Introduction to container orchestration with Kubernetes

Video in TIB AV-Portal: Introduction to container orchestration with Kubernetes

Formal Metadata

Introduction to container orchestration with Kubernetes
Everything you need to know for your next job interview
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
Containers are not new and you can hardly find a job in IT nowadays which doesn't involve dealing with them one way or the other. But once you got your hands on the container technology you are inevitably run into the container management and orchestration topics. Kubernetes is a more or less vendor-independent orchestration platform, which provides out of the box automation for many standard infrastructure tasks (scaling, load-balancing, scheduling..).
Keywords System Administration

Related Material

Video is cited by the following resource
Distribution (mathematics) Data management Computer animation INTEGRAL Software developer Telecommunication Closed set Connectivity (graph theory) Continuous integration Product (business)
Data management Computer animation Multiplication sign Software developer System administrator Combinational logic Energy level Water vapor Product (business)
Point (geometry) Service (economics) Computer animation Software developer View (database) Generic programming Maxima and minima Cartesian coordinate system
State observer Kernel (computing) Computer animation Different (Kate Ryan album) Code Parameter (computer programming) Physical system Number
Implementation Random number generation Software developer Equaliser (mathematics) Source code Debugger Cartesian coordinate system Software bug Kernel (computing) Computer animation Hybrid computer Operating system Energy level Right angle Analytic continuation Discrepancy theory Information security Physical system Spacetime
Kernel (computing) Computer animation Equaliser (mathematics) Real number Software developer Multiplication sign Video game Cartesian coordinate system Continuous function Product (business) Physical system
Medical imaging Mobile app Mechanism design Computer animation Computer file Java applet Network topology Software developer Quicksort Data structure Cartesian coordinate system
Medical imaging Computer animation Computer file Multiplication sign Binary code Object (grammar) Compilation album Physical system
State observer Context awareness Weight Shared memory Data storage device Volume (thermodynamics) Bit Directory service Continuous integration System call Frequency Computer animation Software Integrated development environment Network topology Physical system Reverse engineering
Windows Registry Building Code View (database) Source code Execution unit Continuum hypothesis Continuous integration Product (business) Medical imaging Mathematics Causality Different (Kate Ryan album) Energy level Software testing Analytic continuation Greedy algorithm Data storage device Cartesian coordinate system Measurement Process (computing) Computer animation Integrated development environment Personal digital assistant Network topology Video game Cycle (graph theory) Resultant Library (computing)
Medical imaging Data management Computer animation Computer file File format Different (Kate Ryan album) Image registration Cartesian coordinate system Product (business)
Collaborationism Medical imaging Process (computing) Computer animation Projective plane Statement (computer science) Website Energy level Online help Software framework Object (grammar) Computing platform
Windows Registry Autocovariance Software developer Virtual machine Set (mathematics) Image registration Instance (computer science) Food energy Dimensional analysis Formal language Medical imaging Arithmetic mean Computer animation Term (mathematics) Summierbarkeit Analytic continuation Traffic reporting Abstraction Row (database) Physical system
Injektivität Point (geometry) Data management Computer animation Query language Object (grammar)
Point (geometry) Group action Multiplication View (database) Connectivity (graph theory) Set (mathematics) Cartesian coordinate system System call Number Hysteresekurve Word Computer animation Personal digital assistant Object (grammar) Error message
Point (geometry) Scheduling (computing) Group action Service (economics) Scaling (geometry) Content (media) Combinational logic Set (mathematics) Parameter (computer programming) Instance (computer science) Cartesian coordinate system Flow separation Machine vision Number Revision control Category of being Computer animation Strategy game Computer configuration Different (Kate Ryan album) Website Object (grammar) Social class
Point (geometry) Default (computer science) Service (economics) Software developer Range (statistics) Combinational logic Set (mathematics) Cartesian coordinate system Rule of inference IP address Number Workload Film editing Process (computing) Computer animation Software Different (Kate Ryan album) Internetworking Energy level Whiteboard Object (grammar) Abstraction Task (computing)
Point (geometry) Ocean current Service (economics) Wechselseitige Information Chemical equation Combinational logic IP address Direct numerical simulation Mechanism design Computer animation Operator (mathematics) Energy level Lastteilung Table (information) Form (programming)
Service (economics) Interface (computing) Range (statistics) Client (computing) Mereology IP address Rule of inference Arithmetic mean Word Computer animation Software Different (Kate Ryan album) Point cloud UDP <Protokoll> Data conversion Whiteboard Routing Computing platform Physical system
Functional (mathematics) Computer file Wrapper (data mining) Interface (computing) Multiplication sign Client (computing) Representational state transfer System call Formal language Word Computer animation Computer configuration Cube Video game Configuration space Utility software Summierbarkeit Object (grammar) Whiteboard Abstraction Descriptive statistics
Point (geometry) Default (computer science) Service (economics) Combinational logic Line (geometry) Cartesian coordinate system Mereology Number Word Computer animation Computer configuration Object (grammar) Whiteboard
Laptop Point (geometry) Service (economics) Mapping Multiplication sign Workstation <Musikinstrument> Independence (probability theory) Translation (relic) Set (mathematics) Cartesian coordinate system Mereology Medical imaging Process (computing) Computer animation Software output Whiteboard
