Rubrik SQL Live Mount in the Public Cloud

I'm Andrew Miller you can find me here I work at rubric I don't blog as much as I should came from the customer side for a while and the partner side for a while random certs and my spirit animal is not a koala a llama but it is a kangaroo unlike my other esteemed colleagues and excuse me yeah my name is Mike fall I'm a database solutions engineer at rubric I've got a blog I've got a Twitter I've been a DBA for 15 years before coming to rubric so I'm kind of used to sitting this on on your guys's seats and I have a cert because I could pass a test I'm also a bit of a powershell nut and a data platform MVP and I don't have a spirit animal just don't relate to any animals at this point so go ahead so to keep kind of framing it up we've we've won through cloud clustering rubric in the cloud private cloud out underlying cloud on I've even saw how we took of the M and turn it into an instance drone we're going to kind of tie it all together with this capability around sequel server live mount which is not a cloud only feature but can run purely in the cloud you could use this in a cloud only scenario on-premises scenario or even hybrid or multi we're going to get into that now a bag a little bit of indulgence because for the sequel live mount pieces sequel server live mount thesis to make sense a brief review of how we protect sequel we'll make that work out and I'll keep this in the three to five-minute range right because you've seen some of this in previous ones but first protection as with everything and I love some of the conversation away from Tim and others I'm drawing blanks on names right now but around simplicity so whether it is VMware which we started with were sequel or Linux or VMware or Linux or Windows focusing on operational simplicity the way that I do it is this core architectural core architecture of SLA domain policies focus on the frequency retention replication and archive we actually consider this a little bit of a form of automation because this is how we don't have to configure jobs and schedules I should have mentioned in a previous slides I ran back of exec or ran tivoli storage manager now IBM spectrum protect and whatever it gets renamed I mean all that stuff this is how avoid some of that core architecture it extends across all data types oh and if I can interject there to I think like one of the things that strengths that I see of it is you know typically backup schedules are technology driven I've got to schedule certain backups at certain times not to contend with each other the strength of the core SLA domains is now becomes a business driven architecture that you make business driven decisions about your SLA and we met fear excuse me merely mapping those out using the SLA domains or protection standpoint we use what we call a connector so we don't need a job configuration staging servers we call it a connector because we've kind of lost the term agent there's no UI on this it's sub-10 Meg after you install it it actually can can if you want have automatic updates this is via actually some real specific work around an inner core outer core design therefore the nice little image there right they can help with the automatic updates without rebooting Windows without restarting sequel as you do the initial backups the ingest we've got to get the data in initially we're using flash optimized parallel ingest just as we do for everything else but more importantly after that we do incremental forever and this is relatively unique for a sequel server backup standpoint a lot of times you're looking at full dumps of course that helps with optimizing network traffic and if we can't handle logs you know simple for recovery model we probably should be standing up here that's table stakes we ability to handle that as well as either do copy only backups how often we backup those logs and keep them for continued the operational simplicity theme the point where install the connector it auto discovers you don't go find everything it finds all the instances all the signal servers the instances all the databases and then this is all pulled into a centralized management console the same edge and console use for everything else in rubric and then whether this is Linux VMware etc and if you're thinking through this a little bit while I'm now getting a very different set of users DBAs good folks right but they're now into yes yeah good folks all right I can do that very good because everybody loves data every Alice data M yeah so if you're thinking about it now we've got a different set of folks that are getting into this console so there are some abilities around role based access control today that's administrator and user that can be grandly applied wherever you would think database instance level etc and even a specific option around are you going to overwrite databases when you restore them that's a common need so at this point we have gotten a rubric cluster we are bringing data into it we are talking to the connector after the initial backup we're doing incremental forever and just sending the change blocks across so we've got with a focus on operational simplicity and I don't being that to death a little bit but that's like one of our core focuses here because this is the part you've got to pay the price for if you want all the other stuff that matters around recovery you've got to do this but you don't want to you want to spend as little time and day-to-day management on this as you can so with that I'm going to turn over to Mike all right well so thanks Andrew I'm the protection bit and again some of this recovery is going to be a little bit familiar right if you've seen previous field day stuff you've probably seen some of this but we're just going to touch on it again just to kind of give us a foundation for where we're moving to so we do offer sequel server point I'm restores kind of like Andrew said if we didn't have the table stakes functionality of doing point time restore we shouldn't be up here talking about stuff but we do kind of give that operational simplicity with our screen to say okay well we've got a slider