We're sorry but this page doesn't work properly without JavaScript enabled. Please enable it to continue.
Feedback

Quattor - Configuration and Fabric Management Done Right

00:00

Formal Metadata

Title
Quattor - Configuration and Fabric Management Done Right
Title of Series
Number of Parts
199
Author
License
CC Attribution 2.0 Belgium:
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.
Identifiers
Publisher
Release Date
Language

Content Metadata

Subject Area
Genre
Abstract
Quattor is a systems administration toolkit allowing controlling the whole life cycle of large and very large computer fabrics. It aims to provide great flexibility (use as much or as little of it as you want), accuracy, and consistency (catching lots of configuration errors way before deployment) and scalability, with installations from tens to tens of thousands of systems. In this talk we'll describe the main characteristics of Quattor, its simple language and show how a simple service can be deployed. Quattor is a systems administration toolkit started in the LHC Computing Grid and used in several academic and commercial environments, ranging from tens to over 30,000 systems. In this talk we'll cover the basics of the Quattor architecture, how different tools are integrated together to manage the entire life cycle of a system, and give a minimal example of a configuration with Quattor. We'll explain briefly how the Pan language helps in re-use of configurations, consistency and correctness. We'll give a quick overview of some tools available in the Quattor toolkit to install hosts and show how a part of a host can be put easily under Quattor control.
65
Thumbnail
1:05:10
77
Thumbnail
22:24
78
Thumbnail
26:32
90
115
Thumbnail
41:20
139
Thumbnail
25:17
147
150
Thumbnail
26:18
154
158
161
Thumbnail
51:47
164
Thumbnail
17:38
168
Thumbnail
24:34
176
194
Thumbnail
32:39
195
Thumbnail
34:28
QuicksortGraph (mathematics)Right angleTheoryWordMultiplication signUniverse (mathematics)Metropolitan area networkVarianceModel theoryResultantCore dumpMathematicsProcess (computing)Enterprise architectureDerivation (linguistics)Video gameWater vaporSet (mathematics)Service (economics)Entire functionComputer fileString (computer science)Server (computing)Client (computing)System administratorComputing platformDatabaseImplementationComponent-based software engineeringWebsiteConfiguration managementStress (mechanics)Computer simulationIntegrated development environmentPartition (number theory)Formal languageInformation securityDescriptive statisticsCycle (graph theory)Scaling (geometry)Data managementState of matterDrop (liquid)LogicCodePlug-in (computing)Query languageWindowMatching (graph theory)ConsistencyPreconditionerTask (computing)Parameter (computer programming)Internet service providerElectronic mailing listPhysical systemConstraint (mathematics)Key (cryptography)Declarative programmingCondition numberVirtual machinePoint (geometry)NumberXML
Disk read-and-write headUniverse (mathematics)Product (business)Event horizonVideo gameResultantFormal languageLine (geometry)WindowVideoconferencingCodeVirtual machineGroup actionOptical disc driveConfiguration managementMereologySoftwareMultiplication signDistribution (mathematics)Service (economics)PlanningComplete metric spaceData managementIntegrated development environmentDescriptive statisticsModel theoryServer (computing)DatabasePoint (geometry)Lecture/Conference
Lecture/Conference
Musical ensembleRule of inferenceResultantTraffic reportingFigurate numberCompilation albumWeb 2.0Formal languagePhysical systemDifferent (Kate Ryan album)Hybrid computerComputer fileContext awarenessCartesian coordinate systemPhase transitionLine (geometry)Data managementVariety (linguistics)WindowGame controllerVirtual machineProfil (magazine)WebsiteObject (grammar)Server (computing)Lecture/ConferenceXML
Transcript: English(auto-generated)
So, system administration which aims to change the entire life cycle Run us off for a quick break, please So, installation or configuration, every possible change that can show up during the life of the infrastructure
Changes of policies, changes of mission, everything inventory, and we are also using some of the conditions of the hosts The key points for platform correctness You have to catch any configuration mistakes as soon as possible, if possible
before deploying your configuration Because it's terrible if you see an invalid file and suddenly your entire file goes out We have vulnerability, we have to reproduce our results, we have to be able to
We play our history, we have to be able to investigate what the future will be and to go back easily if we make a mistake We want to scale, we have major package that is running over 35,000 Linux process
and they are introducing 10,000 Linux process It has to scale up, and it's much more important to scale up than to scale down If you want to scale up, you have to allow your users to choose as much or as little as they need We don't want to get in their way
We trust our security in this environment is paramount So, in order to provide that, in order to guarantee correctness and the preconditionality that the guarantee guarantees
is decent in the best of it That allows us to describe every little portion of If you want to control the accounts, you will describe your accounts in this language After that, we have a number of tools that allow you to run the managed machines
and they will align the state of the host to the desired state to the state that you have expressed in the managed machine There are three key components in the water architecture You have all your managed nodes, all your fabric