Point (geometry) Medical imaging Scheduling (computing) Service (economics) Computer animation Software Core dump Object (grammar)
Point (geometry) Digital electronics Service (economics) Computer file Combinational logic Set (mathematics) Mereology Computer configuration Different (Kate Ryan album) Database Energy level Office suite Error message Computing platform Task (computing) Social class Addition Mapping Software developer Data storage device Content (media) Volume (thermodynamics) Instance (computer science) Cartesian coordinate system Measurement Process (computing) Computer animation Personal digital assistant Hybrid computer Configuration space Text editor Object (grammar)
Point (geometry) Implementation Service (economics) INTEGRAL State of matter View (database) System administrator Set (mathematics) Similarity (geometry) Continuous integration Disk read-and-write head Mereology Element (mathematics) Medical imaging Term (mathematics) Different (Kate Ryan album) Database Energy level Computing platform Task (computing) Area Software developer Cartesian coordinate system Data management Process (computing) Computer animation Personal digital assistant Order (biology) Video game Object (grammar)
Point (geometry) Dataflow Software developer View (database) Projective plane Expert system Set (mathematics) Control flow Bit Product (business) Computer animation Database Software testing Social class Asynchronous Transfer Mode
Word Overhead (computing) Computer animation Code Projective plane Interactive television Configuration space Number
Medical imaging Computer animation Integrated development environment Software developer Source code Computer hardware Sound effect Number
Time zone Computer animation Software developer Schmelze <Betrieb> Sound effect Instance (computer science) Mass Student's t-test Social class Product (business)
Service (economics) Computer animation Vector space Instance (computer science) Scalability
Pulse (signal processing) Scaling (geometry) Service (economics) Computer file Different (Kate Ryan album) Data recovery Software testing Endliche Modelltheorie Instance (computer science)
Arithmetic mean Computer animation Integrated development environment Link (knot theory) Projective plane
Service (economics) Software developer Set (mathematics) Staff (military) Instance (computer science) Cartesian coordinate system Product (business) Data mining Medical imaging Arithmetic mean Computer animation Right angle Social class
Computer animation
Point (geometry) Laptop Group action Service (economics) Computer file Modal logic Set (mathematics) Mereology Rule of inference Number Product (business) Strategy game Different (Kate Ryan album) Energy level Proxy server Multiplication Autocovariance Software developer Forcing (mathematics) Staff (military) Instance (computer science) Category of being Computer animation Object (grammar) Routing Resultant
Process (computing) Computer animation Content (media) Whiteboard
Computer animation
so let me start 1st with a close in my and my name is Alexander federal and my Iosing examines Book more maybe you can see me in some other places under his nickname so I'm long-term must adopt and formally I worked in neuritis OpenStack distribution managing really is of the but distribution itself and this will component of it encourage them working as continuous integration Europe Q of and and as no 1 knows what continuous integration in your is that's the person who tries to make sure that the development workflows feeds into a feeding into the Deployment workflows using production so what kind of trying to help developers and defined the of communication way of the way to communicate and to improve on this integration by plants so if you have questions comments and discussions 1 to discuss something and then you can find today and tomorrow it's adorable you can write
me and you can comment on my small at medium so today
we're going to talk about a lot of stuff but I'm mainly let me explain the subtitle so the and obviously this talk is very entry level and it's totally not enough water managing grew in this cluster or for like a real working with we with much what I meant by this time title is that whenever if you're a developer a sysadmin or even Product Manager of product on you need a certain base level of how you know how combinations works so you will be able
to talk with people about who burned and you work and the services so this and daughter presented safe minimum for anyone who's going to work around this topic and I hope you will enjoy it so that we're going to talk about continuous and communities and as a missing engineer and 1 to give it a different context and start from the very beginning and for us to question to the audience who works with Dr. awesome and toward the continent on something that should be called so I'm going to start very very basic stuff I what containers and what have the there's is a lot to talk about that uh mainly for the purpose of from the point of developer who created his of application or this from a very
generic high-level point of view middle mission
has its own kernel parameters are the container had uses a cost kernel from a you this kernel from the host and
this is the main difference in the urbanization of containerized in the theater like courses so these main different solutions has to 1st note that containers are
generally not secure in the it as our continuing user-space has direct access to host kernel that's definitely not secure situation and you can also execute code a whole system from a continue generally there more to the topic but this is the observation number 1 and observation number 2 just it containers and generally viewed it's promised by container I developed 1st
of the year like evangelist of containers systems
that container a cross-platform easy to move from 1 system to another but generally you always need to have in it containers are that represented a very hybrid operating system you take an arbitrary kernel you end arbitrary user space and you hope that
it will work it had it generally doesn't but not always so the generic takeaways
I'd like to point for like a