you pick a point in time I know as the DBA if I had to do a point in time restore I had to find a full maybe a differential I had to get all my log backups get them all in order because if I didn't have the right order it was going to break on me at 3 in the morning 3:00 a.m. in the morning when I'm going why am i doing why did I choose this DBA life because well I did so anyway we go with this operational simplicity to try and make it easy on the DBA or whoever has to do those Reece tours and we have three different recovery options the first one restore this is your typical hey I need to restore the database to a point I'm in the current place that it lives in so we're going to overwrite the existing object with whatever restore point we need then this is just your typical disaster recovery somebody broke the database somebody corrupted the database we need to put it replace the current asset with that database the second is what we call export now let's be honest here restore export it's really all restored DBAs we think of this all is the same thing export though what is a restore to some different location whether it's a new name server new database with a different name new server with a new database however we want to do it that's going to be our export option and that's pretty much what we've had so far since we went GA with sequel server integrated backup stuff now with 4.0 we have what we call live mount and this is some really cool functionality on it's basically data virtualization I like to think of this as you know giving our users the ability to use backups for more than just backups typically you take a backup you shove it off on your tape or you put it on disk storage and it just sits there in - you're like oh gosh I got to recover something right and it's basically collecting dust what we want to say is hey let's give you the functionality to use that back up quickly to virtualize that back up into some sort of usable database and so what we're doing is through our platform we're going to be the ability to mount that backup directly off of our rubric appliance whether it's the cloud cluster whether it's an on-premises device and bring that up in literally a matter of minutes if not seconds or even if you've taken on premises and replicated to cloud cluster do it from there exactly or vice versa yeah so this is at the point where we say show me the demo and so we will show you the demo now I'm going to get logged into so Rebecca I'm kicking you off of the ec2 instance just so you know that's actually this laptop over you know I'm speaking I'm kicking I'll draw off of it okay but there we go and what's our them acquire the C plus seven requirements the sequel server requirements so we support sequel Server 2008 or greater and any of the versions of that and so actually that's kind of the thing that several people asked me about live mount is like well do you support live mount now how far back do you support it we go back to sequel 2008 and I think you'll see why as we get into the functionality so I'm going to go ahead here and pull up the rubric cluster and I believe okay well I'm going to we'll go in here so what we can do is we can use the global search functionality and I can start typing in something so I'm going to go I have a 10 gig database now I think here you can see that I have two different options here one is remote so that's the on-premises cluster that we we are replicating up to the cloud that Andrew used to do the cloud instantiation from and then we have another one that we're backing up locally so we're backing this up in two different places I'm actually going to select the remote and we'll go into that and we're going to get a typical when this comes up the typical rubric calendar view or on the left side we're going to have the the overview option the right side of the calendar view any questions so far is this all making sense to people okay this is my song and dance as I wait for the spinning thing there we go alright so again what we try to do is we try to do this operational simplicity piece and we try to make everything fairly intuitive so we go to the calendar view and again everywhere we have a blue dot is a place that we can recover from where we have recover information so let's go to the twenty-fifth and on the twenty-fifth I have some snapshots and I have some timelines where I've started I've altered the database and moved it into full recovery model now I'm going to go here to this snapshot here at 1:55 a.m. and if I go to the ellipsis and click on it I now have two options if I can get over to it and I can't so I apologize I have mount and I have export so I'm going to go ahead and select mount and much like a restore export operation I get the available instances connected to this cluster with the rubric connector service in this case you know I only have the one machine will say next we're going to select the instance and we're going to give it a name in this case I'm just going to call it CFD - and we'll same out this kid saw get started if I go to the live mount section on my object browser on the left and sequel server DBS you'll see I have one an existing live mount and another one coming up so I want to actually go to management studio here and show you a couple things using the existing cloud cluster so the first is is I always like to show this this is a query against my database file locations using sis top master files so if I run this if I run if I hit the right keys here and now we'll zoom in as Rebecca said I promise I'm a technologist I know what I'm doing what you'll notice here is for my two different data files I've got now for the physical name instead of like a typical like C Drive or D Drive or what have you I now have an IP address a random string and then if we went further out we would see the file path name what this is is this is an SMB 3 share so the way that this works under the covers is we create an SMB 3 share on the brick in our flash space we then expose that to the sequel server instance we're going to mount it to and then we just attach the files just as if it were normal database files I mean