Linux Disaster Recovery as a Service (with rear)
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
Title |
| |
Alternative Title |
| |
Title of Series | ||
Number of Parts | 64 | |
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 | 10.5446/45936 (DOI) | |
Publisher | ||
Release Date | ||
Language |
Content Metadata
Subject Area | ||
Genre | ||
Abstract |
|
FOSDEM 201113 / 64
3
7
10
11
17
19
21
28
33
34
35
37
40
44
48
49
52
55
57
59
62
63
64
00:00
Data recoveryService (economics)Open sourceSoftwareQuicksortPlanningData recoveryProjective planeService (economics)Multiplication signOpen sourceComputer programmingSoftware maintenancePerspective (visual)SoftwareComputer animationXMLLecture/Conference
01:19
Water vaporNeuroinformatikControl flowState of matterLAN-PartyHand fanLecture/Conference
02:03
Data recoveryFunction (mathematics)Process (computing)Continuous functionSystem programmingOperations researchMiniDiscBootingBackupDrop (liquid)Address spaceService (economics)Enterprise architectureMaxima and minimaSoftware testingAnalytic continuationMereologyProcess (computing)Goodness of fitPlanningBackupNeuroinformatikSoftwareComplete metric spaceNormal (geometry)Spectrum (functional analysis)Different (Kate Ryan album)Data recoverySound effectControl flowCASE <Informatik>Decision theoryMultiplication signWordMultilaterationFunctional (mathematics)Computer hardwareData managementLine (geometry)State of matterExecution unitServer (computing)Data centerBitLecture/Conference
05:47
Drop (liquid)Axiom of choiceOpen sourceSoftwareBackupComputer-generated imageryOpen sourceSoftware testingBasis <Mathematik>Configuration spaceComputer programmingElectronic mailing listComputer hardwareOpen setIntegrated development environmentData managementOperator (mathematics)Server (computing)Functional (mathematics)Goodness of fitSoftwareAxiom of choicePoint (geometry)Product (business)Module (mathematics)Category of beingCloningSource codeDifferent (Kate Ryan album)BitData recoveryFocus (optics)Multiplication signMathematical analysisLaptopBackupNeuroinformatikDivisorDatabaseUser interfaceLattice (order)ChecklistWhiteboardLine (geometry)Office suitePlanningSystem programmingAsynchronous Transfer ModeHecke operatorTerm (mathematics)Operating systemSet (mathematics)Lecture/Conference
11:26
Open sourceBackupInstallable File SystemData storage deviceBootingHypermediaTape driveLocal area networkFlow separationData recoveryEmulatorSynchronizationSystem programmingInformationConfiguration spaceMiniDiscRAIDIntegrated development environmentCloningProcess (computing)Data storage deviceTape driveBootingFile systemIntegrated development environmentOpen sourceData recoverySound effectProduct (business)InformationCASE <Informatik>Execution unitSystem programmingMiniDiscKernel (computing)Phase transitionComputer networkVirtualizationComputer hardwareRAIDServer (computing)BackupType theoryHecke operatorChainCD-ROMSoftwareNeuroinformatikUser interfaceCentralizer and normalizerGoodness of fitRule of inferenceFile archiverHypermediaCellular automatonInformation securityComputer programmingConfiguration spaceMultiplication signSynchronizationData centerDevice driverMoment (mathematics)Category of beingMathematicsPartition (number theory)Endliche ModelltheorieDrag (physics)Direction (geometry)Identity managementAsynchronous Transfer ModeStructural loadFirewall (computing)Lecture/Conference
17:06
Scale (map)Enterprise architectureData recoverySoftwareBackupDisintegrationSynchronizationFocus (optics)ComplementarityInformation retrievalComputer networkLocal area networkData storage deviceComputer-generated imageryEntire functionFile Transfer ProtocolWebsiteData modelOpen sourceInternetworkingSoftware developerFeedbackPoint (geometry)DampingSoftwareGoodness of fitScaling (geometry)Integrated development environmentBackupFocus (optics)INTEGRALUser interfaceData storage deviceEndliche ModelltheorieData recoverySystem programmingInformation retrievalSource codeHand fanBitMultiplication signProjective planeReal numberSound effectGroup actionComputer network1 (number)Operator (mathematics)Service (economics)Default (computer science)Directed graphAuthorizationOpen sourceQuicksortMessage passingProcess (computing)Software bugTape driveComputer hardwareSineNeuroinformatikScheduling (computing)Forcing (mathematics)MultilaterationSynchronizationModule (mathematics)BuildingOpen setLecture/Conference
24:14
Open sourceData modelSource codeWebsiteCore dumpInformationSystem programmingScripting languageGastropod shellSoftwareComputer-generated imageryConfiguration spaceComputer fileComputer configurationRootBackupIntegrated development environmentData structurePhase transitionCodeEndliche ModelltheorieComputer architectureComputer fileRevision controlStatistical dispersionWebsiteDirectory serviceScripting languageBackupData recoveryData structureConfiguration spaceComputer configurationSystem programmingPatch (Unix)Computer programmingModule (mathematics)ReliefProcess (computing)Line (geometry)Sheaf (mathematics)MultilaterationNetwork topologyBitLocal ringSound effectElectronic mailing listComplete metric spaceNeuroinformatikFunction (mathematics)Library (computing)Presentation of a groupUniform resource locatorElectronic program guidePasswordType theoryOnline helpSlide ruleInternetworkingBootingPay televisionFlagSoftware frameworkDisk read-and-write headDrop (liquid)Shared memoryDifferent (Kate Ryan album)Cellular automatonSimulationSoftware developerGoodness of fitMeta elementHand fanCore dumpComputer networkServer (computing)Service (economics)Tape driveLecture/Conference