general overview of for a container is is that 1st of all I used only trusted sources steel never trust user in the inside a container unless you invest a lot of our resources into research and container security topics and they get there will be talk about continuous security right after the lunch here and as we have this kernel and the space discrepancy although we shouldn't rely really on kernels system-level level features in a container generally it is usually container applications are being good of essentially user-space applications like the front ends or a generic services but the thing is the it's impossible to not light at least of some kernel features even if your a high-level appllication developers so there wasn't recent example we've PHP 7 of container of we're Everton broke because of the different kernel on the whole system because PHP 7 relies on a certain implementation of random numbers into when you work with random numbers this leads us to kernel implementation of random numbers in there was a bug on container
of PHP 7 application which was able to run on recent Fedora for example just because the kernel different so even if you're in Application Developer you're not safe and you might get into issues we read different kernels and so every time you work with containers you should test containers in the same host system which you will use in production they like developed in any system you want do what you want but before posted to to your real life production system you should always test it on the same kind of host system even if you hold this is a cross-platform application now to containers are all modern myosins of basically container to you
was has been around for 4 10 years what but here comes Dr. and to what Dr. adds to continue relation I think 1st of all look at just appeared in the right time when container article to becomes mature enough to be properly used but Dr. also a lot of the stuff around like around the container
technology itself so don't care as in a consistent and managed continues to work with them and to share and so on so for and again for our high-level overview conductor can be
considered as a way of managing a lawyer a layered images for containers so container topologies and 1 thing but that here as this sort of thing you need to predatory for containers and again this 1 thing which is useful for application developers is to understand the layering structural for images because it's often people consider doctors just isolation mechanism and I sharing and mechanism but the forget about internal learn structure in data with various layers it becomes really ugly sometimes when you have huge images which contain basically nothing I liked in this example you can think that this is a container which is used mostly for building Java apps and you see that we have a doctor file a which is a recipe for
container image and we start from the base layer we add layer 1 on which is saying layer with Grendel binary ends a we and layer 2 we which is the later we've brought about compilers and you can see that this don't care file is it looks looks to be to your because I do
object of data of get clean several times but the reason for that is
again at every instruction creates a 2nd
layer and this where you will keep we you world whenever you move 1 container from 1 system to another so you always want to have your layers as minimal as possible but if you use object for example you always need to clean cash in the same instruction where you updated so you don't carry with you because you don't need it ever and With this are latent images as I said Dr. Panda creates the need for containers there is of course more to it which look here at
the doctor and networking the periods volumes and you can mount directories from across system in a container you can create shared volumes and so on so there is more to the topic but would generally for the purpose of this kind of talk you can safely things that doctor is late layering and observe reversion else can be added later now once you have this container
is then you have the way to store and share and you use them now only continuous becoming weight you package software and to delivery
and I want to put this bit in the context call the continuous integration with containers can like so mainly containers by the are used in continuous integration in very and 2 very different distinct ways so why way you use container in CI is that this is your built environment the built tool and generally you have this the trip is a tree with his 1st call you have an obligation artifact which we want to produce and you have your
built infrastructure environment like dependencies toolchain all related stuff to the new process and you can buy the old-school way of managing by stolen this on the worker slaves is really hard to maintain because every application of that nowadays requires its own environment and if you don't want to agree on a common baseline for use in some dependencies and so on so contender imaging CI is very helpful to solve this problem you put your all your dependencies build cash need catch dependency cacheable and continue image and then you can safely cook because it to build this particular application of greedy application artifact to search to storage and then you can just described the container and there never use it again but this is just 1 application of a continuous containers which is very very helpful in the infrastructure level but this is not enough so as soon as we go for row with continuous we want use continuous and production ends the here is is critical at different container kind of continuum cause but here we use your our application code as an octave and as a source and we build container images and artifacts as a result of a work continuous integration in delivery process in this uh container image is our production artifacts so the pipeline which means this kind of continuous images is completely different is because different away all of i approval of changes like if you build tools I have more libraries and the digital build tools fail by some reason you don't care if it's production container and there's much more curious you need much better testing unit measurement storage and so on so the last cycle
