Cloud Native Computing Foundation KubeCon + CloudNativeCon Europe 2021, 15 May 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Sponsored Session: Kasten - Introducing Kubestr: Identify, Validate, & Evaluate Storage in Kubern...

Description

Don’t miss out! Join us at our upcoming event: KubeCon + CloudNativeCon North America 2021 in Los Angeles, CA from October 12-15. Learn more at https://kubecon.io The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.

Sponsored Session: Kasten - Introducing Kubestr: Identify, Validate, and Evaluate Storage in Kubernetes

Speakers: Michael Cade

As the adoption of Kubernetes grows, so has the persistent storage offerings that are available to users. The introduction of CSI (Container Storage Interface) has enabled storage providers to develop drivers with ease. There are around 100 different CSI drivers available today. Along with the existing in-tree providers, these options can make choosing the right storage overwhelming.

Kubestr is a collection of tools to identify, validate, and evaluate your Kubernetes storage options. Kubestr can assist in the following ways: identify the various storage options present in a cluster, validate if the storage options are configured correctly, and evaluate the storage using common benchmarking tools like FIO.

A

Thank you for joining this 15 minute. Lightning talk where we're going to be talking about cubester, I'm michael cade, I'm a senior technologist at cast and by veeam and cubester gives us the ability to identify our kubernetes storage.

A

Validate our storage options are configured correctly, as well as evaluate those storage options using a benchmarking tool such as fio, let's get straight into the demo, so to start off with the demo, I want to just briefly give you two or one area really where we should um really focus on as to where to get where to get more information, and that would be cubester.io.

A

So, as you can see here is it gives a bit of a brief, a brief description of what cubester is and how it can help it gives you, those three identify, validate and evaluate, and if there's three things I want you to walk away with today is those three things.

A

It gives you a very short demo of how it works, but obviously this this demo, video, will give you a lot more insight into some of the areas that also works also gives you a breakdown of how to use cubesat as the as a tool so and I'll get to the the platforms that that cubester is available for shortly. But then it gives you a bit of a rundown on how to run certain tasks on there, but again we're going to touch on that during this this uh this demo.

A

So in terms of where can we run cubester? So if we head on over here, you can see that we have the ability to deploy cubester on windows, linux as well as mac os, so pretty much multi-platform enabling us to wherever our ide is. Allow us to to run this to to uh ensure that we can test and validate our our kubernetes storage.

A

Okay. So in terms of what it looks like so to begin with, if we have to take a look at the help command of the cubester.xe, you can see here that we have a very simple command line, driven tool in that we have the ability to run a csi check which we're going to get through as well as an fio now on its own, and that's what we're going to do here. This command is going to identify all of the available storage, the storage and storage classes that we have available within our kubernetes environment.

A

So, and what this will do is this will leverage the the cube ctl config or the context that you're using to be able to do that, and I'm going to jump between two as part of this demo as well.

A

So here we can see that I'm obviously using an aws eks cluster, but you can see that it does a validation on the version of kubernetes.

A

It does a role-based access control check and it does an aggregated layer check. It gives you a warning about the csr driver. Beer is depreciated and unavailable after after a specific date, but that's really just that warning that we see within the public cloud when we know that there's a transition coming around csi, which is one of the reasons why this tool is such a handy little tool- is that the the enhance and advancement of csi really does shine a light on.

A

Well, we need to make sure that it's configured correctly, so it gives you some information about the the csi driver that you have so we're using the aws ebs csr driver- and you can see here that I have a storage class known as ebs, hyphen sc and a volume snapshot class csi, aws vsc.

A

You can also see that I have a clone here as well, and it then gives us some instructions as to well. What do you want to run? So we we've got the the ability to validate, which is testing that csi snapshot restore functionality which we're going to do shortly, and we have the ability to perform that fio that flexible io test against the storage that we wish to do now.

A

What we've also got is that entry provider, which also may be something that you also want to get the benchmarking results from, and you can see here that we have a gb2 storage class available for that. But on this instance we can't run the csi snapshot validation because well it's not there. So um the fio test is available still in there.

A

So if we drill down a little bit further and we check out the help for the csi check, you can see that there's a few more options that come up or a few more flags that we're allowed to to start um playing with. In that we can say to clean up clean up the objects and the default is true anyway. So what this is going to do is this is going to go and create a pod, an application.

A

It's going to create a pvc, it's going to create a pv, it's going to create a snapshot of that pv, and then it's going to restore that back, or it's going to clone that back so validating that the snapshot functionality of csi is actually working as it should be. Now, if we didn't want to clean that up, maybe we were running some tests or wanting to contribute to cubester.

A

Maybe we want to keep those those pods up and running so that we could dive into them and and do some more work with them, and we can also dictate where we want the pod to be running by default. It will run in your default namespace that might not be applicable. We might want to run it in a specific namespace. So we've got that option, then a required flag would be the storage class as well as the volume snapshot class.

A

So let's go ahead and pick those two that we just saw from running cubesta and what this is going to do is run through that process. So it's going to create the application which is going to create the pod that we first mentioned and if I open up another window here just over to one side and I run a cube, ctl get pods you're going to see that this cubester pod is container created and it really doesn't take very long at all, and you can see already that that's running.

