Skip to Content
1:10:58 Video

Simplify Management and Protection of SQL Server Hybrid Cloud Environments

This session explores how you can simplify and accelerate SQL Server DR and SQL standup in Azure VMs with automation and built-in integrations.
Click to View Transcript
00:06
My name is robert Quimby and I'm the senior Microsoft solutions architect here at pier storage in the flash drive business unit and what I get to share with you today is how to simplify the management protection of sequel server have hybrid cloud environments. What I'm gonna do is showcase uh some of the ways that we can manage and protect databases within a flash array, across flash arrays and into the cloud.
00:32
We're gonna take the second most complicated setup that I can come up with. The most complicated setup would involve sequel FCS or fail over cluster instances. And I don't want to take all the steam away from my second session, which is a sequel server always on with VM ware, Hyper B. That's where we're going to dive into there.
00:54
So in this case these are gonna be standalone sequel instances. They are going to be virtualized. I am going to leverage virtual volumes of evolves on VM ware and uh let's get started. Thank you Is me. The agenda is we're going to talk about some of the challenges that sequel dbs and you know,
01:21
hyper visor administrators encounter in how Pierre has integrated with both Microsoft and the protocol stack and the storage and the hyper Visor to solve a lot of these problems. We're gonna go into a Flasher, a extension overview, which is our plug into S S M S, which is a sequel server management studio.
01:49
We'll talk about snapshots and and the different types of snapshots that you can have, whether their app consistent to crash consistent whether they're part of a protection group and are replicated. We'll talk about how the extension works and I have Um a couple of slides that demonstrate that in case you want to forward the deck to somebody who doesn't have time to sit through a
02:14
30, 40 minute video and demo. And those, those some of those slides within, you know, five or six slides showcase, you know, how to do the automation and everything. We'll talk about using that extension, how easy it is to use it within smS. We'll talk about how after you've played with it in the gooey, how you might want to automate it and how since everything in the extension is
02:38
calling into our powershell module, which is the pure stories backup sDK Now you can go ahead and call, create your own scripts to call that module directly to automate all of your, you know, backup, restore. Maybe you have more complicated use cases like I want to refresh, test and death and then have a call to action some other integrations that we have some things that you can do, how you can test drive this today.
03:04
So some of the challenges that we have with sequel server um, they all tend to be, you know, Fall within these three pillars. Um, part of it is, you know, the rising amount of capacity being consumed and that's not just that the database is growing. I deal with customers all the time where they have new requirements and they need to provision copies of production maybe an hour
03:30
old, maybe a day old, but they'll have a certain requirement where maybe they need to scrub the customer data and then provision it to different business units within the company. And when you talk about a set capacity of production, let's say this is, you know, 50 terabytes if you suddenly need to create, you know, 10 2030 copies of that. Um now suddenly you need not only a lot of
03:57
storage to land that to talk to her as a target to place that um, but you, the, the time involved and the impact to your production storage and the production server and the production network to get it over and contestant up. So that is something that I want to showcase on how we can do that nearly
04:24
instantaneously without impacting production and how much you're willing to impact production basically the gating factor there is, are you going to replicate it off the production flash array or not? And there are many customers that that's a requirement and there are some customers that are excited to leverage clones on the production flash array because they're tested that environment is not requiring,
04:53
you know, millions of I ops into their, their test ambivalence. So the second one is, and I hinted at this is the lengthy cycle to copy and clone data was recently working with a customer where for them to get it from production into test and deV they had an intermediate step where they had to scrub all the customer data uh and test and deV happened to be availability groups and we'll dive into always on in my next session.
05:18
But one of the things with availability groups is you can't take a crash consistent snapshot and just slam it into an availability group. So they had to back it up Uh somewhere and then restore it somewhere. And then that would scrub the customer data and then they'd have to back it up and then they would restore and dozens of test and dev environments for the different business units and the initial kicking off the backup of
05:42
production to you're ready to go in the test and dev environment was between seven and 10 days and this was just a rolling cycle that was always occurring. And the folks in the test and DEV environments were more than a week behind him. Uh if you take out that, you know, 30 minute time describe the customer data, the pre and the post side of that with our integration and extension was a few minutes so
06:11
that that's a huge many orders of magnitude uh shrinking of the time it takes to do some of these things. And then another big element here is that uh different infrastructures typically have completely different um, ways to automate and ways to do things. And so one of the things that we wanted to ensure is that if you're automating with our
06:40
API is directly whether you're leveraging our python or powershell sdK s or whether what I'm going to show off today, which is our power shell integration to our backup sDK, it doesn't matter if that target flash array is a screaming fast flash array. X A almost as fast but very dense and thick uh capacity wise in the multiple petabytes flash array C or whether it is a virtual flash array,
07:08
which we call it cloud block store in the cloud, whether it's an Azure or AWS, it doesn't matter if your data is on any of that, the automation is identical, it's just your endpoints different. And so if you're running production in the cloud and you need to replicate on premises for a particular reason for devon test, you can do that or vice versa. Um We recently released a nice pure validated
07:35
design that went into it enormous detail, where we had this all replicating in a synchronous active cluster across two sites in a metro environment and we were replicating it up into cloud block Store and Azure and we were refreshing that on an hourly basis for the Azure VMS that were running sequel server and so, you know, you learn it once you've got it it works across all of the flash array portfolio.
08:07
So what is this? Flash array extension? Um most sequel dbs are familiar with SMS Sure, there's a sequel command line, uh and there's a built in power shell module there's also fuller feature set in a downloadable powershell module called sequel server. Um and all of those are available to sequel dbs but most everyone
08:35
whether they automate or not um are familiar with us sms, we wanted to make it simple, we wanted to make it where you could right click a database and back it up. You didn't need to know that that database had two files or 50 files. You didn't need to know whether this was co located on a single pure volume or 100 pure volumes regardless of the configuration,
08:58
we're just going to back it up. So that was one next was the application consistent snapshots and we'll get into this in a little bit more detail later, but basically app consistent means that people knows that the backups occurring right. You're going to do a full or copy backup full meaning we're dealing with, you know, LS engines and you might be uh combining this with transaction log only
09:25
backups to tighten, not only to tighten your recovery point objective, but to give you the flexibility to perform a recovery, no recovery, restore uh and then be able to play forward and restore additional transaction logs so that you have control over the point. Uh in time in which you are restoring um we
09:50
