VMware Velero Community Meetings, 1 Dec 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Velero Community Meeting - Dec 1, 2020

Description

Dec 1, 2020

Status Updates

carlisia
Community support this week
Internal presentation yesterday
Finished the Tilt setup for Velero (PR: https://github.com/vmware-tanzu/velero/pull/3119)
PR reviews + starting on CLI v1.6 issues

bridget
Updated design for multiple credentials support
Needs further thought on the impact on upgrades and Helm
Updated the docker actions used in CI
Spent some time on community support last week over Thanksgiving

nrb
Bereavement & holiday last week
PR reviews yesterday

Discussion Topics
carlisia: Tilt setup demo
Community Shoutouts
@sseago: Design doc for RestoreItemAction wait for AdditionalItems to be ready
@pranavgaikwad: restore progress design doc
@funkycode & Michael Ketslah: pass annotations from scheduler to created backup
@shellwedance: Modify function name typo
@MadhavJivrajani: Add instructions to clone repo for examples
@jenting: feat: add delete sub-command for backup-location
@dsu-igeek: Add an E2E test framework to test Velero across cloud platforms
@damoon: [velero] fixed dnsPolicy position in yaml definition
@jaredallard: [velero] Support setting dnsPolicy on restic daemonset and velero deployment

A

B

Hello and welcome everyone- and this is the valero community meeting and open discussion today- is december. 1St 2020 uh we're in the last month of 2020 we're almost done everyone, uh so we got some status updates and then we'll dive into a uh demo uh with carlesia. Later today, hey.

A

Jonas, uh that's a really cute picture of a little I'm.

B

Sorry, I think.

B

I'm sharing the wrong thing there we go uh that should work better right.

C

No, it looks half as good now now I don't.

D

C

Be reading your texts.

B

Thank you uh all right, uh so yeah, let's start with some status updates. uh First up, we got kalisia.

C

Yes december y'all almost uh done with this year, so I'm on community support this week. I did an internal presentation yesterday to help people who do uh support that includes filler support. I finished so my phone project.

C

Last week there was a holiday week. I I didn't take any time off, so I worked on this tilt setup and so there's a pr for it. I'm waiting for reviews. So what it does basically is speeds up developments considerably and it's pretty cool. Just in a matter of seconds. You can't update the velour code base or the or any of the plugins and that just get the binary gets rebuilt sent to the container, but the container doesn't get remade.

C

It just gets restarted. So in no matter you don't have to do anything manually. You don't have to set it reset the image you just re restart and you have your code change. So that was what I worked on last week.

C

And um what else so I have I do have I there are so many we have so many pr's, so I want to go through them and see. If I can help. I don't know. I don't remember anymore what years I haven't reviewed, so I want to go through them and see. If I can move things along, I am starting to get.

C

um There are some cli issues for that we slated for 1.6, so I want to get started on them, especially the ones that are going to be needed for setting secrets, which is something that bridget's going to be working on. So I want to get those out of the way first and just thinking if there is any way I can set up for when we do have multiple secrets to already start, facilitating that.

C

That's it for me and I'll do a demo of the tilt thing when the the discussion topics come up.

B

Awesome any questions comments for coliseum all right. uh Can you see any blockers this week.

C

uh No, I just think uh in general, we are, I don't know we have a lot of prs to review. I have a pr that's been sitting there for a long time. um In that I don't know, that's it.

B

All right sounds good. um Next up we got bridgette.

E

Hi everyone. um So, as carly mentioned, I have been working on the design for multiple credentials, support um among other things, and but this is the the one piece of work that I'm really going to be trying to focus on now. um I did make some updates to the design dock um and marked it as ready for review. So if um so, if I get some input or some feedback from others on the team, that would be a really great um it does need.

E

It does need a little bit more thought and consideration for the impact on upgrades and also helm. um I had you have some thoughts on that and I will update the dog further, but I mark it's right for review so that um I can get some input on it in the meantime.

E

um So yes, others than that I've just been doing various other pr reviews and I spent some time in community support last week, while other folks were out for thanksgiving I'm in canada. So I didn't have the the time off um and also just made some small changes to our ci process as well, to upgrade um the docker actions that we were using because we were getting some warnings in our ci builds uh related to the versions that we were using uh yeah. That's it for me.

B

All right sounds good. Any questions, comments for bridget.

F

um Are you going to take the pr that's existing for multiple credential support or open a new one, or how? How are you and carly see I going to proceed with that.

E

um So I've just updated the existing pr, that's there. So I had written some comments. Carly and I had just a discussion about it and during that we decided just to update the existing doc. That was there a lot of it. A lot of the the content was still applicable, um so it was more just building on what carly said already yeah.