31:23
BackupConfiguration spaceComputer fileData recoveryProcess (computing)SoftwareInstallation artCodeBeta functionTask (computing)Electric currentHuman migrationOffice suiteEuclidean vectorInformationServer (computing)Musical ensembleScalabilitySystem programmingBootingComputer-generated imageryElectronic visual displayFunction (mathematics)Revision controlComputer networkFirewall (computing)Directed setStandard deviationCommunications protocolImplementationVariable (mathematics)Scripting languageEmailServer (computing)System programmingData recoveryComputer architectureBackupPoint (geometry)Sound effectPhase transitionSystem callInterface (computing)Group actionElectronic mailing listPlanningBoundary value problemVideoconferencingIndependence (probability theory)Real numberClient (computing)MathematicsBitData loggerIntegrated development environmentScripting languageMereologyOrder (biology)Computer fileConfiguration spaceRight anglePhysical lawFunction (mathematics)Source codeBeta functionError messageSoftware developerData storage deviceFormal verificationService (economics)SoftwareComputer networkTape driveSet (mathematics)Process (computing)Directed graphInformationCASE <Informatik>FreewareBootingCommunications protocolStress (mechanics)Firewall (computing)Line (geometry)Direction (geometry)Lipschitz-StetigkeitWebsiteComputer hardwareRevision controlPowerPCVirtual machineStatistical dispersionStapeldateiEmailCentralizer and normalizerFeedbackPatch (Unix)Web 2.0SynchronizationLecture/Conference
38:32
Server (computing)ArchitectureGraphical user interfaceAliasingWeb browserSystem programmingEmailComponent-based software engineeringDemonFreewareMoment (mathematics)Point (geometry)Server (computing)EmailInformation securityReal numberLink (knot theory)Software developerDatabaseComputer architectureElectronic mailing list1 (number)Connectivity (graph theory)Gastropod shellDemonStandard deviationSequelWeb 2.0Message passingData conversionWave packetSuite (music)Lecture/Conference
40:35
Graphical user interfaceAuthenticationView (database)SoftwareConfiguration spaceDirectory serviceServer (computing)Presentation of a groupFunction (mathematics)Sign (mathematics)MereologySource codeWebsiteWechselseitige InformationServer (computing)Scripting languageSoftwareFunctional (mathematics)Web 2.0Kerberos <Kryptologie>CASE <Informatik>Point (geometry)Keyboard shortcutWebsiteMIDISoftware maintenanceDirectory serviceData structureMiniDiscElectronic mailing listEmailSuite (music)Open setCellular automatonShared memoryElectronic program guideReal numberIntegrated development environmentData storage deviceSimulationSound effectService (economics)Moment (mathematics)System programmingStandard deviationComputer-assisted translationData recoveryComputer hardwarePartition (number theory)ECosComputerDifferent (Kate Ryan album)Finite-state machineType theoryComputer networkInteractive televisionCondition numberCloningVirtual machineSCSIVirtualizationInterface (computing)Gastropod shellLecture/Conference
45:40
WebsiteInterior (topology)BootingSystem programmingMereologySoftwareVirtual machinePhysical systemConnectivity (graph theory)VirtualizationData recoveryConfiguration spaceSound effectBackupInformationState of matterComputer fileDifferent (Kate Ryan album)Electronic mailing listOperator (mathematics)Software testingRevision controlIntegrated development environmentMoment (mathematics)Distribution (mathematics)Strategy gameSet (mathematics)Local ringMultiplication signFeedbackPatch (Unix)Computer hardwareSineData storage deviceDevice driverLatent heatWebsiteLecture/Conference
50:46
XML
Transcript: English(auto-generated)
00:08
Welcome in this afternoon's session. My talk will be over Linux disaster recovery as a service. So, welcome. My name is Grasian Das. I'm one of the maintainers
00:22
of the Relaxing the Recover program. But this talk will begin with a broad perspective, talking about disaster recovery and disaster recovery planning, et cetera, et cetera. But let's start with it. Okay, the agenda.
00:42
So the question sometimes is, do disasters really happen? Well, we'll soon know why. We'll talk about planning, about the software that is available on the market, how to do it in practice, how to exercise it. Then I will discuss in particular the open source project Relaxing the Recover.
01:02
It's one of the open source projects that are related to disaster recovery. And as last topic, I want to spend some time describing what a disaster recovery as a service could mean, how to build it up from ground. Okay, the question is, do disasters really happen?
01:22
These are pictures from disasters that are only less than one year old. And fire can happen. This year we had a lot of snow fall, roofs collapsed. So there's a picture from Belgium somewhere. You can have a flute, water breakage.