uh support both volume and protection group snapshots. And this is important because a protection group is a logical container that you put volumes into that you can apply policies to And one of them is if you initiate a protection group snapshot instead of a lying snapshot, then all the volumes that are in that protection group, I will have a consistent snapshot.
10:15
We take care of that regardless. I mean if you have five pair volumes or 50 peer volumes involved in the snapshot set and we're doing volume level snapshots, they'll all be consistent, so don't worry about that. But the fact that the protection group is taking the snapshot well, protection groups can have targets. Now you're not gonna gonna configure this in S
10:34
S M S um but if you're familiar with the flash array and you can configure this in the flash array gooey or you can figure this in powershell, sdk, python, sdk or even ssh into the controller. Um protecting group group can have targets, can have multiple targets. And those targets can be other flash arrays or it could be one offload target and the offload
10:55
target could be um any of our three major cloud vendors google Azure and AWS or it can be an offload target and one awful target, uh which was what I just said, The three, the cloud but the fourth one is an nFS server, so that nFS server could be our flashlight and flash blade is an incredible product because it has a linear scalability because when you need more scalability,
11:26
you add another blade. Right. And so you go from, you know, a handful of blades to dozens of blades. And and the the throughput both for ingesting into the flash blade. But more importantly to really have a really tight recovery time objective because you can push multiple multiple gigabytes a second until
11:50
a lot of folks would use up, but if you just had a clinic server hooked up to some j bod and ran nFS server on it, it could be your target. Right. And so um that is really exciting and that's one of the things that we proved out in r p d D. We had it replicating both to an on premises. Flasher a c we had it replicating to uh an Azure blob storage bucket up in Azure and then we had it replicating to a
12:20
cloud block store running and Asher as far as replication support, if the volumes involved in the snapshot set are part of a protection group by default will take protection group snapshot, you can override that globally. To force the legacy volume snapshot behavior but if you haven't take protection group snapshot and you have replication targets boom,
12:47
you get replication. The other support we put in there was for active cluster with active cluster. Most of the customers that I deal with with. Active cluster run with um synchronous apple active cluster. And so we ensured that the all the functionality would work regardless of, you know which flash array you talked to, whether the cluster was healthy or whether one
13:13
of the flash arrays was down and then a demoted or missing state you would just connect to the surviving active cluster node and carry on. Uh We've recently added a credential manager and the credential manager makes it really easy for you to add the credentials for um the flash array and the uh Windows, sequel server and the the center, if it is
13:42
virtualized on the center as far as volume types, we support um physical physical, meaning you have a server or VM that is connecting directly to the flash array uh and communicating with the flash array whether that's fiber channel or I scuzzy. Um and then for VM ware we support both of the physical RDM and evolves. So um the only thing that we can't support our virtual disks because there's no way
14:11
within a hyper visor to quest iO within a subset of virtual hard disks in the V NFS data store, there's just no way to do that. So be balls gives you that hyper visor. Easy edit the VM and added disc. Um and instead of it writing it down into a VM DK virtual disk, it is provision a volume on the hyper on the flash
14:38
array in your rolling. Right. So all done all within the center. Super simple. And I hadn't done it in about three years, I just uh set up for the demo that I'm going to do, I'm going to do a live demo right after this and in that demo um I decided I was going to make it a little bit more complicated and replicate between one flash array to another in the second flash array.
15:05
I wanted it to the to the target to BB balls and I didn't have the balls configured on that flash array and it took me all of about 90 seconds. I didn't even have to look at a guy. That's really cool if you have the V sphere plug in uh that installed and that installs in like 15 seconds by logging into the flash array, gooey clicking on the settings and plug in and then putting your VCR credentials um
15:31
Pretty phenomenal. Pretty phenomenal. So don't be afraid of devils being new or that you don't have it configured. Uh I have never read a guide on configuring be balls for a flash array and I've now set it up twice, three years apart inside of a few minutes. Um I mentioned that the backup types are full or copy only uh if you have a method of backing
15:55
up sequel today and you're doing full incremental differential transaction logs, etcetera. Don't mess that up unless you're going to replace what you're doing with this um and perform copy only backups the restore types that we support um when you're restoring from the s s uh the the entire volume or volumes in the backup set are being restored. Um the other thing that you can do is mount
16:23
any of the backups both to the original location or two. one of the many um sequel servers that you might have added credentials into the configuration. We also support no recovery mode, which as I mentioned prior enables you to uh
16:47
restore it and the database will be in a restoring state which then allows you to take subsequent transaction log only backups that you performed and restore the ones that you want. We also added a new feature recently allowing a partial recovery. Um this is kind of a corner case where an administrator has accidentally removed a snapshot on one of the ones um
17:16
but you have most of the loans um there um bss by default will fail if it can't do all and so we had to work on a workaround to ensure that that we could allow partial recovery and all easy through the gooey. You go through, you do a restore, it fails, it says Lennox is missing the snapshot, you know, you go through it again and do the restore but you check the boxing and allow
17:43
partial recovery and you're good to go and then I want to go through all the backup sdK automation, which is the coolest stuff because that's where boy you go through the gooey once and you click on, you know, show the power shell and you copy that to your clipboard and um you do that for the three or four functions that you care about and then in five minutes you've cut and pasted, edited it slightly and you're able to do hundreds and then set up your
18:07
automation. So I mentioned this prior, there's crashing out consistent. So what is that? Well, a crash consistent means that the storage you're raised taking a snapshot, the hyper visor over the operating system had no idea that this has occurred in the past. Um This was a big issue because um
18:32
a lot of storage out there doesn't have a global consistent pool of storage. You're going to have, you know, dozens to hundreds of disk devices that are hard disks or flash disks and cashing and all this different stuff and they're basically different silos because you're gonna select a grouping of them and decide which raid type to place on them. Um You may be able to combine multiple
19:00
raid groups into a volume or some a set of disks. And the super administrator really needs to understand that, you know, all the lens that I've provisioned to, my sequel server are within the same consistency group. Right? Um And that is something that is not an issue on the flash array.
19:22
Um We have a global everything on the flash array. Global d do global compression. Um You know, we don't have all these silos within the flash array, where you have to worry about, you know, is this silo filling up and there's that silo filling up. Um But 20 years ago when the SS first came out,
19:39
it was kind of the wild wild west and a lot of both the storage out there and the software virtualized storage out there. Um Unless you were really careful and even if you were careful, there were some of it where it was very often that you have corruption. So uh crash consistent uh before before peer storage