E

So it's still the same pr.

F

A

B

All right, um so do you have any blockers this week, bridgette.

E

No blockers, but I will uh I'll try and set up some time with garlicia to talk about the next steps for the cli changes. Cool sounds good.

F

All right, uh nolan, um not a whole lot of updates. For me, um I plan to work last week, but unfortunately I took bereavement and then holiday today or yesterday I was catching up on pr. Reviews got several of them merged. There are some larger.

F

Some larger design docks that haven't gone in um just because of how big they are in time and it takes to review them, but we got a bunch of of things uh reviewed here um so yeah, that's that's! Basically, all I have for today.

B

Sounds good um any uh any blockers for you this week, nolan anything! You need help with.

F

um I think right now it's gonna be getting the design stuff out. um This is in the community. Shout outs, but um one thing I do want to highlight is genting actually started talk started taking the um the command line um stuff, he started grabbing some of the cli work, um so that is moving along um and yeah. If we can move through these, um that would be great, and I missed some of the community.

F

Shout outs from last week in my script- only has a seven day uh window, but there were several other design docs that got merged um late. I think last week or the week before um that I'll need to dig out and thank people for and let them know they got emerged.

B

Sounds good uh all right, uh let's dive into the discussion topics uh kalisia you got a demo I'll stop sharing my screen.

C

All right, let me try to do this.

C

Okay, so what I'm going to do? Let me start: I was just trying to pull out the my pr and show the documentation.

C

Sorry, I didn't do this before I was trying to get to it.

C

All right, so, basically, I wrote a documentation to help people understand what this is and how to set it up. If you do want to just do it without really getting into it, there is a tldr, it's very, very straightforward. So basically, I'm going to show you there is this. Let me make this bigger.

C

There is distilled tilt resources folder with a the examples folder, so you just drop. All of this will have your personal tilt settings already filled out. If you have your plugins in the parallel directory to valero in the directory parallel to valero, this will just work.

C

uh Your cloud credentials will go here. My because I use medium mine is just like menu menu one two three. This is a sample filler deployment. It's already ready to go um with the other plugins set up, so the the three plugins that we support are included here. There is a rustic deploy demonstrates uh as well.

C

So if you want to deploy that you just put on the tilt setting, you say uh set up breaststick, it's triggered to 2 the setup menu is if you want to run menu in the cluster, so the the file will fetch from this examples. Folder has a menu deployment, so we will just point to download that into the cluster the menu will be up and running. You can use that so and then inside the config directory we have is where our crds reside and we have simple files there.

C

So we also point you should copy this into this. The root of this tilt resources directory, like I have mine, set up here and then just set up the if you're using menu. All you have to do is make sure the url is correct.

C

uh Did I miss anything so this customization, this customize file, will set up the credentials exactly how the leader expects it at least the way it works now and all you have to do is again drop that cloud file. It needs to be named clouds because reasons, and so you just drop that folder under the tilt resources. I mean the file under the tilt resources folder, make sure you get your credentials whatever the provider you're using and then yes, so this is. This is what it looks like.

C

ah Let me bring this up. I'm gonna run to uh show you how what it looks like so here on my right, I have my terminal is on in that directory or my brain on the branch for that pr. So I just do because those files are set up. I do tilt up and and that's it so it's going to give me the url for where things are going to load. You see the plugins already loading and it's building I have.

C

I have it set up to load everything just to demonstrate, so the plugin binaries are are all built already. I mean the okay, so disclaimer. I had already run this, so I think tilt my health. Let me see, let me think through this. I don't know if there was something that already was pre-loaded in the cluster, so it didn't build from scratch. Building from scratch would be like a little bit longer, um so the letter was already installed in my cluster.

C

Yes, so the second time you run it, you will be this quick and there's some warning here. Just saying, if I use the kind.

C

Local, all right, you don't have to okay. So one warning here is uh saying: oh, I don't have customize installed, but they use uh the control customize. So that still works and is also saying if you, if you set up a local registry for kinds, this will be even faster and I haven't done that yet.

C

But you can. Those are just warnings if you do have a build failure, see this one here right at the bottom, where my cursor is, if you so there's an error. If you have a build error, not only like an error message, but a build error, it will show up here now, so you have the logs tab and if you go to the sorry, the either the rustic deployments or the valero deployments, if you have any logs, I mean the photo deployment will always have this running log.

C

It will show up here and, as you make code changes, I'm going to put this on the right bring down my editor, I'm going to close this.