01:40
Computer rooms can go underwater, et cetera, et cetera. Earthquakes is maybe less important than Belgium, but you never know, of course. But seriously, the most critical disaster that can happen is a burnout of your computer. Heart is that fails. You could protect by a good fan. It's a silly picture, but okay,
02:01
it's to bring some joy in the dark. So to give you some small basics, if we talk about disaster recovery, and disaster recovery affects a small piece of a complete disaster. And disaster recovery is about how to restore a function. For example, a computer to a ready state again.
02:21
So it is working as before. In contrary, you have also business continuity. That's the process. The process of the business, how will the business react when a disaster happens until that business is restored as usual, as normally. So there's a complete different broad spectrum about disaster recovery, restoration effect.
02:43
In this talk, I will concentrate on the first part, disaster recovery of computer data centers, et cetera. Albeit the second part would be a very interesting talk, but that will be for another time, I believe. So disaster recovery,
03:00
the question is always, will it happen? Yes, of course. Probably most of the people here in the room already had the disaster, more or less. A computer breaks. You lose your computer, for example. You lose important data. These things happen. Unfortunately, we are not always prepared. So the important lesson, if you go out here,
03:23
is be prepared and react timely on a disaster. So before you can react timely, you need a plan. That's very important. So don't hesitate. If you have a disaster, you have to know what to do immediately.
03:41
So I got often the question, I have a good backup server, good backup policy. Excellent. You have to have that, because it's very important to restore data. But in case of a disaster, backups are not always enough, because you can restore your data, of course, but if you lose your computer, you have to restore your computer.
04:02
New hardest, for example. You have to reload your OS. You have to fine-tune it, reinstall your backup software. This takes a lot of time. And in case of disaster, you don't have that much of a time, because you can lose so much time, days, sometimes weeks. I even had it very long time ago
04:20
that even after months, a lot of tweaks came up because they were from a disaster, because nobody know how it was configured in after months anymore. So it's very important to foresee these kind of things, and it can help you with a kind of inventory of your hardware and software.
04:40
It's very important to have it. There's software available. I will speak about it a bit later. So the plan, disaster recovery plan. I will stress the word plan. It's very important. Before a disaster, you can act on the disaster you need to plan everything. You have to find the money. You have to convince your managers of a decent plan.
05:03
You have to exercise a plan, so you need to have extra hardware available, or you hire it. It depends on the amount of value you get from the managers. And it will guarantee that you, if you exercise it, that you know what to do, that you can act immediately.
05:22
That's very important. And the last line that's very low, you can see it, is the less managers around you, in case of disaster, the better. Because they always tend to influence you, to run in your way. It's very important. Decision makers, get them out.
05:43
And with a decent plan, that's the way to do it. Very important. So what are the main steps in a disaster recovery plan? Well, the risk analysis. Before you can act on a plan, you have to know what can happen and what are the factors in my computer room,
06:00
which are important. Because not all the operating systems, not all the data, not all the computers, assets or functions of computers, are equally important. So you have to make some categories which computers should be restored immediately. Like for example, a database server is most of the times one of the important things to restore immediately
06:23
when a disaster happens, or if you exercise a disaster. It's very important. You have to develop a plan, of course. You have to time it, you have to write it down, you have to make an inventory, and therefore you have to go around in your computer room,
06:41
you have to know the labeling, the operators have to be trained, stuff like that. So there are a lot of checklists to be made and a lot of meetings to be done with a lot of people. But the important point is, you have to convince your managers, your directors, your board of directors, that you need a budget, a large budget, hopefully,
07:01
to give you the time to exercise everything. Because it takes a lot of time to make a decent inventory and to test it, stuff like that. And very important is also the last line here. You have to test the disaster recovery plan, because once you made it, that's not enough, because your computer room is continuously alive and changing.
07:21
So there are new computers going out, all the others going out, and new things is coming, so you have to redo the exercise or the thinking about the planning, at least at the yearly basis. Also the testing, very important. To do it, not all the computers, but one at a time, for example, you could categorize it and just test it.
07:44
Of course, it's always better to prepare a disaster to avoid the fact, to keep it simple and stupid. Some simple things is having good mirroring, good backup, of course. Test your restores of backups, very important.
08:00
And to make an inventory, this, in the open source world, a very good program available, it's called Configure 2 HTML. That makes a kind of hardware inventory of your system, but also a software inventory, and keeps a listing of all your configuration effects. And you can put it aside.
08:22
And of course, when you are ready to start with, with that, you have to find a good, decent program to do the bare metal restores. There are quite a bunch available. I gave you some overview of what is available now. The first choice is, do we go for commercial or open source?
08:41
Yeah, I know, here we are all convinced that open source is the way to go, me too. And luckily, there are a lot of big companies, huge companies that are realizing that open source is not bad. I remember 20 years ago, it was a disaster to convince the managers that open source was better than closed source.
09:02
But the mindset is changing in a good way, that open source does get the chance, also in the disaster recovery environments. That's a good point. But anyway, whatever you choose, if it's commercial or open source, you have to test it. Because not every program does fit your needs. Maybe you don't like it anyway.