20:02
was a danger. Um And so Microsoft created the bss ah which is giving you the shadow copies and before the storage arrays are allowed to take that snapshot all the I. O. Is quest and the databases are frozen in such a way that it's a safe and application consistent, good known copy of the database.
20:29
Um and so on. The flash array. If it's a crash consistent um it's no different because of how the architecture is on the flattery and how rights are persistent to disk um at any moment in time that you take a snapshot and you look at that snapshot. If in that second in time, power had just been pulled to the flash array,
20:54
you would they would be comparable. Right? So um that's what it means. It means. It's crash consistent meaning if you yank the power and when you plug the power back in and bring it back up and then you bring your sequel back up um you'll be in a good state um everything in the crash consistency. Um You know,
21:16
it's working at the volume level uh and it can work at the multiple volume level if it's part of a protection group um zero performance impact because the flash array is just keeping track of metadata. So a few microseconds and you have a snapshot not done with this with the application consistent snapshots. Um It used to be 20 years ago that there were perfect impacts and that you had to worry
21:41
about it and if you were, you know, pushing your sand to 95% of the ops that the array could handle. Uh if you tried to then pause everything for a second or two for B. S. S to take that snapshot um Bad things happened like it would either impact the production latency because you're sitting there short stroking 15 K. Spindles and you're trying to get under sub 20
22:07
milliseconds and up but it popped up to 25 milliseconds for 30 seconds and that's enough to cause alarms to go off because everything that you did and all the knobs that you turned on that sand were to get sub 20 milliseconds or it just, you know, you had too many databases on the disk and it couldn't quest them all in time and bss aborts.
22:32
So I haven't seen that um in the last you know, probably 10 years if you were to um prove out that unicorn scenario in a lab ah without, you know, being a synthetic benchmark it, 99%. I ops um Then I would just divide and conquer. Right. Um Your backup set involves x number of Luns.
22:59
It's not able to do that. Uh you know, take out half the databases or half of the IOP support the databases out of that backup set and create two backup sets and then see if you where you're at. Um That's what we used to have to do 20 years ago. I have yet to have evidence of the pure uh customer having to do that today without
23:21
patient consent snapshots. But that would be the simple remedy. Right. Um The other thing is that the app consistent requires a hardware provider. The hardware provider is the bss hardware provider, which is part of the bss infrastructure, which we'll talk about it. It's all installed and updated and and manage
23:39
all through our sLS extension. One of the things that we really wanted was an installed once and forget about it, uh scenario. And then if you upgrade, you know, grab a new build of the extension and install it on smS um every at runtime when we do something, you know, backup mount, restore etcetera.
24:00
Uh We check on the target sequel, sir, are you ready to go and if you're an old version behind or missing it. We push it out. So that's really cool. And one of one of my favorite features, I like not having to go to, you know, hundreds of different sequel servers and manage which version is on here or is it
24:21
doesn't have it here. Do I have to update it? Um et cetera. Um The other thing is that the irish quest, which I mentioned and that's all handled through the S. S um and uh you know, it still uses the same snapshot technology. The only difference is that the operating
24:41
system and the databases are in this in a in a known good state when V. S. S tells the provider to go ahead and take that snap. The other thing that enables this point in time recovery since the sequel is involved. If you also take transaction log backups, then you can roll forward to the time that you care about.
25:05
Whereas a crash consistent um by itself. If you just Reverted or mount it or restore it. Um and that was 20 minutes ago, that's where you are and everything after 20 minutes every right to that line is gone. So this is uh you know, I don't want to spend 10 minutes on this slide. Um This is kind of in the weeds about B. S. S. I just want to call out that their four big
25:34
pieces of bss bss itself, which is a service. Um and then there are writers. Uh and each application has its own writer sequel has a writer exchange as a writer. Active directory has a writer cluster, shared volumes have a writer if you're familiar with Hyper B or for uh see SVS on a Windows cluster. Um then there's a requester and the requesters requesting the backup,
25:58
we're requesting the operation. In this case we ship that as part of our extension and it's called pierce equal snap. Um We have customers that just uh leverage the inbox Windows disk shadow requester uh to do things like uh snapshot their file server one or there sequel server in some cases, um We've only had
26:25
our integration for a year and a half. And so the majority of the customers that used snapshots for using something like this, Shadow or one or one of our partners um that are able to call our um bss provider and then the final pieces of the provider, their software providers and their hardware providers uh and are harder providers, what knows how to talk to the storage to ensure that the snapshots occur
26:54
on the storage and anytime our providers involved with a snapshot what you see on the bottom, there is a volume one dot bss dash and a and a unique identifier, which is the snapshot set I. D. Of from V. S. S. When the, when the backup occurred. So this is a single pane of glass, it's kind of hard to read the slide but in the middle that
27:23
should look familiar. That's S S. M. S. And when SmS goes to do something, it's going to call the power Shell backup sDK and that power Shell backup sDK right as the Sun is coming out and we're going to move over so that yeah, there we go. Um that's the power shell back of SDK is called
27:49
and then it is going to connect to the target sequel server through windows are M and a remote powershell session and then take care of business. And so um it's kind of a corner case but there are some cases where I've found in some customer environments where they have things locked down where maybe with group policy they completely disabled remote power Shell or remote Windows.
28:17
RM. So we in our user guide, we talked through, you know how to ensure that that that's working well. But if you can, you know, create an inter partial session, you're 99% there. Um and then the app consistent snapshot mobility is all part of the P group's.
28:36
So if your volumes are participating in a protection group and you have lots of targets on this protection groups, every time a backup is run, it now exists in many other places, which is really awesome. This slide is talking about our hardware provider. So normally with the hardware provider and this is what folks have to do if they're not using
29:01
an integration. Um I mean they're going to have to create a script to call this Shadow. No disk shadow. Not only do you need to call uh, you know, call out our provider need to ex explicitly state which volumes are going to be part of the backup set.
29:20
Um you can say that, you know, you don't want it to be a fuller copy backup, things like that and then it will, it will work but before it will work, you'll have to have installed the provider and configure the provider and that's what you see here. So the first one is you were getting the service and you can see that the pure storage bSS hardware provider is installed and running.
29:43
And the example below the pure provider Config list is showing that um We've added credentials for two flash arrays, their online and the credentials are good. Uh and it's they're compatible with the version of the provider if that is removed and this is stored in the registry. If that credential information and that Foster is removed from the provider then if you try to
30:13
