious solution for that situation out there that some users of but I think the idea is that you don't really want to have to my scale databases writing to the same file system or if you do you really want to them to write in 2 different uh folders or something like that so you have to take care of that race conditions and what I would suggest that you really have to network storage is that you can attach to the 1 container so each container has its own network price that the I hope you replicated data not this way but actually by talking with each other between the data and I think that's that's going to detail afterward afterwards I think there's a lot of the difference in areas have to do that and yet sure this is the 1st of its own you know you have to build it yourself but LOI of we still have time I can show you a project where you can just run 1 command and it builds everything for you break time had been you had to hit him if you may know on demand right 0 no who who was 1st very and go-betweens you all all the all the how dramatically again let's let's push this question a little back and I'll talk about how chorus things best how to bring up the net clusters that right yes you all all and this all out and do I always say that deploy service in front of the pot and I really depends if for example if you have a batch job and that matched up nobody needs to talk to that you don't really new service in front of it but uh whenever you need to talk to that thing you very when searching for and so full different story around stateful applications but I guess I can't cover everything in the park refrigerator afterward sort yeah kind of I as our near the so this is really difficult set it so it's a lot of of in the the in and it's the world Alex and so the the question was do I and what happens if I have a local local this local volume claim and I'm now my in my no dies and now I have to move it to different nodes right and so I think and brave Mr. explained that room where let's go back a little bit so you don't have local
storage but and go fast enough yeah
network storage right so you don't really you don't care about the the hard drive
on you know it on your machine but you that that the managed for exam by a 3 something by you cloud provider or something you build yourself was open source solutions and then I demand network drives into so whenever it now there's no dies we have never written to the local disk so we can just move the application to a different server and mount the same network storage in there and we have we don't have to care about moving the data further there are either dead or something and I'll just go into a level of details of course runs this and if there's the question them around afterward I think that's where more helpful and
skull these are all the slides
and so we contributors to the upstream project of of anatase and we have a very opinionated way how to run communities and so we know how to make it most resilient as possible and I wanted to cover really quickly how we do that so there the 1 click solution that the
proposed earlier and might have over integrated love it but um it is the tectonic in solid and don't be afraid of the words tectonic in here in the end you can still spin up just a vanilla its cluster with it as so it is the the chorus way how to run the bananas and uh this X build was terraform on our even below the bottom was terraform at the very nice way out to integrate with pretty much any provider the so you can extend it nicely how we're right now supporting because mental better and there's an code focal stack and there's gonna be support for the UN where so what does this installer actually do well at this and it starts up command that is in a very special way and that is the idea of self-hosted is anybody familiar with self-hosting compiler for example or self-driving compilers the anybody ever heard of so the idea is that if you write your own language and in the end you can compile your language with language so it's had chicken egg problem here in what we do is we sell pose so um the idea of analysis you have the application ID can nicely scale it up and down you have great tooling around it if it dies we just restart yeah know why don't we just don't not just do it with our applications but also without cluster and that's the idea of self posted to run for the net is inside itself in the now that's a chicken and egg problem and I think over there so someone quite surprised and I'll go into detail it works but 1st of all we have different stages so for example the innocent
add-ons that's easy to define your own class it's in the end just a deployment then it's getting more difficult than level 3 scheduler control and proxy because tried to schedule a scheduler without a sketch of that is really difficult and then the 2nd level where we are at right now with the tectonic in solid CAT server as started by the same API Server the Congo experimental with at City and you can go crazy with cube lives but we're not there I actually self-posting triplets so how does this work
well it's an open source projects called boot you and it helps you running as insect minutes and I wanna go through the steps how this actually works so on your machine and the blood add to blood just being the Damon that can in instead of containers and that's all that the kid needs at the beginning you can think of this you just as a 1st just creating stuff so 1st of all we spin implement it might be difficult but if you automated that's actually degree and ended spins of a temporary it's the cluster a temporary at the server and temporary sketch just our temporary components just to bootstraps once we have that on that Cuban and fast on our bootstrapping green of the we start our actually production running kabineta plus so we start an additional it's in the cluster we start an additional API Server study an additional schedule and so on now this ADI server sees the temporary API Server and sees all that to some not gonna do anything right just idols around we do leader