09:21
So the testing is very important. Never test on a production server, of course. Always a test server. So what are the solutions available in the open source world? Commercial, I don't talk about. You have a lot of categories in the open source world. You have disaster recovery modules
09:42
that are available in, for example, a backup solution. To name one is Bekkala. It's an open source backup solution. It's a very good open source backup solution. But I'm not too fond of having a disaster recovery module that you get as an extra thing within backup.
10:00
Because the main focus is in backup. And you can only be good at one point, backup or disaster recovery. So you are very linked to Bekkala itself. And that's not always the best opportunity. But okay, if you like it, fine. Another possibility is the cloning. Cloning is nice to make new systems,
10:21
but it's not the right way to do a disaster recovery. Because it's a snapshot, it's always old, and it doesn't give you the flexibility that you need sometimes on new, a little bit different hardware. That's the problem with cloning system. It's not always 100% alike, and then it doesn't work. I had it all ready a few times.
10:41
But it's okay, it's good for one purpose, but not for disaster recovery, in my opinion. But okay, you don't have to believe me, of course. Test it yourself. And then you have the true open source disaster recovery software. Their main focus is on disaster recovery, luckily, and not on backup.
11:00
But okay, you can do backups with it, but it's not really a backup solution. It's a disaster recovery solution that gives you also backups. But okay. And their main focus is to give you a very quick environment user interface, not a very fancy user interface. That's not important when you're in disaster recovery mode,
11:21
you have only a laptop or a small terminal, sometimes only a prompt. So what the heck are you with a fancy user interface in that moment? You need command lines. You need to type something very fast, and it has to act immediately. Okay. In this category, we have three open source
11:41
disaster recovery software available in the market. We have Mondo Rescue. That's about from 2000. We have MKCD Rec, make CD-ROM recovery, and we have Relax and Recover. I can only say something about MKCD Rec
12:00
because I wrote it also. So it's also my piece of software which is the predecessor of Relax and Recover. It's a wonderful, wonderful program, but it's so monolithic. It's not flexible enough to react in chains, and you need chains also in disaster recovery models.
12:21
Therefore, we rewrote, in fact, MKCD Rec in a new program, Relax and Recover. So MKCD Rec is still existing, and the user base is still very huge. I try to convince people to go to Relax and Recover, but sometimes difficult because they like it so much. You have Mondo Rescue. That's also a good program,
12:40
but okay, I won't talk about it because I never use it much, so. Okay, in disaster recovery, the medium or the data is very important. Never store your data of your disaster recovery on your local computer. Always use external data storage. Very important.
13:01
Whatever the bootable medium is, it's not that important. You have to boot your new system via an ISO image, a CD or whatever, or via the network, pixie booting, via USB, via a tape with one button disaster recovery. Everything is possible, and you can mix the two things.
13:23
You have a boot medium, and you have your data storage where your backups are stored on. It can be on the same media. It can be different. You can have a LAN-only solution that you boot from a central boot server, pixie server, and that, in fact, your archivals or your data storage
13:42
is, in fact, on a NAS server, on a different system. Perfectly possible. Or in a different environment where you don't have a fancy NAS server, you can have it on a tape, an external tape drive. You can boot from the tape. You can have it on a USB disk. These days, the USB can be a total solution.
14:01
You can boot from the USB, and you have the data also on it because in USB, you can remove and store it in a vault somewhere, whatever. There are different solutions. Even tar of rsync is possible. You can do it over the network. For example, on firewall, you don't have a lot of rules. You only need the secure cell rsync rule,
14:22
in fact, to go through your internal data center. It's all possible, but you have to configure it, of course. How does disaster recovery work in practice? First of all, you have to gather the system information. So the program that you use will collect all your data,
14:43
network information, boot information, your disk information. All kinds of information has to be gathered. It has to be stored on the central place. Also, the disk layout, LVM, RAID stuff, if you're using RIP, LILO or iLILO
15:02
for a tenium-based systems. You have to make a system backup from that particular time and also the user data, but not necessarily. And you have to make a bootable image effect because in the case of disaster recovery, you have to boot from that image that you created.
15:23
That image can be stored on a CD, on a network, tape, USB, whatever. And all these steps are done online. That's a very nice thing of it. So you can have a production server. You can just launch it via CRAN or another queuing system,
15:40
daily, weekly, whatever, and you have it online available. That is the practice for making the rescue image. Now, if you are in a disaster recovery mode, you need the rescue image to boot from. So you boot your system and you have a toolbox. Okay, and here I'm describing that for the rescue image,
16:03
you need also the Linux kernels and the device drivers, the network configuration, stuff like that. And it's all done in RAM. So the unit are the environment. Now we come to the recovery phase. In the recovery phase, you need to boot from the rescue image.
16:23
You have to restore your disks layout. First of all, you have to recreate your partitions, the file systems that were originally created. You have to mount it. And then once mounted, you have to restore the data from your data store, whatever it was. When that is done, you restore the boot loaders.
16:42
And in fact, that's it. Once the boot loaders are there, you can do a manual inspection or not, depends, and you can just reboot your system. And that's finished. That's in fact the simple steps in disaster recovery, making an image and then restoring the image back on the new hardware or a new virtual partition,
17:02
doesn't matter, and just boot it. And it should work. Let's talk a little bit about Relax and Recover. Relax and Recover was created in 2006, together with my colleague, it's a German colleague, Shlomo Shapiro.