of life cycle of a certain application can look like this that you have in your source code you put you put it in the trick is a for hopefully and then you take a quote from a get trapped for you to do some building you produce an artifact in my case this is a job you publish job to your tree again this this modern which is a tree for example then you take your job you built your of Dr. image containing this journal publishers doctor in which the doctor registry and this imagine that the registry is your final artifact which goes to production environment and here are the 2 containers so we're talking about so for a rental about you use this view container for CI which is your disposable continually falling below tools and ends 4 of the final artifact there's a container for production which is totally different from the 1 you use for
building so currently there's a lot of issues when people try to a merger both of those containers into 1 and this is why don't care for example create indented to the format which is a kind of remember how it's properly quarreled but currently
you can have it basically 2 stages in
the doctor file the bill stage in the production of the you look at the production stage but I prefer to have this is a completely 2 different files and different manager differently build differently maintained if if so now we
have come to the registration we've lower and upper images in but we will lose the want from our application in production and here is where the fun starts because you know container images are not enough you need to
understand how when you all those processes with these continuing around how we interact how to update them how to roll out cultural background to reschedule 2 different costs and so on and this is where the converted is comes into
our discussion because government this is an artist Thracian . this is taken from the collaborators main site so it's kind mission statement for commodities project so the whole idea of carbon this project is to operate levels of our orchestrate most continue continuous and container images we have produced in the previous step and very this is a whole new things because if it is a platform which provides certain framework and we in which you want to dig in and out containing Copernicus has a lot of helpful objects in
abstractions which are helpful for you as saying I mean a developer or anyone but this means also that it has a lot of new terms and a whole new language you need to learn before you start working and this is the sum of the sum of the mean permit terms in the covariances and set up which would everyone who is working with companies was learned and must know because like this is the language used to describe what you do with this registration clusters so obviously we have
this image we record of our appllication the energy stored in the registry now we have a container containing around images hamas some has some continuous can run the same image some containers from different now we have this new term which is what and this is basically just a group of continuous and we have a known to which is the whole system for reports no this means you can imagine it's a bare-metal postwar or of the it'll machine it can be initially it can be the dimension in OpenStack can be very mindful it can be on Amazon instance and end of obviously you have is governed is cluster is a set of nodes and then you have these 3 more abstractions which we're going into details later so just this is
of the layout of a query is clustering generically out how I would look like so we have cluster amount amongst every node has different what of has a lot of once every point has a lot of containers and generally there is a one-to-one mapping between contain in we probably will see why is that so this is the the layout of this object and now we a leg putting the objects in this kind of layout is 1 thing but now we need to work with this and this is
where the management obstructions come coming to our tool coming this so how do you manage spots so from from a
religious point of view could is never words with individual containers it always work with have predefined groups of continues to so we we just don't are not interested in of managing container along we always have this set of continuous which we put in a book and it's called a mean it's the most common case there's 1 continue their formed and made this means that for this just a proper of all objects around your 1 based application contained so
1st when you start to to work with the wants and contain errors so the 1st notion you need to know is the replica set so I replica set is when you define a certain application you want to go make my hysteresis on this new 1 1 and you once you never wanted 1 container we 1 application you want a set of this call in this is called replica sets a replica set his account for the number of can't to see it containers which belong to this replica set all of them all of the number of points all of those points are equal this is just scaling of 1 point into multiple multiple components so here in this example would have a nginx container which at its put into nginx point and this point is the member every point is a member of a replica set from jinx into the
replica counter here is for so main the like Main option main property of the replica set and a Command is classes at the replica set can be scaled up and scale down so we at the counter is a parameter which can be changed during the lifetime of a replica set and combinators who deal with scheduling this new point to some knowledge which if it if it finds that if you will increase the replica counter couldn't is general find a way to run 1 more of the Board of the same kind of some of the nodes so deadly is saying is that the common this everything is done by comparing it is you just basically said this replica number 2 plus 1 but I having replica sets is cold but uh it manages like the number of points but it doesn't that actually content of what so once you define the replica set you define the content of a point and then you just scale-up scaled up that's all you all you do with this object so our the there is