election by default so doesn't really do anything this is just sitting there are the same as with the yet city in the scatter now what we can do is move all the data so all the the brain of from at city over to the word long running at City and then we can tell kill the temporary server these the but production API Server means that Cecil there's no of API Server had better take over from here and then it takes over and now you have fewer communities cluster running inside it for the next class and you might be asking why sealer confused faces here and I will 1st of all we have small dependencies and we run out of stuff how run our occupation um we have deployment consistency as a set um we can run everything the same way we have featured a easy introspection brings a lot of tooling how to introspect into your applications and now in addition you can see the interest backed into your cluster itself like for example metrics pollarding and so on then you can easily do cluster operates on on a lot a lot of people you're running common in production but it's actually should paint how to updated right so I + upgrades are just like application of great you change your deployment gamble and give it the analysis and that takes the rest and add in addition we have a high-availability notes so for example we can scale server now we can scale applications we can also scale the ice over now this high availability Wheatsheaf by the election as I said if the ones degeneracies 0 there's still the temporary API as controller I won't do anything and we do checkpointing so the sound we restate checkpoint the API server every now and then on to this so whenever a node for example restarts we can start up the same API so again that is for example how you can now update you command as fast as you just use your tooling that you use for command and just and API server and from there you go OK that's my in my
and out the most specific arenas close and I think there are a bunch of questions around
that I'll just finish my slides and then you can ask questions around that right so I we're
hiring if you wanna get involved in the disk and you're going the land hiring for instance as well for a Prometheus developers upstream far too many engineers protester dimension engineers FEE much at all over and up
and down to 3 to reach out and I think now
it's time for question at the love 20 minutes is very OK OK I'm going to repeat the questions right yes please so what do the of the this is all in I the the we know that some the a wall in order for the war but yeah so that what and was you will so again so well what why would they use to coming in solid why don't they just use food you why don't they used to do that when they use cube ATM and all the other tools all they all that's happened on in a country really represent the other tools here on for sure check them all out and just go with tectonic installer and they're dying and so as I said we know how to run command that is so the entire idea of self-hosted you get by tectonic installer and you get the whole story of how you get chorus and the need to go over container Linux underneath and so you don't need to buy and operating systems and so you get the whole package of forests and tectonic installer I think they're cute ATM is probably more basic approach and less opinionated to the entire package either per be a lot of talk shows around comparing those tools and I don't think I do a good job comparing them here OK other questions these this is so if if this so we are low on a low testing there's a whole lot but and as you know where production company where the product company we're not actually opposing this I'm not able to talk about my clients here horrifying are for guys yes to the of what and so are we using terraforming our 2 continents solid and yes we are and that is and we chose that on purpose so that people we don't just give them tectonic installer and they and can only run OWL version that they can actually just all this that's all open source you can play around with the terror form and it's in the end just terror from code theory of I the you and I know do we generate tariff from scripts no we don't we don't do any macro programming here which is leverage terraform itself so we use telephones it and you can extend it you can you can clone the github repo and use it as well we have a U URI installers well if you wanna go through that it's sperm easier at the beginning that so that helps you bring up a concern AWS for example OK of oppression yes please you don't want to buy a new so is means whole thing the it's not this all the work of the you you you all of us a lot more of these all of your using your home it was so um if I just my replica count can I say that it's not supposed to run on on 1 node or can I say that my database note please run on this really high storage nodes and so on and yes for sure that and a lot of concepts and given that you can actually just to specify where what should run and for example you can also just for example we would only 1 3 API servers and all them running on the same note so we can specify hey and if there's 1 API Server don't schedule a 2nd API server on the same note so you can do a lot station toleration so some current any other questions yes please that will be the kind can Menendez handle applications there clusters and synchronize their captures what would be the wrong 1 not OK and in the end uh there's no real like it's just a virtual network so in the end it's just the network that you can use and leverage yourself but you have to pay attention of course I have to talk to the service not right not just single parts for that to take the torque to stateful set the value we can go into detail after which was what you on OK other questions that again right and I'm here live longer you wanna reach out and I think that the only thing that you need to know the basic stuff of and confuse you too much about the complicated stuff and selflessly right figure image for the attention and thanks for the Boston Tea question