 Okay, I yeah, I think I think we're live now. So Let's find out a little bit about KubeCon and we'll jump into this extreme of cloud native live It's more than just open source It's about Connecting with people. It's about being part of the community It's about sharing what you know and helping others KubeCon is the best place to get hooked into the community and learn from everybody and let me tell you people This is just the beginning Cool. So welcome everyone to this week's episode of cloud native live where we dive into the code behind cloud native I'm Bill Mulligan and I work at CNCF Every week we bring you a new set of presenters to showcase how to work with cloud native technologies They'll build things they'll break things and they'll answer your questions So join us every Wednesday at 3 p.m. Eastern time and this week We have Michael and Cherise and they're here to talk to talk to us about Kubester and also just as We talked about in the video join us for the KubeCon cloud native con virtual at EU May 4th The 7th to hear the latest from the cloud native community And just to note this is an official live stream of the CNCF and as such it's subject to the CNCF code of conduct Please don't add anything to the chat or questions that would be in violation of that code of conduct Basically, just respect all your fellow participants and presenters and with that I'll hand it over to Michael and Cherise for to kick off today's presentation Yeah, thanks Bill. Yeah, so just a bit of a brief introduction. So I'm Michael Cade I'm a senior global technologist here at custom by Veeam Yeah excited to get into what Kubester is and how it works. Cherise, why don't you introduce yourself? Sure. I'm Cherise. I work at casting by Veeam as well as a software engineer there. I'm glad to present Kubester to you guys Yeah, I think a bit of an introduction to What Kubester is is a handy little tool that allows us to well identify Validate and evaluate our Kubernetes storage. So taking Taking those potential Hard tasks, especially for the platform operators the new guys to Kubernetes Bringing taking that that potential hard tasks bringing that into an easy Really easy tool or handy set of tools and running that against our Kubernetes storage Whether that's day zero or or even existing clusters and giving us some information or giving us some validation about that about that storage so One of the key parts here is that it's going to give you It's going to one it's going to identify the storage that you have available within your storage And we'll get into what that demo looks like shortly It's going to validate in this particular instance It's going to validate against your the the CSI driver is configured correctly. I The snapshot functionality is up and running and working And then from an evaluation point of view It's going to enable us to benchmark your storage or this particular storage class against flexible IO so another open source project that enables us to Get some benchmarking Stacked out of your out of your storage and Yeah, why not let's jump into its duration and actually jump into a demo just to show Show everyone what what it is and how it how it looks? Sure thing. Um, I want to let me share my screen Let me see my screen Not too small That's good So, I mean kind of where Cube sure came about is You know, we've had customers and the customers generally have are like new to Kubernetes and have tough time kind of figuring out Their storage options and so we kind of built this tool in-house and then realize that maybe there is a Benefit to the community and the community could actually cease Use this tool and gather some insights onto their storage. So You know, I downloaded the tool here Cube sure and You know, normally what somebody would do when they want to see what storage they have is they'd run something like cube cuddle get storage class Storage class right and it would tell you storage classes that you have on your cluster. This is digital ocean cluster. So it has You know one storage class deal block storage and this is a driver or the provisioner But that doesn't tell you a whole lot But if you run cube stir on this cluster instead It does a little bit more it the first thing, you know, it checks you that you have a valid Kubernetes version Some RBAC check check gives the API layers enabled. I mean accessible some warnings here that come with The CSI CSI drivers changing because yeah, it's been rapidly changing in the recent year or two And now we have more information about our driver, right? Initially, we just had a storage class But now, you know, we headline it with saying this is the provisioner that you have on the cluster Points out that the CSI driver It'll also tell you, you know, some additional information about the driver itself. What what kind of features it supports And that points out a storage class Which is what we saw earlier, but also that there's a volume snapshot class here, right and Once you have both of those you should be able to take a volume snapshot and Hopefully restore from that volume snapshot, right? So it's a cube stride again has those checks or has the ability to validate Whether your your provisioner setup takes snapshots correctly, right? Digital ocean does a good job of when you create a cluster it sets up all this infrastructure for you correctly But for a lot of provisioners, it's not always straightforward to set up that and that kind of Snapchat and capability Many steps involved. They have to install many different types of objects and then even after all that You still may have some errors. It's if you're new to