1 more thing which you need once you start working replica sets and again as we years going to scale scale down we obviously want we don't just kill optical doubtful for the fun of it we want to load balancing off a certain service on too many instances of the application we that's the reason why we scale so close that's why I'm here comes 1 more a concept which is the concept of a service service is a common endpoint for a replica set it can be common . 1 replicas that but you could also can have several replicas sense under the umbrella of services you aiming at over more to it you can have services and based on certain select 1st so you can have both you can choose quotes buying some label and to assign service to them so and there is a flexible tool so mainly service is of the common endpoint for the group of points and which are in replica set so in this example have to replica sets with different version of vengeance supplication site and I have a common vision common services are assigned to them now are so we have replica sets we have service so we can scale I replicas service will be adjusted accordingly so everything is done on magically by commodities once you add more stuff into your replica set service will include them as well but again this is not not enough you need a way to update your content of your books and this is where deep and object comes because deployment object it means you update strategy to your replica so why all this you can do everything manually you can set up your replica sets you can set up the
2nd replica set with Wenjin sufficient to and you can degrade your all the workload from 1 to another but combinators is going because it's all it's all really does everything for you and you don't care so deployment object that manages updates for replica sets so in this example I have the deployment object jinx of which was at 1st set up as a deployment object we've replica set with aging session 1 but then I want to roll out new engines notion you I set up a rule rule out of date for a process of think could burn and then converted a 6 so it creates the 2nd replica set with the circumvention it starts to scale down the 1st replica set by 1 by 1 so it's cuts 1 point from all the replica set and it's 1 point to the new replica set and keeps a service working on to wall replica sets in between so it just replaces 1 replica set by enough heard of steadily 1 by 1 and keeping up everything in working state but in the meantime so deployment objectives the will of higher level abstraction which provides this romping around the updates 50 cents the so now all of it falls all and abstract but no we come to more interesting tasks which are the network and so we have points would have services with the plications how we talk to each other so the basic difference of of the combinators if you compare it with bookers warm and usually doctors work is that combinators and you have 1 flat network internally in every point you have there is connected to the network and has its own IP address there so you don't have of these problems of port numbers overlapping between different applications and services because every service every board and has a key it's called idea is assigned to it by the converges scandal or and thus you have freedom as an application developer as a creator of fools sports and containers to expose to use any kind of sports you want to leave it doesn't matter if you the developer
sitting next to you want to use a port 80 he can have because you will use obviously you will use different IP addresses and to everyone has the whole range of ports available so there is external network of which you're nodes are connected to and there's a flat internal network where bullets leaf ended by default to these networks are not in any way note in the notes have me and know this is through exactly so points can access the internet usually but uh the uh from from outside of the question is how do you assess that puts you affairs they just internal matter there's some IP ranges IP addresses but no 1 knows how to reach their MIT in those that can assist from the outside
so on before we get into that we should think about services 1st so all founders load-balancing works with these services mechanisms so I can convert it this uh we were is set every service hasn't yet all IP address assigned to it so it's it does not and DNS entry as is often happens and this is not a DNS because of the genus is too slow in delivering the updates basically for combinations of dating service the endpoints in IP addresses happen very fast
endianness is not reliable enough of war this kind of micro-services operations up to deliver a faster they of like can that's why it was decided that services in a burning desire represented as a mutual IP addresses and there the rooting and created a new IP tables level for services to balance of request to me my backhand and points which are forms of signal of of current serious so this again internally every service has its own personal account from the IP range and so it everybody can access a service by gatherers or by but now we want to reach their from the outside in
the there are different ways of doing that I mainly now do you need to route traffic from outside through this some node interface to internal network so our this is called a migrant histories exponent a service through outside and you can expose the service the different in different ways but most basically others they ix service can be exposed as not port so what it means you have to choose a certain worked on and i in range and all of that you assign this board to for service and then every node in your question in a will have the support of what I'm tryin directed to this internal services I could bring so you kind of know don't have an IP address here you have just a port so every service represented as a report on on on the known and the EU can access any known with this support to get to missing service our obviously accessing services ports is more fun you don't want to remember his words by means so converters has connectors through different cloud systems for example if you have to bring this platform deployed on India's or Google Cloud then it can talk with Google API or atheist API and what once you create internally service this service will be registered by India CPI and the EVS will create a there a rule which will rule to this traffic to sort Ford again on the cluster and who to this service so you don't want to have your client services to discovery services by part you will you will discover them by name and Amazon will do or google or you were like certain bare-metal of posted system
what quote so can do that for you but generally the underlying concept is the same so each node exposes support ports is mapped to a service you know all there is more to it but