you can do this with the UNC path and sequel server today one of the things I do like to call out with this particular UNC path this smb3 share this is only visible to the machine that we're mounting it to so if I have another machine in my network it will not unless it will not be able to browse to this smb3 path so this is kind of a security feature right to minimize who can get because I can go into a file explorer and browse that path from this machine that is Rubik set up in the security on the rubric appliance to only is it the username and password to connect to it I mean isn't it as far as the smb3 share I believe it's just the limit of like IP addresses so we recognize the network address of the machine that we're mounting it to and basically allow it only that way another thing that would be potentially a bit of a performance limitation because you are reading sequel files over the network rather than let me let me get to that here in a moment I want to show one other thing here and then we'll talk to kind of how that works so if I go a little further down I like to run this other query this is against this dot databases and what I was again this is more of a a DBA thing and once again I can't seem to hit the right keys or fake it was perfect well no I'm going to zoom in now Kerr Emporia but you're going to see here that this is again sis data bases so this appears is just a regular database to sequel server you know a single server doesn't see this is any sort of strange black voodoo magic and what I like to call out is you have to create a database ID that source database ID is null what this means is we're not making use of sequel servers database snapshot technology we're doing our own thing here and that's the the mounting of the files off the smb3 share we have it a recovery model of simple and is read-only is zero which means it's a readwrite database and this is kind of going to what you were asking about when we do this mount we're creating basically a readwrite copy of your data it's a virtualized copy of your database however our filesystem is immutable we don't want to be messing with your backup file we don't want to change your recovery chain or give you you know somehow disrupt that snapshot that you're going to use so what we do is we track the deltas that go on with that live mount as it exists and over time so kind of to your point there is a performance hit but we're a data protection company we're not like trying to get into the Tier one storage business we're not going to come to you and say hey you can just run this run your databases off of rubric don't worry about your regular storage know you can run your databases off of rubric and maybe a disaster recovery scenario but there is there is going to be a degraded performance to that so does that answer your question and kind of where you're going at awesome any other questions on that I've been talking for a bit here more minutes couple meant four minutes wow that's a long time so if I go here to databases and I click refresh you're going to see now that I have CFD to appear there so in all that time my live mount has come up now you're probably thinking yourself well yeah Mike but that was a 10 gig database I can restore a 10 gig database in about 4 minutes on the right scenario so what I'm going to do is I'm actually going to switch I'm sure you all appreciate the challenges of getting one terabyte of test data up into the cloud but we have a one terabyte database in our lab so what I'm actually going to do and this is also another thing I'll show you everything we do in rubric is there's a call to an API and actually some collaboration Chris Wall wrote most of the original PowerShell model I've done some collaboration with him so everything we do we have PowerShell code that we can wrap around this so I have a PowerShell script here and this is going to go to my one terabyte database and spin up a live mount of it on one of my lab instances so I'm going to go ahead and initialize this we'll run it there we go it's going to ask me to confirm my action yes and we're off to the races so now if I open up open up my browser which will go to Google News but we'll skip that for now and then we will go to our lab instance and this is where I in the Wi-Fi come on up there we go so I'm logged in I'm going to go to live mounts sequel server DBS and in the amount of time that I was that it took me to log in I mounted a readwrite copy of a one terabyte database so this is something to kind of start to wrap your brains around again we're not offering this as Tier one storage this is not necessarily going to perform as quickly but think about it I mean as a DBA and actually this is probably a good time to jump into the use cases not also where we just say isn't that cool well yeah yeah somebody this is where we can start talking about the use cases right I mean the ability to get to quickly get at those backups your backups aren't just sitting there waiting for you to run through a full restore them anymore so ad-hoc restore is granular restores or object level recovery however you want to call it somebody came along to my one terabyte database and ran a query without the where clause and deleted a bunch of rows right because that never happens and now instead of me having it like I've got a allocate additional storage I've got to go through the entire time of the restore now I can quickly live mount a copy of that database and make it accessible to whoever delete her and bring the data back this is actually as we're previewed before for auto was released I was talking with a friend who's a DBA he's got an infrastructure sense but when he heard about this he got more excited than I thought he should may be but it was because he'd been through so much pain around those scenarios event bring big databases back they're teeny little amounts of data that was critical to the business or critical or thought to be critical to whatever executive wanted it back that kind of thing so please this was this doing a remote live mount as well that I mean so so the first database that we did we mounted to the ec2 instance and