Kubernetes. It's really hard to kind of debug these errors So we thought this would be a handy tool to kind of validate that. So let's let's go ahead and run this check CSI check Just to look at the options that we have here We do create So in order to run this check, you know, what are the things that you normally do you create an application, which is a pod in a PVC then you take a snapshot of it and That creates a volume snapshot Object, right and then eventually you can restore using that one essential object So then you know the cleanup flag that to clean up cleanup those Those specific objects and then you can always write in a different namespace that's not default And then yeah, so I mean the more required things are in storage class in the volume snapshot class So let's run this check really quick We also have a question that chat from Carlos shirt Is there any application for this tool to backup at CD? So this is not a backup. No, no, there isn't a backup Functionality, it's not a backup tool. It's more of a tool to validate your storage, right? It's not necessarily a backup tool on its own Yeah, that's your question Carlos. Yeah, hopefully that's useful Carlos, but I think This is another area where we thought from a custom by beam point of view or a custom point of view is that by someone coming in that wasn't affiliated to the storage vendors and the storage market it actually allows us to Create something that that is more universal across all storage arrays, but also public cloud storage But yeah, this isn't really focused on on the backup of of anything It just like Sarish says it validates that the CSI driver is configured correctly and that's what we're going to get to and show that but Yeah, it's not it's not a backup tool say Okay, so Maybe I have a question that might help the audience understand a little bit more Could you maybe explain like why someone would need to benchmark their storage in their club? Yeah, you go you go no you go ahead So I was gonna say around so benchmarking storage isn't a new thing and it's not definitely not a new thing for Storage in general whether it was whether it's cloud native today, whether it's cloud whether it's Virtualization whether it's physical Benchmarking our storage for particular workloads has always been a bit of a dark art and a challenge and Like there's no reason why anyone today in fact that was the way to do this is you would go and create your own application you would create your own set of Tools to be able to go and run this this test and I think what we've found is that it's a very manual process and it unless you Know the depths of both Kubernetes and how you create a pod you create a PVC you create a Persistent volume you attach it and then allow your application to run Then that becomes of it becomes even more of a challenge I think the other the other element to that is that we're obviously there's a lot of people coming into into this space That are traditional platform operators coming from a virtualization point of view Storage point of view that maybe don't know Kubernetes to be able to spin up That that pod and the PVC etc. And this is just taking one of those tedious potentially long-winded processes away and automating that process for you and I'll get into a little bit more about the challenges later on as well in between Sarich's demo around things like making sure that we choose the right storage, especially when you've got the We've got so much choice. It's kind of overwhelming a little in that you've got different various different types of storage I think this also helps Narrow down that choice. It's great to have that choice. We're not complaining about that But being able to choose the right storage option for the workload that you need is is important and to do that as fast as possible because Ultimately, if you're especially in the public cloud is well, you're you're paying for that storage And if it's over over provisioned under provisioned Obviously under you're not gonna have the performance that you need over provisioned Well, you're definitely gonna be paying more money or over the odds for that. So There's a few more challenges in there and I'll get to them But I think if we jump back into the demo series, then I think you're probably gonna answer some of those as well So, yeah, and we'll get back to that Yeah, so let's go ahead. Sorry, there's also like two more questions in the chat I don't know. We can also answer them as we're going if this is something yet you have coming up. So Um, no, we can answer these I see the one from Rajesh saying they're in limitations The limitation. They're not really any of the limitations to do the CSI Snapshot restore check. We obviously needed a CSI driver And then in terms of running the FIO check As long as it's on Kubernetes and it's a Kubernetes Provisioner storage provisioner, then we should be able to run the the FIO test against it So so yeah, there's no real limitations. It's just Kubernetes Has to be Kubernetes But yeah, what's the other question on here? I Don't know whether it's possible to connect sand storage in Non-premise data center or object storage buckets from the public cloud to my Kubernetes cluster Is this something this project could help with or is that outside the scope? This is outside the scope But if you were able to connect a sand storage to a Kubernetes cluster, then you would be able to Validate its performance and whatnot using kubecture, right? I won't help you actually connect it But but yeah, I can definitely definitely help you validate that it that you've seen the performance that you want