17:21
He was a very fan of MKC Direct, and we decided together to rewrite it from scratch. He also has some open source projects, and together we rewrote it. In fact, in two months time or so, we have the first release available. So it was an incredible solution. And the way it was,
17:40
because it was completely written in a modular way. But I'll come back to that later. Relax and Recover has come from a far way, from 2006, where we have no user's effect, until now that we have a very broad user environment, going from big companies, even in Belgium.
18:00
Don't know if Jeroen and Dak are available. No, he's not here, he's sick. Oh, he's sick, okay. Because we can say that the Federal Police is going the open source way, and they use Relax and Recover for the disaster recovery feature. So, thank you, Federal Police. But not only the Federal Police is using it,
18:20
the German government, a lot of German governments are using it, and also quite a lot of commercial companies, even very big ones that are using for the global disaster recovery policy. And important is, it's available since beginning last year in Fedora, so you can just do human stall, rear.
18:41
And also it is shipping with SLES 11, SP2, SP3. It is in the image over there. And it integrates, which is very important, we find that very important as a modular system. You can integrate very easily external backup software. So there are already a few external backup software
19:03
linked into rear, so that your backup is in fact stored with, for example, Tivoli or data protector of NetBackup, Bacala is also available. So you don't care about rear with your backup, it's done by really good backup software that you trust because the companies
19:22
that trust their backup software because the operators are trained for it, so it's a very good selling point that your disaster recovery system integrates very well with your existing backup software. But of course, we are in an open source environment and we, first of all, we delivered Knutar
19:42
as the main default backup solution. Our sync is another possibility and now we have since the new release 1.9, Bacala also as an open source solution. And it scales very well, that already told you that. Okay, a bit of history,
20:00
just mentioned already at this effect the spinoff of two open source projects and it was released and it was even in three weeks time, not one month or two months. Okay, what are the features? Relaxing recovery or rear, I will mention rear in the future, is focused on disaster recovery only.
20:22
Backup is not, well, backup is important but not the main focus. So we are not really interested in doing in, let's say, incremental backups. We try to do a complete full backup as the main integration.
20:40
So it's effect a simple full backup integration. It compliments also the backups of software, of course, because backup is very good in incrementals and give you a long time data storage but backup software is not really intended to do disaster recovery. So the two software pieces, they fit together very well.
21:04
And that effect is what we try to propose or what is our methodology, is say, use the best tool for the job. So I already mentioned that external backup software or commercial backup software is supported
21:23
and included today. Even last week we had the new release 1.9, we are very excited with it because there are a lot of big fixes solved and also a lot of new features were added. But I will talk about that a bit later. So the integration is very transparent
21:41
and other backup solutions are still available and still are missing. But for commercial backup solutions, we are working with a sponsoring module. So we don't write it just for the fun. We can add it if you'd like to pay for this. And very important, in my opinion,
22:01
that rear is integrating very well in the network. So you can have a complete network only solution. You don't have to fiddle with tapes or with CDs or with other external hardware. You can do it, but you don't have to do it. And that's the fundamental important point
22:21
if you want to make a service around it. You have to make a service, you need a good network. Not everyone has a good network, but okay, for the big companies who can afford it, they do have it. A good network also means a good SAN solution so that NAS or SAN is available. So you can have a very quick storage
22:41
or retrieval from your storage from your environment. But it's also important that the companies, if we deliver the software, it's open source of course, that they can make their own branded RPMs or packages from them, but it doesn't matter. So they can tweak it a little bit
23:01
and they can repackage it and just distribute it to all their computers. I think that's an advantage. And the scheduling, of course, that's done by their scheduling team or whatever. So the development of rear is an open source based model.
23:20
And we're using Sourceforce to store our data. And you can use Subversion to get the snapshots available, but the snapshots are also available on the internet. You can have a daily update of our builds. Development is done based on sponsoring or also done by the community.
23:40
We have some, I already told you, we have two main developers, but there are quite some active developers in the community. Two guys from Federal Police, for example, they are very active the last months and working very close to us, with us, to make new features available.
24:01
Which I'm very grateful. And that's the strong point of open source, it is. You have a community and community you can feed back, you get feedback from the community and you get very good ideas. And together with it, you can have a discussion and you can work to a good solution because sometimes we get patches and we are very grateful that we get patches.
24:21
There's something very weak. And then we have the community and we can improve the patches very quickly. And that is very nice. I like it very much, that framework. Why is that framework so important? Because Reel was built with, in the back of our heads, we knew about the faults we made with MKC Direct.
24:43
It was very monolithic. It was a very strong program and still very strong. Some monolithic, not only I, were able in fact to do a very good code design, code development, code patching. That's not very good module effect. The model that we now have is a framework.
25:00
And you can just plug in modules. And it's a very easy module because it's written in BAS. We don't have C coding, not yet so far. It's just a BAS script, very small. Sometimes only 10 lines long that do a job. So it's easy to debug and you can just plug it in. I'll explain it a bit later. Documentation is online.