it was the database itself was being backed up locally so it was a remote instance it was on-premises the one care bite database is all on-premises but again I want to you know touch on you know rubric is a software based product so it's that whether you're running it on premises or in the cloud it's the same software okay I'm asking the wrong question okay I'm sorry no apology so that I'm this miss AI strim because if you then powering up a one terabyte database but it's remotely connected to the rubric to recover from a failure that's only half the job because then you need to get that database back onto the original sequel server the originals might see what you're saying so and what we tend to call this internally is so you have with our instant recovery for VMs right we make use of the storage B motion technology to move off of that we don't have that for sequel server yet we are targeting it for a future release it's definitely something on our roadmap is this is kind of you know that this is just the case of iterative development we get a good core feature we start with it we find out what other use cases are to be you'll get out of jail free cuz exactly to the row in your database yeah but I mean going on going out of business of this database is node up I can get it out of up and then in a maintenance window you could exactly so we start talking about the ransomware I mean want to cry hit databases to write it didn't just oh I'm just going to hit VMs and if they have a database I'll be nice no I mean so if something hits you with ransomware you can bring that database back up on a live mount to get up and going quickly now again right now we have ways internally we've got some design patterns that we can get you off of that but it's not baked into the product on as we go forward again on the roadmap we're going to get at the point where you basically just push button and say okay you know bring up the live mount but also start seeding it out to a physical location that we can eventually move you off to you can use other you could use all shipping to one of them so and have movies I data might you know just to talk about a couple things here like sequel server recognizes this just as a database it's there's again no black voodoo magic it's just another database so we could log ship off of it we could actually we've talked about how can we get the live mount added into an availability group right so that when you get everything up right you just failover from one to the other that's a fairly complicated scenario all doable and actually we're going to be working on some white papers on how you go about that but again it's a database so if you can do it with a sequel server database you can do it with a live mount alright other use cases that we can kind of talk through it's health checks so one of the things DBA s are usually fastidious about is saying I need to run a dbcc check to make sure my database isn't corrupted odd but running a DB CC check against the production database is a pretty invasive workload so with this you know usually what is suggested oh I'll just restore your database off somewhere else and then run against a DVCC check against the restore again you have to allocate storage for that you have to get resources for that in the case of this I live mount my database off to another instance I don't need any additional storage and now I can run that offline dbcc check and validate the consistency of my database dev test workloads I mean to me this is where my automation my DevOps brain starts the world so you know think of if I can just get a copy of my production data I can see what a production release is going to look like before I actually release the production be my upgrades I mean a lot of painful stuff you want to script ahead of time instead of having to do major restore probably not that big of a deal because we all know that Devon production looked the same right oh that was a joke what do you think is right again and if we go back to if we go back to that power ships PowerShell script I wrote to spin out the live mount I can integrate that with Jenkins or with teamcity or whatever your other continuous development continuous integration pipeline is and suddenly you can spin out a copy of your database unit test your release send the stuff out automate the whole thing so for the the gentleman who was tweeting out about clicky clicky we're trying to get away from that and there are other units right there are other cases you're really good you didn't looked at was like in the ceiling when you said that I think so anyway the thing is this is just what we thought of right and and rubric the way we view this is we we we engage with our customers as it's a partnership right our customers you guys right you're going to think of use cases that we may not have thought of so there's going to be this back-and-forth engagement between us to say what other things can we do with live mount what other use cases what other scenarios and we support with the product and that's pretty much it so with that I'm going to go ahead and turn it back over unless does anybody have any other questions I'm sorry No alright either I did a really good job explaining it or I did a terrible job I'm going to turn it back over to Andrew to wrap this up I feel like I've just got a beat to death one terabyte you had to log in it was already up okay done beating the horse so you could have all of this with rubric 4.0 that is GA now right so that's cluster cloud up for wild cloud on instantiation this is where when we were thinking about coming to cloud field day because it you didn't want to come in and try and cloud wash existing stuff some of these have applicability on-prem so the ability for a cloud perspective even both maybe call it hybrid this is all stuff that you can use today and with that our tagline as you know is don't back up go forward

Channel: Tech Field Day

Views: 2,724

Rating: 5 out of 5

Keywords: Tech Field Day, Cloud Field Day, Cloud Field Day 2, CFD2, Rubrik, Michael Fal, SQL Server

Id: MdiGKeZZ3Sk

Length: 24min 2sec (1442 seconds)

Published: Sun Jul 30 2017