Yeah, and I think it might have been a broader broader question from Deepak as well is that Is it possible to connect to a non-prem sand storage in like in one location to a public cloud? Kubernetes cluster say in AWS or in Microsoft AKS or something along them lines and I guess the answer is yes if there's connectivity in place, right? That's the that so it can be done There's some storage vendors out there that offer close to cloud type type offerings, but yeah, this tool is not going to Get that configuration for you. It's not you're gonna require that direct connect type type situation that connectivity but what cubes to can do is well validate that it's been configured correctly from a CSI point of view and And just give you the the forefront that is actually there and configured as a as a storage class within your within your cluster Cool. No, the question is keep on coming in like fast. So pretty popular. Yeah Um, just keeps to extend beyond the standard CSI benchmarking checks, maybe something that is available like out of the box When what do you mean by CSI benchmarking checks, let's put it that way because are you trying to benchmark? So the benchmark kind of benchmarking that that keeps your does is Like storage Performance, right? It's a performance benchmarking So whether it's CSI or a native entry provisioner You know, it's still all just storage and we treat it treat them all the same But is there anything more specific Sumon that you had about CSI benchmarking checks? I guess we can get back to that But I mean FIO is a test that we run Cool. And yeah, maybe you want to like dive back into the demo and I'll answer some of the questions that they have Of course, um, so yeah, let's uh, like I said where we left off was we wanted to run a check to see if You know We can take a snapshot and restore from a snapshot so that basically checking if this provisioner set up correctly And now so you see that the first thing it is is doing is creating an application to create a pvc and it's creating a pod it's waiting for this pod to You know become become live and then once it's live I'll take a snapshot and you should see a volume snapshot object getting created here um, and like I said, these these are not very, you know If you're used to kubernetes, they're not very complex things. These are things that kubernetes operators do on a day-to-day basis Taking a snapshot restoring the application, but if you're new to the the landscape It can seem like a bunch of different steps just to validate that you can take a snapshot correctly, right? So, you know, hopefully this creates that, you know One command tool that creates an application takes a snapshot and restores it and then validates, you know, that Um, the data that was in the application originally is still there on a restore, right? So it kind of does that end-to-end workflow without You know all the various steps that it takes to do that um So, um so that's That's just the csi check tool. We also have The fio tool, right and Let's go ahead and run that on The same storage class, right? Um Again creates a pvc creates a pod And i've seen this test take about um Maybe around Like 40 seconds or so So yeah, so we the pod spun up and essentially the pod the application here is an fio application, right? And uh, we have a 100 gig pvc And it's running this fio test and you know the default fio test that we have here Um, it does 4k random reads and writes and then 128k random reads and writes. So it's actually four separate tests and um You will see the output in a second Uh, well So it shows you that it took to take around 30 seconds around this test. Um The four different jobs that it ran and you know, let's maybe keep note of these numbers, uh somewhere here um Let's just Just say around, you know, 1900 uh read iops Uh, 1300 write iops at 4k, right? Um, you know, they seem like reasonable numbers And you know, maybe you're like, hey, that's good enough for me I'll go digital ocean for my application But you know, maybe you want to try some other another cluster, right? Um, and I have another cluster here. Let's see I'm switching over to my gke cluster And now we can run custer again here Carlos fio center flexible io flexible io tester So Again, this is you know, it looks a lot like what we had in digital ocean shows you the kubernetes version Uh, our back checks are okay Aggregate api layers working correctly But here I have two different provisioners. I have the csi provisioner And then I also have the entry provisioner. You'll see the csi provisioner has a volume snapshot class um And you know the entry provisioner doesn't have that kind of functionality. So you want to see that there Um And you know in this case, you know, we ran we ran it on a csi provisioner last month started on the entry provisioner at this time We're on ssd. Hopefully it's the fastest Storage option they have um, again, we can follow it along. Uh, it's creating a pvc 100 gig pvc it'll create the Um pod and then it'll run the test And I've seen this test run about a minute in in the past. Um, I can add some questions in the meantime, but um See Yes, it's starting to run the test Yeah, I guess one question that I have is Like why would I benchmark my storage in the cloud? Like what is this kind of information? Does it give me? Sure, um So like if you had an application, right? And maybe it's writing some information writing something to a database right Storing maybe uh, some some records to a database, right? You want to know if you know your storage can handle the bandwidth of your application Maybe you have thousands of users. Can it can it