A

If I also run a cube, ctl get pvc, you can see that we've got our cubester pvc in there as well. We're then going to take that snapshot. So again we could do a volume.

A

Snapshot- and you can see there that we're just keeping up with what we're what we're doing here and what this is then going to do, is create that clone or restore that data back into the namespace, simulating that the csi is configured correctly now what it will do, and eventually this will this will complete. It doesn't take very long at all, but what this will do is validate and give you an okay or a thumbs up now.

A

If your csi is not configured correctly, maybe you don't have the snapshot controller configured quite right or within the correct name. Space is one of the areas that I've seen, especially within aws and ebs, and the csi driver that you manually have to implement.

A

Is it's going to come back and give you some information about what is happening? Why it's not working at that point? The cubester pod and the pvc will still be in place as well for further troubleshooting alongside that.

A

So you can see here that we've created the snapshot and now we're restoring that application, and you can see there that we cleaned up the resources. So if I now just go up and check that volume snapshot, we can see that that's been removed.

A

We should also see that the the pods have also that they're being terminated as well as then. Finally, on the on back on here, you can see that the csi checker test is that the application successfully snapshoted and restored, and that's really the the key key to success from a validation point of view to say everything is everything is good from that angle.

A

So then, like I mentioned, this is around being able to test against not just day zero, validating your storage options, understanding identifying your storage options, but also being able to be a useful tool on your existing clusters, and you may have clusters in various different locations. So in this instance, I'm going to quickly just jump over to my microsoft aks. So it's your kubernetes service cluster and I'm going to run the same test against my azure disk csi and hopefully that is still up and running.

A

But again what this is going to do is he's going to go and create that it's going to go and create that pod. It's going to go and create that that same that same process is going to happen here and if I go okay, so we're definitely on our aks cluster.

A

If I then run get pods, we should see our cubesat container.

A

Creating we should see.

A

The restoring of the application.

A

A

Now you can see the clone there in this instance, so this is the pvc that has been cloned from the from the snapshot and the restore is, is ongoing and happening within there, and you can see there. That was super quick because I'm using different nodes, I'm using different storage and compute nodes within my aks cluster, with much more memory, much more cpu, which does actually affect the storage capabilities that you have within your within your cluster.

A

So we could also run that against the azure file csi as well, but this is really to highlight that this is around. I validating that your csi driver, your snapshot and restore functionality or the snapshot. Functionality from the csr driver is configured correctly and fit for purpose from a data management point of view, but you can hopefully see the advantages of being able to see or validate that your configuration is. Your storage is configured correctly.

A

So then we get onto the evaluation of the storage. So let's jump back into our aws cluster and clear that down. So we can see so now we want to focus on running flexible io. Now flexible io has been around for quite a number of years. Open source gives us the ability ability to benchmark our storage systems.

A

So here you can see that we've got our options, our flags, that we can use again, namespace size string, etc. So if we just want to run a basic fio test using the default, which is a 4k, random, read or random, write as well as a random read, write 128k block size, then we can just get going there and that what this is going to do in the same vein is going to go and create another pod or another application.

A

But you can see here that it's now that fio pod- and this is going to take about a minute to to run through and run that scan against the storage, ebs, hyphen sc.

A

So it's running that fio test that I mentioned the default fio on storage class, ebs, hyphen sc, with a pvc size of a default of 100 gig, now you'll see in later commands that I can run that against the 400 gig or bring your own or pick your own size that you need for your workload, and that's really. The key here is that I'm demonstrating something against some storage that I really don't know what the workload looks.

A

Like but you will be looking for well, what does my database need to look like in terms of performance with x amount of users, so you can then choose what that looks like from a from an fio perspective, and what you can also do is obviously run this from a day, zero point of view, so that you know that the storage that you put putting this on is performant enough, both the storage but also the compute, and then also being able to run that against um like day two operations and make sure that the storage that you've already purchased or the ones that you're already consuming is also configured correctly.

A

So you can see here that just from a random write and this, this is how it, how it outputs that fio results.

A

So we can see that for a random, read iops point of view with a 4k block size that we get an iops value of 461 for right, it's 507, then you get to see that block size, change to 1 1 to 8k, and you start to see the the the iops decrease actually, but obviously gives you an idea of what that looks like. So we could go through that and it would run a similar and potentially give us more iops against that.

A

We can obviously jump back into our azure and run the same test against that. The other interesting piece here is that we can also jump into the pre-configured or community-led fio configuration files and use those as our as our testbed. So if there's one that you generally use for your application or you've created your own for your application, where you can bring your own minus to f, bring your own bring your own file, and here you can see that on that it just walks the walks. The um the file system backwards.

A

Reading the file so and then gives you the the output to that. Cubester is a handy little tool that enables you to explore your kubernetes storage options by being able to identify the storage options that you have available within your cluster, both from a day zero point of view and onwards.

A

It validates that the storage options are configured correctly, especially around uh data management tasks such as snapshot and restore, and then it also allows you to get some performance benchmarking in the way of evaluating that storage using a flexible. I o hope that was useful thanks for.

A

A