25:21
Adjust to release 1.9 of Reel and also the release notes. The presentation will also be on our website. So you can then download it from the presentation section. I will do it this evening. So the architecture of Reel,
25:40
I explained it's a modular way. So if you do a command Reel dump, you will get an overview, like on the left side of the pictures, of course, that gives you what architecture are you running on, what is the OS vendor, which version of the vendor you're using,
26:00
and you have a configuration tree. On the left side, you see the configs which are available. Most of the configs are a bit made by us and best practice, but there are two config risk files that are very interesting for you. It's called the site.conf and the local.conf.
26:22
The site.conf effect is there you can put some variables for your site. For example, for your company, this are important. If you all work into the same pixi server, you can define it over there. For your local computer, you can just use the local.conf. And these are stored in the slash etc slash Reel directory.
26:45
So how do you use it? All the scripts are stored under the usershare rear subdirectories. There you find the complete directory structure and we have a list of methods, a method to make and rescue image,
27:02
a method to restore it. So it's the same data. Your rear executables, well scripts, are also on the rescue image and also the configuration files. So each configuration file that you modify on the local computer is also on the rescue image. So it's copied to fact. So you have the exact same definitions
27:22
and the exact same programs available. And you can make backup or make backup only is available and the recover important if you want to restore. And also you can plug in models like I said. For example, the config to HTML is already available,
27:42
not as a model, but if you install, for example, config to HTML, you can just enable it in rear with a variable. And then it will automatically detect it and run it and store the output of config to HTML in the subdirectories of far, slash far, slash lib rear effect.
28:07
Okay, already explained the site and the local configuration files. You can see some examples. An example output, output could be an ISO, could be a tape, could be one by the disaster recovery,
28:21
for example. You have also a backup definitions, NetFS. NetFS is for example, going using, we have an S server, NFS server, CIF server. Also USB is a NetFS solution. And we tend to use a lot of the NetFS solution.
28:40
And the NetFS goes together with the NetFS URL. And there are some other options if you need, for example, user password credentials, you can have some options over there. These options, you don't have to know in their heart of course, it's all described in a concept guide. You can find it on the website. And we also have a new file
29:01
in the documentation directory or subdirectory which gives you an overview of the possible configurations that you can have to start with because I know it's a bit confusing at the beginning. But in fact, it's very easy. Once you understand how it works, it's very simple. There's a simple output of the rear command.
29:22
If you don't type any option, it will give you some help and a list of commands which are available to type. Always handy to start with is the dumb command which you saw in a few slides before that gives you an overview of which is your current OS. And then the second option that you can start with
29:42
is make rescue of course, that makes a rescue image without backup. If you say just make rescue, it makes a disaster recovery image, you can boot from it. So it's handy, you need to test it. You need to know if my image boots on my system or another system. So before you do a make backup and recover.
30:02
But okay, don't worry, there are a few papers already on the internet that explain the commands and how to use it in different scenarios. The backup or rescue method, that's almost the same. There's only one difference,
30:20
then make backup makes also on backup and the rescue not, for the rest is exactly the same. It has some phases. The phases, I already discussed a bit in the disaster recovery analyzing. You have a preparation phase, you analyze your system and it's all stored under the varlibrerecovery subdirectory.
30:42
So there you find your data on your local system. That directory is also copied on the rescue image. So you can always look at your local system to see what the latest status is of your configuration files. And it does different phases. It's not that important that I explained it
31:01
but you can find it in the constant guide and more detailed explanation of it. An example, if you use the flag minus S, it's a simulation. That's very interesting. It gives you an overview of all little subscripts which are included in a certain method or phase.
31:23
But it's not readable but okay, if you install it and just try it once, it doesn't harm anything. It doesn't do anything at all. It just show you which commands or which commands methods effect and which scripts will be executed in which order. The order is also very important.
31:41
So from the left to the right. And you have a log file. The log file is stored on the slash DMP. It's called rear minus hostname dot log. Very interesting to look at. Not much output is given but it's very good if you have errors, you will see them there.
32:03
The recovery phase is, well, it's the opposite. You have the recovery system. These are the steps that are required is the verification. It will verify if your rescue image is capable of doing a recovery. If you only did the make rescue and don't have backup
32:22
and the method says, for example, Bekkala, but you didn't define any Bekkala environment, of course, nothing can be restored and it will complain about it, okay? Very important to know is if you decide a method to use,
32:43
for example, backup and an output, the recover phase will use exactly the same one because I explained that the make rescue, make backup and the recover are using the same set of conphrase files. So don't expect to do something else. It's important to know.
33:02
Of you have to foresee it. Nothing is impossible but that's part of the planning. Also here you can do a minus S. You see here it is a very small output because it was a request restore that is a very simple method that will just do its job and then wait
33:21
and tell you I'm ready for the restoration. And then you have to do another kind of thing to restore your data, for example, in our sync or something else and then it waits for you and when you're done, you just click return and it does the rest of its job, to make a bootloader active, for example.
33:40
That's a very stupid way of doing disaster recovery but it is available. In some cases, it could be interesting. Already explained the config to HTML. Cannot stress enough, it's quite important. If you activate it, you will see a line coming up, the red line over there
34:00
and you will have some extra lines or text files available under the subdirectory VAR, librear recovery, config to HTML. Very handy in case of disaster. What hardware did I add? What operating systems were operating systems? Which configuration files are available?