do something on a lunch, that's on that flash or it will fail with the extension if that's missing. The extension at runtime will populate this. So you never have to interface this with the extension, which is a really nice feature but dr revere peer provider config is where you would change the global override for I wanted to perform volume snapshots instead of instead of if it happens to detect protection
30:45
group too. The default behaviors will take a protection group snapshot. So this is just a couple of minutes and in the slides in case folks don't have a chance to see the live demo that I'm going to do after this. I wanted to sure that you can see within a few screenshots. Um It's a simple M. S. I.
31:07
Double click the M. S. I. You know, you hit next a few times and it's installed, We require sms 18. Ah all the versions of SMS, all the energy versions um can be installed side by side. Microsoft made sure of that.
31:27
Um however, um we do not work with the old 17 and whenever Microsoft releases a new version, they deprecate support for the last energy release. And then if you're familiar with S. S. M. S, it's like on the monthly or quarterly basis, there always releasing updates necessary. But then you, so you go through, you install it done, you open the best of cement um and then
31:53
what you'll notice is that at the top of the title bar there is now pure storage and everything that is accessible there in the title bar is global to the S. S. And S. Machine that you have installed the extension cord. So everything that you go to configure in here will be global.
32:17
So if you've created backup jobs for You know 50 different sequel instances and you click on the back of history um it's going to be a long list of every backup on anything, right? So if you want to have it filter on the database that you care about, you wouldn't do it up top here, you'd right click the database and click pure storage there.
32:40
So not only can you create a backup. Config that's like basically saving a backup job Or edit one. You can kick off a backup, you can look at the back of history that's global. Or you can open the credential manager and the credential manager is where you could go in and do a few things that will show off live.
33:01
This is what it looks like when you say, hey backup can faith. And as you go to create one, you're going down the list from the top to the bottom. The configuration is the friendly name that we type in. That's the name that you need to know for your automation because the command lets you know, invoked a backup job desk and fig name and then what's the configuration?
33:23
So you wanted to be helpful so that you know what's getting backed up. And the S. S. Is very particular. If you say I want to back up in this case, you can see that there are a couple of different databases and you can't see the database name because it's the names longer than what you see there. So we look into the database,
33:50
you see a server name slash san ramon which is actually the sequel instance name. And then if we look to the left outside of that backup convict window you can see that well there's two Databases. There's an Adventure Works 2019 and the Adventure Works 2019 underscore no recovery. Uh huh. So the databases that that's selected there is
34:11
the Adventure Works 2019 database. Um and why this is important is uh you might have more than one sequel instance on the server and and and there's nothing prohibiting you in sequel from having the same database name ah Multiple Times. Right? It has to be unique within the instance but if you have 30 instances installed You can have DB130 times.
34:36
So which DB one. Right. And so this is a way to help you clarify that when you're going through and selecting your database then you're selecting the flash array. And the next thing you can do is select the remote flash array. If if uh it's football and it would be applicable in the case of active cluster or active D. R. These are features within the flash array where
34:57
you're going to have a copy of These snapshots are more than one flesh shirt. Now, either synchronously or asynchronous. Um and in the case of active cluster if you know, Flash Array one is down, we'll just connect the flashlight to and continue on in the case of neither of those but you're just using protection group replication.
35:25
The flash array you place in here will be where we will look if flash array one is down. So flash everyone is down the line that adventure works 2019 is on is down the database is down, you can right click the database going into the back of history and mount the most recent snapshot and it will go to flash everyone while it's down so that it will try flash every two and if it's there, it will mount that snapshot on flash Harry too and
35:51
connected to the server or the BM in europe and running. So that's pretty cool. The volume types here is what I mentioned before. It's going to be physical B ball rDM in the case of Ebola RDM uh we, from a storage perspective, like from a flashlight perspective,
36:11
but if you're sitting in, you know, the Master Control program inside of the flash array And I want to give sequel server one alone, like create a volume and I say this volume needs to be added to this host or host group. And that host or host group is connected via, you know, in Q and I Q N F C W W N and it connects it all up. And then suddenly the the initiator can see
36:36
that line uh so what do we do on the flash array if it's a B m? Uh we don't know what a VMS Right. The flattery just knows the endpoint. So if the endpoint happens to be an E. S. X. Host, uh, and and it's a B ball and um what are we gonna do will work if you go and you snoop around in the foster area, you'll see that when we mounted it,
37:00
we created a new volume and we copied it from the snapshot and that volume is now connected to the E S. X. I. Host that has the VM running done. That's all we can do as a flash rate. Right? So that's why we need to be centered. So you'll create a dissenter credential. And by doing that after the Foster has done its
37:22
work, then we call them to be center and said, hey baby center, this new b ball that we just connected to. You really needs to be connected up to, you know, Super B and one. And then we issued the call to the center and then boom, the beach balls connected up to be a sequel BM one and then now we continue and uh it's
37:42
mounted restored what have you. And then another issue that I had about a year ago is we had um some sequel dbs that aren't global V center administrators. So they went and set up all these jobs and then the V center administrator went in and changed the name of the VM.
38:05
Uh and well we need the name of, we needed the name of the B. M. Because if you're going to mount something we needed to tell the center hey you're rounding into sequel BM one and they renamed it to, you know, product group dash sequel VM one. And so it failed. So now we require when you're doing the config we go and we pull and we grab the being
38:24
persistent I. D. So now it doesn't matter if the be centered man or the super admin. If they're the the center changes the VM it will just work for the metadata location. This is where we're saving the metadata. Um This is kind of important information because if you want to mount or restore through our extension, you need that metadata.
38:48
If that metadata is lost, you're not going to be restoring it through this. You're going to have to manually do it. Uh either through the sDK or through the power flash or a ui the gooey. You have to go in and say okay, which snapshots uh with this I. D. Where are they?
39:12
And I need to restore them also. All of the lunge and the backup center consistent. Um and then the backup type there is just gonna be full or copy only and at the bottom you see, you know, there's a powershell script. Thank you. For example, this is the different context. This is when you right click on databases.
39:32
In this case we right click the database. Ah and we've clicked something and in this case they clicked back up and by default it picked the configuration, the most recent configuration created for that database and in that case uh you didn't click edit a backup configuration or new backup configuration, you put back up so none of the fields are editable and you can
40:00
either you can cook back up, you can look at the power shell script and in this case the power shell script is just gonna be invoked dash pf a backup config name. And then that conflict which is central and underscoring the recovery and you'll see a little uh progress bar, green bar on the bottom as it's backing up. It typically takes between 10 and 30 seconds unless it is on a CSP.