actually handle thousands of users writing data to your database at the same time? right And that's something that you know, you need to figure out what your Uh, IO requirements are to your storage and then you can How to figure out the right kind of storage to use for that, right? So maybe this will help you say hey, you know I cannot use slow storage like slow standard Spinning this any flash like high-speed flash storage and you know, you maybe want to benchmark that That's kind of that's a use case. That's a use case. I'm sure there's other use cases out there, but I'll talk to my head That's when I could think of Um, does that can answer your question? Yeah, absolutely I think it totally makes sense. You don't want to have something that that doesn't work for your application and give a bad user experience Seems like we have a couple more questions um in the chat now too So the first one is from j warner could be a silly question the benchmark function. Is this optimized equals HD parm but for the cloud Um, I'm not sure what HD parm is. Um It's not necessarily going to optimize your storage at all It it's just to to validate that your storage is fast enough For your application or it's suitable for your applications, but it's that way Yeah, so They said there's the Linux command HD parm. I don't know if you're familiar with that I'm not familiar with that but uh fio is also I don't have it um but uh, no, it's it's actually simulating the fio application um here, maybe if I can post a link to fio Yeah, if you just send it to me I can post it on the screen Sure. Yeah, that's what i'm gonna do Yeah, I think I think um HD parm is more aligned to like art 64 type frameworks But so it runs similar things but gives back Like information of disk does a varying amount of Like spindle tests. It's not really the same. Maybe there's a bit of an overlap with fio So it could be seen as similar, but yeah, I think it's like 64 or arm 64 um Yeah, cool. There's a couple more questions in the chat now Is there a way to benchmark volume provisioning attachment with kubester? Maybe compare multiple csi implementations from the same storage provider Yes, um So, I mean, if I understand the question correctly, you're saying Is there a way to um run fio with kubester or check if the snapshot Functionality is working because you can do both and you can do them with multiple csi Providers, whatever csa providers you have in your cluster, right? Um, so yeah to answer your question. Yes, it's possible Cool, and it looks like we also have an update. Um from joe about hd palm is more about getting the hardware parameters of the physical Hard drive. Um, so it looks like that's more on the hardware side where it kubester is more on like the software side Is that right? Yeah, kubestrian on the software side. Yep. Um, I Joe, I'll I'll take a look at hd palm and you know, see if there's some sort of, you know Benefit that kubester can provide using hd palm, right and that's part of um Part of it or it's still kind of in its infancy and we kind of want to see What other tools we could add to this little toolbox to make it? You know A kind of a a go-to for all storage storage needs in the cloud, right? A little toolbox Yeah, cool. Um There's another question. Um, can we simulate load testing using kubester? um If you have an fio configuration that simulates a load test you You might be able to maybe I don't really understand the question but But I hope that answers it Yeah, and I can um send a link uh the the link is here in the conversation if people are Looking for it And I can drop it. Yeah, this is a It's a direct link to the fio um, so cool, um If you want to get back to demo really quick You'll see here that we have You know the iops that we're seeing here are nowhere near what you know, we experienced with Digital ocean, right and it doesn't mean that google is is worse by any means. It's just that maybe we haven't configured it correctly, right? um, and What you'll notice is that if you look at google's um storage I feel if you look at here, let me maybe I can post a link. I'll tell you that um Yeah here. I'll see so And this is from google and it tell you how to you know get the most performance out of your um Out of your storage and what they say is if if you um Use bigger volume sizes. You're going to see better ios, right? So, you know fio. Sorry our kubster test has the ability to Do Um It has the ability to to specify a size. So, you know, let's try you know our default's a hundred gig. Let's try something bigger Let's do 300 um and see kind of what that gives us instead Instead of the hundred gig size This one actually runs a little bit quicker than um the previous one because Like I said, the the bigger the volumes in google the better performance you're going to see This one takes about 40 seconds I think that's another thing to mention saris as well, right is that okay. So this is going to be testing The the blocks that the io is testing against iops within the storage that's available to you within your kubernetes storage But also so as saris just said by changing the size of the persistent volume or persistent volume claim That dictates what iops is available to you But also think about the underlining node Architecture the compute nodes that you use that can also dictate or change depending on what cloud you're in they change on What what storage performance is unlocked? So saris has done an awesome cncf blog on on like it's not a comparison But it just goes to show that overwhelming choice that we have out there from a computer and storage point of view And that no one's a