I want to go to him I want to go to them more client related stuff here and so all of these were abstracts con concerns but how do you work with them we live in your like daily life so combined this has they could be huge CGL comment this is a common like it online utility is very extremely your both because it's kind rocker around the full risk the PI of when this cluster so every object is represented in this common language initially and you can get objects you can
least object described them up the data and so on so this is the kind of a REST API hemlock engine while you can of create all objects through command line by running sum by cubes CTO commons you obviously don't want to do that all the time manually typing all over options all their souls of all the object descriptions can be stored in young files in these young files can be consumed by cubes at Yale or the quantity consumed like on that but it's also possible to have adjacent farmland but generally you have the anticipatory review a young all configurations and then you just write it like this the calls folder of your young configurations to the cluster to do anything there there is also play it converges that words interfaces the graphical them interface which provide you erase you we've overview of what you have which known to have which puts you have how this is all going but this interface again it's a wrapper around the same breast DPI but it has limited functionality so should be your main option once once you really work with it and convert it is there's boards is an option for of having a generic of now all of this is 2 example also
how do you work with who cubes Yale tool from common line so for example I keeps you'll run common by default creates the deployment objects we talked about so I set is a replica of number 2 was the number I want to have this starting number of of my points I said to the image of because they look as I said the default option is to have 1 container per blonde and this is what this here if I set up this is specified this 1 dimensional teleport created for it and I have 5 instances of the and then I am simply expose this deployment object to the outside world and here and specifying the internal part of my application in my application listens by some reason to port 5 shallow and I am exposes to the outside world i don't choose which known words I'm exposing this service on because combinations will take care of it so I don't I have the problem of overlapping of ports again so the idea is that you you create your deployment objects you services is
service accounts and everything you need as the user of this cluster you shown be interrupted by another user who already have this board BC that's why you don't choose here which in an old port you will use competitors will find find it and will are registered this service trade
then you will be able to discover the sport walk your Amazon EPI will EPI will create useful so you will access service by name and this board will be somewhere behind you don't care which 1 exact reported it's so every time of the is the is the this a feature that you don't overlap with other people's work so you have the full independence a application fully independent applications don't clashing with each other and don't taking each other's for sentence and they and 1 more thing is that this networking is
is not easy and uh over the network and you see from the outside is different from the network and set up you see from inside and this is the the very huge differences because from outside you get a sport of mapping and the award of translations in between so sometimes you just want to know what's going on in the internal network and to what's happening can puts talk to each other without going out of sight and that's where debugging point is helpful but you can create a simple point i'm just with 1 container temporary so it's not a part of a huge deployment object the it's not a part of a replica set is just on 1 point with 1 image which was run temporary and soon as you close of the process it it gets killed and removed from a cluster but this kind of dividing input provides you the way to interactively uh getting through the the internal network so you run it from your uh desktop laptop workstation you get inside and you will you trigger is
uh some dividing point in my example too busy books but these books so that we had a thing to use as a dividing point because it has no nice tools to in which you need to divide so basically you need to create the debugging image with tools like TCP dump amount of work or we get and so on so you have your I mean toolbox in this in a container image then you run this continuing that should get your interactive common light from inside this image and then you all it can work on internal network from this in in this thing it was going so of course there is a
much more to the topic of various the like or I covered I covered only the the most common of deployment object because this the 1st object you start working with when there is a as well then themselves which allow you for example to deploy plots of at least 1 point per node so if you want a service which should be local to each other to each node you can set up a special scheduling algorithm Soviet portal with gentle in such a way so every node
in the class will have at least 1 instance of the point of this that you can have stateful sets supine have more volumes that can have jobs with you can have our country maps and cigarettes which means complete maps is they just things you can store in in 2 in the internal and is cluster storage of and they can be editor containers so you can have some configuration options stored as a file keepers phylum in the Cobra this cluster itself and will let containers use this file all on the fly and you can update is file and continuous will get its updated and so you can have the same but for circuits with additional safety measures you can have service accounts which error like again in Europe for the can use certain out and education methods of which are stored on the ground is level and so on so combinators and ends more and more abstraction of abstract objects to help you with solving this difficult task so you don't do this stuff on your own so that you we even can start with basics and D from them so I and