40:29
So um I don't recommend this. Uh most folks that do have ci s and I'll get into this in a lot more detail in my next session deploy it where they have created pure volumes, added those pure volumes to the cluster, added those volumes which are clustered two disks to the sequel server role, which is the incidents that you've installed in those disks move to the different nodes of the cluster as things fell over.
41:02
Mhm Microsoft allows you to install to a CSB. In fact you could install multiple people instances to a CSP and run those instances on different nodes of a cluster. Um But if you in book, use the C S V. Um I mean yes, potentially you have a couple of fewer disks to manage
41:23
um but now you're not separating the system databases from user databases. Um Right, and That's dangerous. And if you have more than one instance, right. If you're trying to restore sequel instance one and there's 12 instances on the CSB, your overriding the CSB so not recommended. The other problem is that a cluster, shared volumes is read and writable
41:51
by all the nodes in the cluster and it's up to the role that owns the files on the C S V to ensure that only the role has write access to the file. Right? That's why um like Hyper V. Right. You can have hundreds of VMS running on one C S V on a 16 node cluster and the VMS are all online on the different nodes and nothing's overwriting anything else because the VM itself
42:21
is a role that is now given permission to write to that in the case of hyper v, a v h D x file, whereas that would not be the case. Um otherwise, so the C S V writer is now a new writer that says, hey, I'm involved and then instead of it just being within the operating system environment, it now has to communicate two every node in the cluster that's an order of magnitude slower.
42:49
So you're looking at, you know, a minute or more To take that snapshot and I haven't tested it with like a 30 node cluster, so I don't know how bad that would be but I wouldn't recommend that. And then here you see, yep we took it back up. Great. Um it's just a couple more slides on the demo and then let's get into a live demo.
43:11
This is right clicking a database and saying, hey let's look at the back of history. In this case you can see you know the pertinent information, what is the source driver drives or mount points um and then in the case of a snapshot, you know, second from the top there you see that that snapshot is mounted as the H drive on a particular server. Uh and so when you go and you click on one
43:33
that's not mounted and you decide you want to mount it, you cook that mount button and that's when we'll rent a query on what dr letters are available and then you can pick one or you can erase that and just in the drive letter box type in and out points and you'll be good to go. Uh And that is backup history and again this is uh
44:01
showing multiple configurations because this is the global black of history view if you had right clicked the adventure works 2019 and then she said backup history it would only show the configuration on the left that was applicable to that news here is where you would say I want to restore uh and so in the case of restore, I mentioned the recovery mode that will leave the database in a recovering state.
44:31
So you can restore additional transaction like tearing files. They allowed partial restore. Like I said before it was in case of a strange corner case where something got deleted. Put the new power shell script and see how to automate that. In this case we want to mount and the mount computer can be any of the sequel servers
44:56
that are like this disc type. So if the source was an RDM you couldn't mount it on another BMS view. All right. It has to be like for like and then if the source was available, you couldn't mount it to a physical symbols are it would be fairly trivial to copy the snapshot and take care of that outside of this.
45:21
But our current workflow doesn't allow that if you're interested in me directly are Quimby pure storage dot com and also be a sample script on how to work around. That. Pretty trivial to do. Um Here's where and this is something that we added uh in the last couple releases um If you in the past would select one of these items in the back of history and said delete,
45:48
it would delete the metadata and it would delete it from the back of history. Now we give you the ability with this checkbox or it's a switch on the Fan line where you can destroy this snapshot on the flash rate. So at the same time that you're removing it and this can be done, you know through para shop, you can also remove the snapshot on the flash array.
46:11
Um It's also trivial for you to run a second job, right? Uh and we have customers that do this, that have a set number of snapshots that they want for each particular volume. And so the olders, any older snapshot beyond x get removed and they run this daily nightly or weekly or what have you. That is how you um
46:36
oops tap too quick. That is how you look at the script here and in this script it's just now. This is getting the list of the history and that's how you would do it. And this is clicking on the credential manager, this is where you can go and see. Oh yeah, I've got two different Windows servers here added and their credentials and I've got
46:58
one flash array and if this was virtual, you add in the center credential as well. And before I jump into the automation we recently added some functionality in here, not just adding editing it because I see the password changed or removing it but verifying. So if you know one of your backup jobs stop working and you're like well jeez did the password change or something, you cook on one of these credentials and click verify and that
47:25
will tell you if there's a problem. Uh And then for additional troubleshooting, if there's a problem with the provider, you can click, can figure targeted to clean things up now for automation, this is where everything is in power Shell, so I've got a couple of things here and then I want to get elected.
47:47
Um This is the back domesticate so you can see that you know we need power shell 5.1. Um We need uh you see the operations that are available to you um any power sharing module, you do get that command space, dash module, space, the module name and it will show you the commandments then you know what what is this commandment And how do you do this if the module was written appropriately they'll have help in with each commandment.
48:17
So you should be able to do get dash help and an example here would be you know get dash help, add dash pf a backup cred and it should dump some help about what the heck is that, you know, up arrow and then add a dash examples, give you a couple examples up arrow and do a dash full and it should dump everything. Um And you should with most power Shell modules on the market, that should be your documentation like you should be able to get
48:49
functioning scripts just with that and that was one of our goals. Um you know we got a lot of that from a lot of the Microsoft modules and I have a lot of background with the exchange and have the exchange admin tool has turned into power shop. And so being able to do all of that is something that is really important. You can see that, you know, since things are objects, right?
49:13
You then can then pipe them until you see the bottom right example where you know, yeah, we want to get the back of history, but let's just grab these particular values and dump them. Uh Right. You could, you know, also very easily um get the pf a backup history but only select the config names. That or SAn ramon, right?
49:34
And you would only have gotten That history 80 # seven at the bottom and that kind of stuff is where it starts to get really powerful in something that I use a lot myself where I'm like, you know, give me all of them that are just on this config name and then give me just the most recent one and now I know that in the list, you know, the the property zero is going to be the most recent one.
49:59
And then I can go do something with that with another script or with another commandment. And then here is that get command right? Really simple. You're seeing the module and showing the version of the module and then getting the command and it's listening to commands and then you do it get help and then run forward from there. Here's where we're issuing a few different
50:21