winner in in across the Best performance, but it does show that there's so much options here. So many options here And yeah, I just I just sent you the link. Um, yeah, you posted it out there It's linked to the blog and I kind of you know go through the same kind of test Of along like, you know the four different cloud providers digital ocean Gke azure and aws And you know just kind of compare and contrast and kind of see How tweaking little things here and there between the different providers gives you, you know, much different like end results Right and yeah, like it's hard for you to if you're jumping in to make that decision Of which one to choose without having a tool like this that can quickly, you know Run some sort of a test and get you some metrics that that means something to your application, right? um oops Yeah, it's let's see. So There you go So it seems like there's a question from a user on linkedin. What are the best storage Back up third party solutions for aks and azure and it seems like you could use cubes to help figure out Um, like exactly what your application parameters are and like which storage is best to meet the needs of that application. Is that right? Yep, precisely um, yeah, so like I said there isn't It's hard to say. What is the best storage without knowing what your application is, right? um And say it really depends on your application and your budget and then you can pick what storage works for you um No, deepak. I'm not right writing. Uh 300 gigs of dummy data. It's it's a 300 gig volume and I think um So I I think the the file sizes here are two gigs. So it's two gigs of dummy data on a 300 gig volume Um, but the size of the volume itself dictates the kind of performance you're going to get at least with google, right? Yeah, that's a that's a good point though as well. Um and Out of the box cubes to use is like so the default fio test is a random read and random write at 4k block size As well as 128k block size, but you can there's a there's a huge library repository of fio configuration files that you can kind of bring to the bring to test Uh as well, um Yeah, yeah, so so that so it's a rich there showing what the fio configuration file is so if you know that your workload works in a 8k block size or You need to walk back the file system or in a reverse manner. You know that it's a sequential write All of these changes could be could be modified in your own fio configuration file So you can run that against the storage as well Yep Um, let's see. So we want to run it against our same ssd storage class Um, let's make it the same 300 gig volume. So it's fast and then Uh, let's see. I think it was minus f and then Basic verify Let's see So so yeah, I mean you don't have to use our um You know default test really comes down to your application, right? So it's kind of know your applications Uh fio need like signature or what what that looks like What your what your um load looks like in terms of an fio file You can pass in that fio configuration here and then, you know, we'll run that particular fio test So it's running this particular test on that storage class on a pvc of size 300 gigs Let's see. I think this one takes about 30 seconds or so So what else we can also, you know write all this output to uh Out as a json. So if for some reason you wanted to scripting where you wanted to run multiple fio tests or Whatever whatever your needs are you could easily, you know Get this output back as a parsable json file and then, you know, maybe right collect some data that way or If you actually try to make some pretty graphs or something like that to kind of showcase something about your data, then That's what you can do. So yeah, this is uh, you know a different IO pattern and you're seeing Much more different results, right? Like I said, it really comes down to what your application is and if you figure out what your application needs are then You know the stool can be very helpful Yeah, we just posted some of those examples for fio as well. Um that you can you can run on there um I was going to cover some of the ah, okay, so there's a There's another question. Do you plan to have a crd and a controller for cubes so that it runs directly into Okay, it's cluster and is declarative um Yeah, there there were plans of that initially For now, this is like I said, it's infancy and it's what it is right now. It's just uh executable But yeah down the line we could make it a crd CRD base, you know operation um Yeah, these are all kind of things that we have in our pipeline. Uh, we also kind of want wanted to make this more of a A community-based tool. So we really wanted to get um Like people kind of pitching in like running their own fio tests and telling us What their results are so maybe like a leaderboard or something which other people can go and Get a better understanding of how other people's storage is performing, right? We kind of wanted to we're hoping to kind of move in that direction where we have more of a community involvement with it or community participation um But yeah, like I said right now, it's infancy We just want to see if it's actually useful for people before we take that next leap and And do some cooler stuff with it Yeah, so maybe if people want to get involved like how can they do that or how can they try it out or test it out? um I'd suggest start off going to cubes your i o installing it and running it and um, yeah, we also Again, we're open source. So go to our github page. There's links on that for that on the website as well. Um, and yeah Work the repo, uh, feel free to make some changes Yeah, visit us at github as