34:20
How are the settings, et cetera, et cetera. Again, the log file. That's an example. What is the status? The status, already explained that it's very mature on Intel and 46 based architecture. It's also working on Ethereum and PowerPC
34:42
but it's less tested. It is working but the more users will be using it, more feedback we will get, of course. This release tells an RPM tar in depth and it ships with OpenSUSE in Fedora. And well, support is available via SourceForge
35:01
and we are open for patch submissions, of course. The more, the better. The current development, already mentioned 1.9 has just been released and it is a basic steps for starting with cloning. It's still in beta. It does work but you cannot rely 100% on it because we still need to tune things.
35:21
So in the next release, it will be much better. But it is already usable for migration from physical to virtual, vice versa, whatever. And that worked quite well. The one button disaster recovery method is available. Don't think much people are using it still but okay, it's there. And what my point is, the DHCP client is available
35:44
and you need that infrastructure for having a server service afterwards. So the network, again. And there's a toolkit available for doing other stuff than just recovery but to do some inspection of your system.
36:02
What is still missing is a central point of storage. I already had, we had the central point of storage of your data via NAS server. What is still missing is a central point of your rescue images, for example, or your conference files. Because you do it locally, you put it on USB
36:21
or on the tape or CD and you copy it manually to a central point. These kind of things could be automated. And that is a fact, what we are working now as a future step is called disaster recovery as a service. Collect also the rescue images, configuration files to a central point so that we get a service around it.
36:41
And that point software we will call real server. So in the first phase effect, we are collecting the information and making a web based interface around it so that you can just make a listing of servers that you did in disaster recovery.
37:01
Did it work? What were the failures? You can look at the log files, for example. You can restart it. These kind of actions should be possible in the first phase. You can group it by department, by host, OS, stuff like that. That will be the planning for the coming months. And the plans are now after six months predicting
37:23
and getting very real. So you may expect that within the six months, the first release of real server will be released. What are the requirements? It should go, for example, through firewall. Other boundaries, network boundaries.
37:42
It should be protocol independent. So we don't want to have fancy protocols used. We only want to use the standard available protocols on the systems. So we don't want to install too much other stuff except the real server, perhaps. But even the real server on the side of the client shouldn't be any additional software.
38:03
Should be just a new real version effect. The real server is only on the central part. Maybe a few changes are required in the real software, new variables. You have to point to a real server. The sent results, how do you want the data
38:21
to be sent via mail or via HTTP, for example, and some other small things, and maybe a bit of bunch of scripts to plug in in the right spot. The real server architecture could be like this. It's not 100% defined. We are still open for discussion.
38:40
Discussions are ongoing on the mailing list or developer's mailing list of real. So if you like it, just subscribe yourself to the real mailing list of development and you can have a live discussion with us. How do you want to see it? What is the best approach? Because we have our own ideas and we already know from the past. Our own ideas are maybe not always the best ideas.
39:02
So the more interactive there is with the community, the better, and that we can choose the best way to go forward. So a central point could be a NAS server. We could use MySQL or something else. It's not 100% defined, but okay, Apache or PSP is probably very usable and will be used.
39:22
We could have Postfix as a delivery point, but we had now at the moment discussions to use WebDAB, for example, and using the HTTP or maybe even the SSL link for enhanced security. Very plausible discussions and we are really having good conversations around it.
39:42
So we are starting with it and that is very good and very challenging for everybody. So the design considerations. Okay, I mentioned a few of them, but don't take it for granted that these will be the final ones. We are still open for discussion.
40:02
I won't talk too much about it. But again, thing is we'll use standard components like for example, Postfix, like Apache, PHP, MySQL or another database. And we don't want to use a daemon. So it should be something that you push and decline push.
40:21
Of course, there have to be another way. If you want to do a restore, how do you do it? Do you do it via secure shell command execution or something else? That's still not 100% defined, but okay. Okay, the web UI interface could be,
40:40
or that the case could be done by LDAP. So Apache can handle that for us, Kerberos, stuff like that. This is only PIP scripts behind an Apache server. So that's not so difficult to create. But it's interesting to have, to have a central point that all your systems are available and listed.
41:01
Okay, the RIA server, standard software. Okay, the roadmap, and almost this is the last slide. This is the roadmap, the 1.5, the 1.0. Hopefully it will be released mid this year so that it is as a compliment for RIA,
41:21
the RIA server that we can have creating a service around disaster recovery, which is very good and very nice that have an extra functionality around disaster recovery. And then of course the next steps, but let's go to the first step first. If you want to contact us, this is the website.
41:42
We are the maintainers. Do not hesitate to subscribe yourself to the mailing list. And that was it for my talk. I'm open for questions. Please shoot. Yes? Can you show there one of the directories where all the scripts are in?
42:00
So these are all shell scripts. Posh? Yep, and the question was, can you show us a directory where all the scripts are available or stored in? On this PC I cannot show it because it's my disaster recovery PC. My PC where everything was on did not respond on the... Is it written in Bosch? It's all written in Bosch, yes.