commands were first getting the back of history. Uh and then we're deciding, well let's kick off the backup And one of the convicts is the San Ramon _ one. And then we're getting the back of history again and you see that now we have history I. d. 10 and that's not try to have occurred.
50:39
And then if we get information on that recent snapshot that we just took that history 8010 when you see not only the serial number with the name and the name in this case is the volume name. So if you went onto your flash rate to that there's a volume called, you know f 0 4-0 9 dash cycle dash data one. Um and then each snapshot is being, that was taken by the provider is going to have a
51:04
different dot V. S. S. Dash and then that number after the dash is the V. S. S snapshot idea. So this is my last slide. So before I finish this last slide I want to thank you for your time And I'll finish this last slide and then I have about 15 minute demo and if you like live demos where or you know,
51:28
things can go wrong and things alive, fun and exciting. Hang hang tight and if if you don't then I hope you appreciate uh this fellow. So first the call to action is download the flashlight extension. Uh you know, we've got a couple different versions out there last year, right before Christmas we we delivered 2.5 which is what gives us support for our virtual
51:52
volumes. Um we have all of our best practices on the Microsoft platform guide uh and that's as easy as going to support that peer storage dot com and then clicking solutions and then clicking Microsoft and you'll see all that kind of stuff. Check out our pure validated design. Um that is something that is like 100 page white paper,
52:14
tons of stuff. We did tons of working Adger with Block Block store, take a test drive. We have virtualized flash harry and virtualized VMS and everything in an environment and if you reach out to your peer assistance engineer you get a voucher and you can check this out and run it yourself. I got a little guidebook that pops up and then boom you have a live environment,
52:37
it's not nothing's a sandbox and fake and broken. It's it's full of them. Full sequel, full S S. S, you install it, you can figure it, you check it out and you We'll see that in 10 minutes you've taken a backup and you could look at the power shell and boy you're ready to roll and it's a lot of fun and on the right side there you can see that we have lots of other
52:59
integrations, we have Powershell sdK for both our rest api one dot x and two dot x. And in fact the, just a couple of days ago last friday we released two new updates on that. So for both one and two dot exe and we also have an extension in the Windows Admin center. So if you finally gave in to Microsoft on recent operating systems,
53:25
whatever system manager pops up, try this in Windows Admin center and you cook the button and you download it all you have to do when you're in the Windows Admin center is click on extensions, you'll find pure boom, you got it and you're good to go. And there's some fun stuff you can do there. We also do a lot of work with system center. One is Operations Manager,
53:43
which we just recently released a new update to the management pack. So not only can, you know, within scum have an alert that hey, you're using too much of the storage capacity is too high over the threshold. Maybe the ops are too high, The latency is too high. Some things down replication networks down, somebody bumped a cable and one of the flash
54:07
uh, fiber channel lynxes down etcetera. Nice to have that if you use Operations manager. And then on the B mm side this is a kind of Microsoft's V center SCB mm because of our s M I s support. You can add flash arrays to Vietnam and that will help you rapidly clone. We have a lot of customers that use this for beady eye, You know, they update the golden image of all the updates or the new build of,
54:40
you know, Windows 10 and they're able to push that out and replicate that to thousands of virtual desktops, you know, within a few minutes. And then we have a Parasol toolkit. The parasol toolkit has lots of basically command lets that we've added that customers have requested things like being able to and check that the best practices are in place. Things of that nature.
55:08
So with the so the first thing we're gonna do, uh, we are on a VM and this VM is got a little screenshot here is sick will be M4 and we'll see that and we've got a production running here and this production databases, you know, A couple of gigs and it's on flash, right? 32 and a ton of pure volume.
55:40
And what we're gonna do is we're gonna take an application snapshot and we're gonna asynchronous to replicate it to a different flash. Right? And that different flash array is uh Where sequel DM three speed balls live. And we're going to refresh test and dev. And not only are we utilizing devils, we are utilizing asynchronous
56:06
replication where it is on multiple Foster race now just because I don't have infinite physical hardware. Um, this is a single ES. ECs cluster. Um, but the different VMS are connected to different um, data stores and one data store is on one flash rate and one data store is on the other and I'll go through and show off that.
56:29
So that is it in a nutshell. So then if we look here and we look at this, we see that. Yes, we are so called the M. Four and then if we go into um this happened twice, we go into S. S. M. S here, We're connected to sequel BMW four and let's look at the data bases.
56:48
Uh we got DB one DB two, we've got to develop a group database. This is the one that we're going to play with today. What I want to do is get rid of an old table that we don't need anymore, let's do some changes. Let's just create a new table. Um In this case today is 19 May 2020.
57:09
Um Let's go ahead and say that 18 2020. So we'll just, you know, it's just an empty table for now. Um But by refreshing this and smS you see that we have this new table. Um so now that we've done that, I mean I've shown you everything of this and slide wear so I'm not going to go through all this um ah and then if we right click the database,
57:33
you've seen this where you know under peer storage you can kick off your back up. Um You saw in the slide where you know what kicking off a backup, good and how it did a little green bar at the bottom and if you click on the power shell script, I mentioned this um it's that simple so I'm gonna copy that to the clipboard and invoke it here um in power shell and this is, you know, you could schedule it in
58:00
your schedule er of choice whether it's some sort of super agent or whether it's some sort of, you know, just scheduled tasks and Windows or whether you use something outside of the operating system to call into it. Um lots of opportunity here. So this is going to be done in about 10 more seconds because this is all live demo and I'm
58:19
not gonna, you know, pause and cut things out and so it's finished and it's dumped out, you know, the information that is important. Um you can see that the local flash array here is flash array 30 and we're replicating it to flash array 40. Um we can see that that was the name, it was the B ball P group database, that was the config name um and that it was of evil and
58:43
that we just took this snapshot to if not completely mounted anywhere just um so that's all I wanna do here. This, this was basically okay, production, we just took a snapshot really, really fast. We changed some things, we deleted the table, we added the table. Now what we're gonna do um and just so that, you know,
59:02
we don't have crazy things in zoom with multiple windows. I'm gonna rdp from within this shared window here. So we're going to go ahead and take a peek at what's going on here and then we'll open up um a browser uh and check it out. So here what we have is test and deV and then test and dev and this VM is
59:28
connected to the other flash array. Um we can see that it has the old image, right? It's mounted, everything's happy, it's the old image of um what's going on. And so before we do much in there, let's see what's going on in um let's see what's going on. And um let's see
01:00:01
the flattery so here Heres the flash area that is equal the M4 and if we look at it um we'll go in here into storage and we'll go into volumes. And since this is the balls, it involves volume groups. And if I go to school be um four we can see um It's two p.m.
01:00:37
2 or three on 5 19. And if we look we look at in the volume group there's no snap trucks. The last snapshot was two months ago. So it's like uh what happened um and so what we should do is figure out which one of these meatballs is it?