You have a configuration management database and they should talk about a very good implementation of the database Basically, the model is you produce bank codes, it gets compiled in the configuration management database and it's served via HTTP and then the installation infrastructure and the management fleet
are going to consume HTTP files and scaling up this HTTP is already a problem and it's a difficult issue with HTTP, it's also a problem Something we already know from you But, as she said, it's the guarantee
You don't want to send commands to get some code You want to send it to install something Because if one host is down, it will become consistent You want to declare the state You want to say, this is the job list of parameters and then they will consume that state
and then they will have to do something along that So we have a task that gives us that declaration It allows system administrators to declare their own data types, their own consistency constraints including cross-references between the process of the configuration
or between clients and servers There are consistency and clients and servers A very simple example would be to restrict a given string to a well-known set of values As I said, there are tools
in each planet
There will be a planet dedicated for configuring the monitoring and you will use whatever you need
and you will do that with your systems We try not to reinvent the wheel and since we are currently supporting distributionists break out our derivatives We have Insta, F, and VMC
And again, if you want to use Quota for installing your hosts you have to declare everything that is relevant the profile, the file systems, the partitions I think our Insta simulator is the only one so far that can do LVM on top of software-like
Arbitrary stress that I haven't found any other tools We don't support installation over IPv6 but no site in the community has came with that request If somebody needs it, it can be discussed Very easy to add
Reading presence is easy You have a cool new service You have a very simple API and you have to download the description of your services and match it to the behaviour on the node If your workflow is to run the file and restart the window
which is very frequent We have generic queries to do that and if you have something that is more complicated your plugin will be as difficult as the logic for your service There is no big drop there
The downside is that you are requiring system administrators, service managers People who may be very smart at their job at their mission
but they are not coders and we are asking them to wait and your infrastructure is only as needed as the code that you are providing You will need some way of having a code review You will need to teach them how to control James is going to talk about a way of addressing that
The first thing is that you can contact us or the doctor and we'll see what we can resist and follow us through the party
but the language is not as strong a time as it is so there are some odds that we can get to the video point
and we'll understand Papek is more chatty, tries to send out facts from the node to the server
and the server will tell you what to do next I think it's lighter than that regardless of what the model is like Yeah, I'm close with you but I never heard of it The description sounds very interesting
Could you give me a very simple overview of what Katara exactly does for services for example because in your example, you can see the environment you can see the data line distributions so it cannot be cross platformed and you can't find any windows services
You can use it for inventory biggest user issue for inventory We currently don't try to manage windows services We don't try to manage any approach It started in the late C complete entry and the requirement there was to support the data delivery
and that's where we focused and became very specific It's very easy to port part One planning for your configuration manager and maybe one planning for your network configuration manager How do you populate the configuration management database?
How do you interact with it? That's the result We will show you That's the gating document for your configuration language
Yes, it's super thoughtful We have a full group with exercises It's simple and we are able to close it because it's similar to what's in our machine
License, what does it say? The new code is Apache license The old code is European database license Do you still have Macs?
We're installing Macs everywhere We've got 10,000 machines and 3,000 to one variety of Macs
and a couple of Macs and some windows and they're already different applications Is that different in this case? The main purpose is to make your fabric as finished as possible sometimes it's not possible for you to have different needs and support different requirements
but tools that require DVM or Windows they usually come with their own management systems Windows you really have to be interactive and you have to do something in a way that is not difficult by Microsoft We didn't have it
It's not difficult It started here So if you want to have some demonstration
you define it with this line Here you can have the context
Have I enabled the hybrid configurations or not? and then I use these configurations So the language is very good and using recipes in your site is trivial and sharing them with the rest of the community So sharing the configuration is very, very helpful
Compile is something that we showed you It's an AP compile
Compilation phase and this puts the file the compiled object under the control of the web server and it sends notifications to all the nodes so that they will download the profile and if the notification gets lost go for it There are other ways for the node to use the profile