basically this was my own content so if you have any questions you can ask him yeah anything interesting in that part of it and have I and that the question was what I'm thinking about running databases in converted so I and don't think about it ahead of their main reason and like yeah obviously databases of the completely different use case from a common stayed with microsoft office and if you really needs to run them and encourages may be true right but I think this is the way out of the scope of what could this provides apparently so for me the could this as you probably so from the top of how I frame it's uh it's a hybrid is a platform for deployment pipeline to where you deployment pipeline comes to and it's very useful when you have those of hundreds of applications the development teams working on them in the living independently and everyone can manage their stuff and you can be tested properly and so on so this is a very
nice platform to solve all your integration deployment tasks but uh it doesn't add anything in case of a database management from my point of view at least now also I know I'm not thinking about anything else what you think about all that used for the case where the to what do you think about using to run uses for different application area classical 1 like having a classical contain also c continents or similar and and of all Koblitz stateful approach not a status Asuras this novel and in order to marry this both worlds of taken to be honest about thank you very a lot of stuff happening in Cuba this world so we cannot even imagine where will end up in in 2 years were might be very many different applications of the of the the this approach but I from from my side as a continuous integration engineering of the low-level maybe sysadmin I see a lot of benefit of using could burn it this kind of API out with different beckons so extending uh having the discovered this concept of replica sets deployment objects and service accounts management tools but replacing the back to a different kind of an implementation the doctor be the hands the process of horse you might not yet so this is also could be a nice interesting because look we're really eager for that way of managing Bauer deployments and way of giving developers a way to manage these deployments but we're flexible in terms of how exactly this is implemented on of basic basic level at the I have from my side of the head you have elements of life the name of a is of the of the that it's for according to the but there was 1 no part in this question can you handle state full images but yes the Coburn as a has this concept of stateful sets which is a the
way of handling stateful images but this from what i remember it's an early stage of development and mean it was it appeared in in
1 . 6 Release or something what this year so I don't know the current like production readiness for this kind of set up and I've never heard about this kind of production readiness bit maybe it's just me so I'm not exactly the experts in this particular topic because for the purpose of our set up will have very nice micro-services friendly internal project which is like really truly truly stateless and can connect to the database which is remote and we've separately so I haven't dig into this topic I heard that there's social sense but I never used yeah into that uh do you have experience with the
amount of money that's the clinics clustering the company so we currently have been set up on the W S where we create for each set of problems pool projects of 1 could but cluster for 4 tests bombers stating in 1 for production so we have flow for example than 20 to so qanti discuss some of the way of thinking is married to create just 1 what tool really become when it is class so was so really you to mode of notes sold yeah but from my point of view and I think that's is leaking out this work close to different clusters is a better way to go I mean even tho for a kind of just the development of 4 different teams I would have different clusters to for them to play with because they will have a flexibility and they don't fool break each of our
stuff this way so so I believe in in many clusters because I believe that in this young approach to grant this cluster configuration is very helpful in this way so you can degrade your configurations from 1 cluster to number isn't so unless you really need the interaction between these projects why do you put them in 1 in 1 place so once you have these infrastructure the code approach you this all you create your clusters as you need them and you know that's what I think about it so let me clarify maybe because she from the questions I heard and so apparently and we are working on a word like new greenfield projects were beginning so we're not maybe that experience as you for example so we this is our of how we look into it and how we plan to do this maybe like in 1 year you ask me on this will completely different than the this is a different way but for for now and again this is the way you know it I the and what's the
what's the overhead of deploying it could when it cluster how much resources doesn't need to just do nothing and sit there of this is a good question which I cannot answer you right now and
this is something we need to investigate more on our is set up so like just to meant to cover a understanding and no I don't I don't have the numbers from Europe so what each of the question to the auditorium yeah images of all of this we because we have because
we have experienced that all always creating a new cluster for some environment just
eat up a lot of resources it is that of course it will be weeks more expensive or but the thing is a like what's more expensive me cluster resources all the sources of the developers who will work with that of and this is always the trade off here you can go a very effective hardware or vise but then the developer will struggle and