01:00:55
And what I did is I renamed the B ball because I hate data data dash and some weird number. I want to know what all my balls are in this case, I only renamed one B ball but I heartily recommend rename your balls. What the heck is this? So this is the B ball that had the production database on it. So if we click on it, what do we see?
01:01:15
We see that? Oh, At two o'clock today, uh you know, three minutes ago there was a snapshot but the snapshot is called a Sikh demo. And the reason it's called a sink demo is because this volume a sync dB is part of a protection group, which we see here on the right, it's part of the protection group demo. And so since I didn't have the default behavior over overridden a protection group snapshot was
01:01:43
taken instead. If we click on the protection group, we see that We have a target here. And so you know, a 29 not flash array is a target. Um you can see that, I'm also playing around with the snapshot schedules and the replication schedules, but that's not necessary in this, in this type of configuration.
01:02:06
Um and so you can add additional targets here until the clicking here. Um we would add targets of additional targets that are added to the flash array and right now I cleaned it up so I deleted the other three or four arrays that I had connected, but if you add additional flash arrays here. Ah and then you can add them as targets if you want to add an offload target,
01:02:30
you can have one, I mentioned this earlier, you know, there's atmosphere, there's S three, which is aws and there's Google Cloud, but if you select NFS that could be, you know, an NFS server flash plate. Uh so you can have more than one target for a protection group but you cannot have more than one schedule. So the schedule applies to the global
01:02:52
protection group. So we have some customers that are like, okay, I'm gonna use this protection group for the Super Bowl back up but I need a longer term storage somewhere else and I want to have a replication schedule and a retention schedule that will keep that stuff for, you know, a year because it's going to the cloud, then you have multiple protection groups and kind of play with that.
01:03:15
Um Okay, so that's what I wanted to show you on the source Flasher, right then what we should do is look at the target flash. Right. So the target flash right here, um, it is just sitting as replicated protection group snapshots is not connected to anything.
01:03:40
What's connected to things are already there. Right, so here we're going to have volumes for sequel, BMW three sequel, BM three um has an a sink target uh the ball and this is the one that we're going to overwrite when we refresh the test and a copy of it, which we already did. We just haven't looked. Um and so in this case, um, if we look at the um,
01:04:09
yeah, protection here, we can see that this which took like, you know, we'd have to double check, but within the same minute, mm hmm, I received a full copy of the snapshot that we replicated when limited the snapshot. Uh and so this snapshot is what we want to use um, to refresh testing death.
01:04:35
Um, and then the last thing is all this really quickly show you the center just so that this is all visual and I'm fed up And here we can see that Super VM three, which is the target. It's big hard drive here is living on F a 40 and C equals B M four
01:05:27
which is production. Um, it's big two, it's gonna be on 30 and you see that at 3 30. So, and the b balls are different luxuries and we are in the center. Oh my gosh, it's so easy. Um, you basically go into pure storage here and um,
01:05:53
this is where you would add a new array, right? And then you can go in and Set up the hosts all within a few clicks and you're good to go and then we've got a guide. You got to support research.com solutions somewhere and it literally took 90 seconds of guessing without even looking at anything. Okay, so that's the center, so you saw how that is and then everything that I've been talking
01:06:18
about um because my demo is almost over, I don't think so. Microsoft on the brain storage dot com. So if you just go to support dot here storage dot com, that's like the main landing page, I mean you can dive in and get like the flash array user guide that goes into everything, you know, things like safe mode and stuff like that, which I have enabled on the production flash array but not on the D R flash array in this
01:06:42
particular example. Um and then if you click on solutions, we've got the top two are going to be, you know, Microsoft and being worse and this is everything that I was mentioning, right, we've got our support matrix here, this is our PDT the pure validated design. Um this is a, this is a white paper I wrote on super server and active er which is our
01:07:03
continuous replication for the flash arrays, which is really cool. Um we have some customers, it used to be that synchronous replication was cost prohibitive, it's like, you know, most of the sand vendors were charging, You know, hundreds of thousands of dollars for the license for that. And uh so not only did you need to have really close latency but it was very expensive.
01:07:21
Now it's just do you have the network and the network fast enough and for us fast enough means 11 millisecond latency or better, but now you have this Ferrari, you have this blast array that's given you microsecond latency and your latency across cities or campuses to say eight milliseconds. So now you're eight milliseconds plus, You know, 150 microseconds or whatever.
01:07:45
That's an order of magnitude slower. Some people don't want that or they have certain workloads that can't take that additional hit. They're used to the sub one millisecond latency. Okay. For those workloads consider active deal um or availability groups within sequel. I mean there's many different things that you can do here is our bss harbor
01:08:05
provider, the powershell sdK that have talked about here, talked all about the Windows admin center but you can install the extension through act Uh and then this was the recently updated management pack 1st. Calm. So now that I've done at the web, let's look at um let's look at our test and dev environments.
01:08:27
So here in test in DEV. What we're gonna do is I'm gonna run my refresh script and what it's gonna do is take down the database by taking it offline because it's running right now and and testing data and then it's going to take off offline the disk. Then we're going to override it with the snapshot from the protection group and then we're gonna bring it back up and then I'm gonna dump off the tables now,
01:08:47
you know right before I do this that it has the old for 4 20 here And if we refresh that, I mean that's live and we created the five 19. Right? So let's run refresh and then use the password for my flash rate because I don't have it encrypted on disk yet and this will take no it's done. So you can see it did it now, the tables are different and if we refresh it in the gooey,
01:09:13
um we see that it's different. And so all of that took, you know, 15 seconds on the primary to take the snapshot and it took 15 seconds to replicate and it took 10 seconds to do this. So, um in this workflow right with your looking at sub 60 seconds to refresh testing. Now, if the interval of how often you would refresh, test and dev um is very
01:09:41
long and because it's very long you're right or change rate in production is very large. Right then the thing that you have to increase in this time is the replication. So in this case the replication between flash rates 10 seconds. But if a terabyte or 10 terabytes of data had been changed and that needed to be replicated across the wire,
01:10:06
that's going to take a little bit of time and that are some of the metrics that you need to keep in mind of, you know, well, how often do I want to replicate because I don't necessarily want the replication engine to take, you know 30 minutes because we're a competitive fight of stuff at the end of the month because of whatever. So this is the demo.
01:10:28
I thank you for your time and I encourage everybody to go out, and if you have a flash array, uh, to check this out with a test dummy database on a VM or on a physical server that's connected to that flash array, and you're gonna find that as fast as that is. Um, we're even working on some new stuff here for the fall. So thank you very much.
  • Video
  • Cloud Block Store
  • Pure//Accelerate