well And maybe I'll already I go ahead But can you tell us a little about like the roadmap or where you see like the project going? Sure, um, like I said the first a few things are Additional tests that may be useful, right? there's Some people want to be able to benchmark their object storage. So that was something that I was looking into This is persistent storage. It's not the same as object storage. So You know, maybe we could kind of foray into that venture into that a little bit. Uh, but yeah, like I said The other bigger thing thing is to make this a more community driven project have like like I said some sort of a a leaderboard or some sort of A place where people can come because you know, we don't want to run the same test 100 times 100 different people are in the same 200 times if you can just go look say, hey, I have that exact type of Uh infrastructure. I can see what kind of results I'm going to see Uh, it makes that decision making process easier, right? Um, obviously we'd like, you know To get more involvement from some of the bigger storage vendors. Maybe have them Write specific Fio tests like that say, hey, if you're looking for a certain type of application, this is an fio test that Would benefit you and you should run run this particular test for example Um, so yeah, I mean, there's there's a couple different ways that we've been trying to Uh, a couple different options on our roadmap. Um Like I said, we're still kind of seeing what what's the most important need right now and then moving on from there Yeah, I would I'd just add to that point as well. So we should round the uh So we mentioned storage providers storage vendors that obviously make sense but also like database Vendors as well. Like if they have a they're going to have a better understanding of what they're What their database looks like at load? Obviously, that's not going to Uh reflect on every single individual customer of theirs but if we can if we can team up with database vendors storage vendors and get a much Wider view of what that application looked or what that database requirements look like And you start putting fio specific tests around that to understand what that looks like and then have a A library of these results. I think that becomes very useful And I've been calling it over the over the last week that this has been released Is this is really that easy button to understanding your kubernetes storage really is So as serish kind of Very easily went through it's a super simple way of being able to test against your your storage But it's a lot quicker than having to manually go and create pods create pbcs create pbs Then understanding actually how the application is going to run against those This is literally download to your mac os to your linux box to your windows box from github And get going that's that's as simple as that and then yeah to go back to the roadmap point the other the other Thing would be nice. I think around that leader like leaderboard would be something around Like vision. How do we visualize? How do we make that that output that cli output? How do we visualize that? How do we put that into some sort of visualization so that we can compare and see it Much easier than having to scroll through endless command line type text And yeah there to that point there was um You know right now I just give you the fio results, right? Like maybe down the line we could do some more analysis on the fio results, right say that hey, you know your Tell tell the customer or tell the user why What is impacting their performance and maybe you have? Um a bottleneck, you know because of the storage size or something like that, right? So I mean that requires a little bit more understanding of the storage and the results of the fio test itself, but Down the line we could see ourselves giving you more A more descriptive output than just these are your fio results Cool, and it looks like there's Actually people trying it out right now Um as we're on this so I got this error message when Cubester pulled the image failed to pull image user cannot be authenticated with the token provided Give a quick way to allow cubester to pull The image it's a live debugging Yeah, um I wonder what image it's struggling to pull because kubester I I believe the images that are all publicly available. I think they're docker hub Um Is it gonna be authenticated with the token provided? Hmm. It's kind of hard to tell without a little more more details. Um Uh Boon to a twitch, uh, do you want to you know? Maybe add some more details to our github a pretty github issue and add some more details there and I'll help you debug that time Yeah, definitely Yeah, and it's cool to cool to see people are actually trying it out like as we're as we're live right Now It's pretty cool Cool. Yeah, so I think if you create that issue on github, you'd be happy to help you out. Um, and then I guess going back Um, Going back to like the leaderboard a little bit Just checking if you see like any fluctuations in regards to the IOPS over time from different cloud providers So if we do benchmarking and like have like a leaderboard, would it change over time? Would it continually update? What have you seen so far? Well, like I said right now, we don't have enough data To see if there's been any like trends, but I'm sure if like the cloud providers Um, I mean they have so many different offerings, right? Like it's not always Like they like hey, I have one type of storage and one type of node. They have Nodes optimized for storage or they have faster storage slower storage. So Yeah, you I mean you will see them like the