they're like your effectiveness that brings you with so what we have done what how experiences what you need so when we have from 7 native years for example and you need to the high available so we have this master notes which are used to control the km cluster and then you want to spread some nodes or was it easy so those that are relevant to them so we have some measurements and on all set up 3 years uh Masson notes for each class and then the Balkan notes and even the visible nodes as the good at all was the availability zones so we have also set up was the 3 years and mass molten we worker nodes that's all minimal and what you can weaken the work was especially when you have such a cult set up is so how you size and also for example when you've just development of so you can take small of some these 2 instances 7 of before production and then of course you can scale up on dog but if the especially for those of I available to set up you needs of minimum spread it's the notes all was the available zones and the
that and on my questions also about sustainability because I had some experiences with student at this
and from isotope also use those 3 nodes spot at had a problem so I don't like the scalability of Cuban is because as soon as we want from 1 node of this you prick stone so the containers I'm not give you wrote about but our across the different so as to adjust the running so as soon as running TCP the instance straight down from my but home a CBIR cluster of the EC ESA service from long that Dong and 0 I'm not skewed vectors of the notes on but it's not true that I mean that if this happens in your set up it's obviously the
but because this is the idea of convergence cluster if your 1 node gives now you have URIs candling of
pulses to the other available models so probably ways of this as soon as I thought it was of course go down to see but as soon as it was going on thus read it different scale add it attracted to these really running notes to have side ability that's an interesting topic to discuss select the idea sure it's not it wasn't it wasn't just resources were not available both the main and also you poisonous most 182 ETC deep the instance freight conference I'll given Veryzer like to city services can be also cluster ice in different ways so it might depend on your that happened when the contribute because of obviously it's not designed to work this way and obviously the idea is that once you have is the file available classifier you can take out of the it is and I had the deployment of current is done by cues brains at which it was cobbled projecting recently but now it's spray and I tested is killing of 1 of the most announced in the walls of recovery and the words and so we can't think about hold on hold was there another question Alyssum how can set up a kinetochore to so on my local up serve as a matter of looking as a dual of thought when the wrong love how can a 17 year material can tests deployments that's an awesome question because
I almost forgot to tell you about it as so 1 thing don't
care and you I think the development environment generally is that I now people start to care about the onboarding of new people rather than just inventing in developing the technology itself so converges project learned its lesson from Dr. and created a lot of documentation and what has to happen around to help people to get to work because like
it appears that it's very important so now there is this meaning called tool that you can download it from the debate and it from the top and this is the mean of hope looks like a the OK they want to have some link by the thing so there is this meaning to
which you can download this meaning
cultural allows you to set up your own developer sandbox we've converted so it's there is a set of of of but what it does under the hood it goes to google fetches the the an image downloads it to your little instance set up your right around this you have to imagine this little image internally contains the this cluster is is the one node cluster so it's not the development of rail production ready for this set that but that's a full cluster we've services which you can use as the developer to test staff to run this replica sets to run ports and to deal with to play with the quadrant is classes so I haven't mining
and I think this is even worse sometimes and I
way you can see that it's Coburn in this new thing called our utility of
automatically configures autumn dictation L locally so that's your CTO starts to work with as many good cluster so it as you can see that I can access of this cluster with Q particularly tuning I can also tried to look into until finish for also this
particular I mean equally instance so you can see that I have only 1 note this is a VC-dim which Monaco download I have a lot of some deployment objects actually this is a wonderful and objects right in here and we 5 instances and that they probably can't see the point here B 5 points find this could bring cluster and so this mean equal was the book full covariances here on your laptop you you work with the API and the things you do with this API you store them in the other files you go to your production cluster upload the same number files and get the same result obviously in the multiple on them which an old level because users only 1 node we have but we you can test the deployment strategies you can see how updates is vul rolled out 1 by 1 it's also look as a playground for developers this is not yeah anything else what this that what are rhesus so our interest is a way to route to tragic not by the ports along but by rules so you can assign a set of rules and say if something they basically it should appear on you can the creative proxy on a certain level fly over look neutral force level if you specify different it'll hostages beside different unit of different
parts we will be related to different services so as it increases the next level of and node port as well when you drop around the known and you do watching on a more reached uh we reach a rules that yeah it's not a replacement for Dionysus necessary it's the way the proxy of occurrences the low some level of property yeah it OK individuals yes so I hope you got interested in and will try this many groups staff are at fault and so enjoy how it is it is to
we with that and because it it's really work very nice to to work with of course this is not the half that is problem left over their complications and this is just the tip of an iceberg and as I said is the content of the talks you just helps to get you on boards then like there's a lot to learn and for me as well it's just that we have just started with his job and so
yeah as I said the question discussions and we obviously hiring so if you interested come talk
and this and