Users can easily protect on-premises SQL Server data by replication to Azure. SQL Server Management Studio enables disaster recovery in Azure with Pure Cloud Block Store. Both crash-consistent and application-consistent recovery are possible. This session explores how you can simplify and accelerate SQL Server DR and SQL standup in Azure VMs with automation and built-in integrations.

Additionally, DBAs and Storage Admins will learn how to enable application consistent snapshot management on-premises and in the public cloud with Pure Cloud Block Store. We will cover choosing the right volume types (Physical, RDM, vVol), succeeding with common use cases, and discovering the benefits of the Pure Backup SDK.

Continue Watching
We hope you found this preview valuable. To continue watching this video please provide your information below.
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
CONTACT US
Meet with an Expert

Let’s talk. Book a 1:1 meeting with one of our experts to discuss your specific needs.

Questions, Comments?

Have a question or comment about Pure products or certifications?  We’re here to help.

Schedule a Demo

Schedule a live demo and see for yourself how Pure can help transform your data into powerful outcomes. 

Call Sales: 800-976-6494

Mediapr@purestorage.com

 

Pure Storage, Inc.

2555 Augustine Dr.

Santa Clara, CA 95054

800-379-7873 (general info)

info@purestorage.com

CLOSE
Your Browser Is No Longer Supported!

Older browsers often represent security risks. In order to deliver the best possible experience when using our site, please update to any of these latest browsers.