cloud providers offering faster or better Storage options as you know, they they progress as well. So Yeah, you'll definitely see fluctuation in what cloud providers often offer All right, they're trying to be competitive amongst each other as well, right? I am I just posted the the slack as well. It could be a good good area to start that debugging process as The other thing I was going to mention was around Um, obviously CSI especially in the public cloud. I've been playing a lot around I think Saris you mentioned it like digital ocean out of the box is CSI first they They deploy with CSI. You don't have any entry provisioner, which we know is Going to become the norm. Um, I think it's the next version of kubernetes Maybe even the next one, but it's very much CSI first as we as we move forward And it and it potentially won't take that configuration route or I would say the pain points as a new beta to this world is the CSI driver implementation having to having to install the driver is a is a A little bit of a challenge at the moment. It's a manual task at the moment. So Clearly on the error there, but there's a there's a there's a task that you have to go through And I found that cubes does a really good kind of full stop to understand even whether there's something wrong. So In Saris's demo, he shows you all the good stuff that it says okay that the CSI snapshot was was successful And he could restore back to a clone I've seen many more of the errors that come with that where the snapshot wasn't successful Um, and it's just really really short sharp to the point Let me know what it is And and then we go and troubleshoot that and it's generally going to be down to you Probably your I am or your secret credentials that that um at least in AWS and again probably user error, but I think that as we're in that transition period of CSI becoming the norm This is going to be an awesome tool moving forward and it's not just for that day zero cluster rollout. It's going to be kind of to to the last question around Do we see IOPS fluctuating? Well, this is something that we should keep checking because At the end of the day, you're paying for that service that storage on the cloud provider You keep using kubester to validate that you're getting the resource that you you should be getting right No, definitely Is there anything else you two want to show today? um I was going to kind of show the same thing. I've got an AWS cluster that I could jump in but it's going to be very much the same Same stuff that saris has shown. So I don't know if we need to go in there Um, I think the other thing again saris might have touched on it is around So as long as you've got kubectl access to you to your cluster That's where kubester will run. It will use the kubectl context to run these tests in So if you don't have kubectl kubester is not going to work um Also, if you if you've got many different contexts like saris showed Is you're going to have to jump manually between those? Before you can before you can run kubester against the other The other clusters, but yeah, I think I was just going to show kubester.io as a as the first resource because on there um Is a very brief description kind of what we touched on at the very beginning around identifying Validating and evaluating your kubernetes storage and hitting the easy button And kind of touching on that csi challenge that that I just touched on Then a short demo from saris on or how to use it And then some steps on where to grab where to grab the source Source files so you can install it on the on the os that you want to run it on and then a few Brief instructions on how to how to get started and run that um Then I've also shared around the flexible io the configuration like the the ready made community driven fio configurations um Yeah, I think That they'd be and then also the slack that that we've just started. So it's very Fresh in there. This is only this has only been rolling out like since the The first of april so it's really really new and uh, yeah looking forward to the To the feedback from the community and and driving What what this could be I think yeah, it'll be be fun Anything you'd add saris at all Uh, no, I think you covered most of it. Yeah, like like you said, we're I'm excited to put it out there I started to have people actually using it and you know really hope it benefits people and they they can You know see its value and Eventually contribute to it, right? Um So yeah, I'm looking forward to that. Um, I can't I think you're all A classic like Yeah Clearly I got to go back to purgatory like okay one more one more round of lockdown just so you can figure out mutes again Okay Thanks everyone for joining the latest episode of cloud native live it was super great to have both michael and Sheree To talk about kubester. I'm really excited to see how this project grows and the community that comes up around it And really seeing like those benchmarks. Um, so we can track story performance over time and see if what you pay For is what you actually get. Um, I also really loved all the interaction that we have from the audience I think this is really tons of questions coming out Super excited to see that even like people trying it live like as the session is going Like super fun to see it interaction Um, and yeah, looking forward to see the audience again Just a reminder we bring you every single week the latest cloud native code every wednesday at 3 p.m. Eastern Um, yeah, and if you haven't gotten your ticket yet for kubecon, uh, then there's a code for you to do it So, uh, see you around the cloud native community. Thanks