42:21
It's all written in shell scripts, pure simple shell scripts, and you can find everything under the slash usr share rear subdirectory. And then you have a structure depending on the method you choose. But it is explained quite well in the concept guide, how the structure is aligned and how does it work.
42:41
And of course, use the minus s simulation to show which directories are in fact executed and which scripts in the directories. Yes, please. Is this Linux only at the moment? Yes, the question is, is this Linux only on this moment? Yes, it is. We are open for others
43:00
because this morning I took a CD from FreeBSD just to try it out. But okay, for the moment it's only Linux. We have Ubuntu, we have Redcat, we have Fedora, we have all the others. Is it also copied? Yes, the question was,
43:21
partition layout, is this also copied? Yes, of course. That's the basic of disaster recovery. Everything that is supported to have in a restoration of your system is copied. It's not that you bear metal, sir. You have to reconsider the conditions. Is that standard for me or you have to do it manually?
43:42
The question is, if you want to do a restoration of your partitioning on another hardware, if you have to do it manually or not, well, that is the purpose of the new release that with the cloning P2V, that the flexibility is built in the scripting that it will recognize other layouts.
44:03
And if needed, it will propose you a question. Yes, please. You said you don't like the cloning method because when you restore your clone setup, one of the problems is that how do you solve it in P2V?
44:22
Yeah, the question was, I did not like cloning software. Well, I didn't like it for disaster recovery purpose. How do we solve it with rear? Well, the thing is, rear is made for disaster recovery. So we had already in the back of our minds that not all hardwares that we're restoring
44:41
are equal. So therefore, we're using, for example, UDEF to recognize different hardware systems, other types of hardware. And it will propose you a method to change, for example, an SCSI disk with a SAS disk or another type of disk. It will propose you that, that this major step forward in the 1.9,
45:02
that the cloning functionalities is more flexible. In the previous two releases, it was much more strict. And then you had to do some manual tweaks, but we tried to avoid it because like we said, in a disasterly environment, it has to be automatically as much as possible. And these kinds of things we can, in fact,
45:21
trigger and interact with it to propose you a decent new style of underlying hardware, for the network, for the storage. Yeah. What about virtual machines and virtual services as a base system, which you have to restore first
45:42
and then to restore the various virtual disks? Anything integrating that here? The question was, what about virtual systems within a basic ESX system, for example? For the moment,
46:01
if you are talking about the main system that carries the virtual systems, it will restore the main system and the backups, if you want it. But it will not restore the virtual systems if they are running, because that's another kind of status. So if you want to virtual systems, you also backup or do disaster recovery of the virtual systems.
46:22
In the companies that I work, we do it like that also. And so you basically make a bootable ISO, which then you boot within the virtual machine as the boot part of the virtual machine? The question was that we restore the virtual machines with an ISO image. We restore the virtual systems effect
46:40
like we restore physical systems. There's no difference in the methods of booting, methods of restoring. Another question, yes? What is the actual state of support for that? Especially with L-DAT, it's what was working in Fedora, there is L-DAT, so what is missing in that unit?
47:02
Yeah, I believe the question was, what about LVM in Debian? Debian is on our list of supported hardware and software. So normally we have a bunch of test systems, because I know disaster recovery is a pain to test always,
47:22
so we cannot test everything. But normally, by my knowledge, Debian is working quite well. For the moment. We don't have tested everything, of course, but as far as we know, that it is working quite well with SLES, with Fedora, so RedCats, Ubuntu, and Debions.
47:44
They are on most tested environments. And there are others on the list. Other questions? Oh, there's four. What kind of SAN components do you support? SAN.
48:01
The question was, what kind of SAN components do we support? Any kind of SAN that Linux is supporting. If the Linux is supporting the drivers, if you use them, they will automatically be copied to the rescue image, and they will be activated automatically.
48:23
So we're using, for example, in some companies, the XP storage, or EMC storage. That's quite transparent. Another question? Does Maria need distribution-specific configuration files?
48:43
Do you need, with every new distribution version, do you need to update a set of information gathering files? Or do you simply have the strategy to copy everything? Or do you need to update some configuration files for every release, and how much work? The question is, for every new release,
49:01
do we have to update local configuration files, or your side configuration files? No, not really. It's always backwards compatible. So if you do an upgrade with RPM, the local configuration file is always saved. I mean, does Maria need to have knowledge
49:23
about the Fedora operating system? Do you have special configuration systems? And if you go to a new version of the operating system, do you need to prepare a special release-specific version of your files? Okay, the question was, with a new release,
49:41
for example, with Fedora, does RIR also need new versions of it, to understand the new version? Yes, that's true. For example, to give a very simple, this is the last question, by the way, a simple explanation on it. We have Upstart to boot your systems. We have the 6.5 init environment.
50:02
And with the new Fedora, which is still not available, the 15, we will be using systemd. systemd is not yet supported. It's on my list for the coming weeks. But for the 1.9 release, I decided to not support it yet, because I did not have the time to roll it out and to test everything again.
50:21
So it was decided to make a newer version within a couple of weeks that supports the new boot environments. And we do it, in fact, for every new stuff available. A lot of patches are coming from that. Thank you for your time. Thank you for your feedback. And hopefully, enjoy, relax, and recover. Thank you.