C

I want to show a code change, so you can see how quick uh the code change gets um gets reflected. So one of the uh directories we I we are watching with this configuration- is the package directory. So we are watching the whole thing. So I can go here and I know the backup storage location controller runs continuously, so I can say- and I also have configured the the valero deployments.

C

Sorry before I show the code change. Let me just show something that I really like too. So the the little deployment that I have here. If I make any change here, it will also rebuild restart the content, rebuild the binary and restart the container, so it will reflect immediately, and so let's say, but so now I have the deployment set to the bug, and I add I just want to how do I make this show up in that log? That's running with so many things.

C

There are so many messages there.

C

Maybe this will be obvious.

C

So I'm gonna save the first thing that happens. Is that see the valero manager here on the right? That's the actual, valid binary that is being built. It's like it's just a gold builds of the source code. It's been done, that's happening locally, so now the valero deployments is is restarting.

C

So this means the the binary was pushed to the container and the container is only restarting is not being remade and that's what makes it super fast because you don't have to rebuild the container itself, you're, not pulling images from anywhere, and so let's look for this.

C

C

C

It hasn't finished building yet.

C

C

Sometimes it's faster than other.

C

Times but you you get to see everything that tilt is trying to do.

C

And here's my the message that I added say so they had already yeah. That was and that message is going to pop up again because this is going to keep looping. So that's it ta-da questions and then there's the tilt file. I try to document, oh by the way um I copied the setup that cluster api used. uh Indeed, goldstein set this up for cluster api and I copied and tweaked to make it fit for valero, and I try to document things that were not super obvious.

C

So basically, this part is just setting up the velo deployment and then this is setting up any plugin that you have configured to use uh with with the tilt setup, and with that in mind, I've just remembered one thing that you we need to use any plugin, so I have prs for each of the plugins, the plugins that we support with this file here tiltprovider.json.

C

So you have to give it a name, because this is the name that when you say oh, let me show you some sorry slowing down all right. So you have this set up and all you do. Is you say what? If you weren't running with tilt, what image would you use for this plugin? So you don't have having it to give it a version. You just give the image repo a name.

C

You point to uh what directory you want to reload and you also point to where the go main the main go: dot go file resides in now cases. It coincides to be the same directory, but with other plugins. You know they might have that somewhere else, and you give it a name, and this name is the name that you are specified here in your personal.

C

Personal personal personal setup file, which is under the tilt resources folder, you have distilled settings which is your personal settings, and here, when you say enable providers, um you use that name, the name that was used in the config for the provider, and these are the providers that are going to show up here. So you have to do two things to enable the provider you want and also point to where the repo is so it so it can even find that config file.

C

Now you can specify all the repos here and if you just delete this, these things.

C

So I'm just going to show you.

C

C

Okay, my vim just stopped working and so.

C

Gosh demo, nothing is working, okay, there's a properly formatted json, so immediately those binaries are got unlisted and all I have is the aws binary here, so the tilt file. I just try to document uh things that weren't obvious should be pretty easy to understand. So the two main things are, the this local resource function is. Is this is a tilt function?

C

It builds. The binary in this docker build is what is also a tilt function, because uh there's some python code here is people may recognize.

C

So this function is what takes that binary puts in the container and restarts the container so for the plugins, just uh as a heads up the plugins, because they go inside the init container, it would be awesome if we had a way to just put the plug the the binary in the little container that's already running in that deployment, but unfortunately it will trigger a restart of the valero container, but still it's relatively fast, and if you look at.

C

At any rates, if you look at where is my thingy, I was just adding the comments to nolan's question on my pr. If you and I took a screenshot of like the make container, that takes almost five minutes to run, and this uh you know might take 14 seconds depending on what it is you're doing.

C

C

F

I think you hinted at this, but uh the tilt file is that basically a python file that gets invoked like when I was when I was reading it. It looks like it's a python file that has some well-defined functions like docker builds and um some you define yourself.

C

um I'm uh concerned to say the wrong thing, because I didn't okay.

D

C

So I don't know if, if so, this is there's, this is uh these functions. Here are definitely python functions and.

D

I'm saying this because.

C

I don't know if you know, because I'm not a python, I'm not super it.

F

Looks a lot like python, I'm guessing it's a button and if you uh go to line one very top, I I bet it's just python and they just interpret it. I don't know how tilt runs, but it probably says invoke the python interpreter.

C

Yeah, so so what I don't know that this is definitely python codes. What I don't know is, if you can use other languages here or it has to be python, I'd.

B

F

C

Python, it might yeah yeah. I just said.

F

It's like um vagrant, um the vagrant file is like a ruby file.

F

um I was most mostly curious because, like the thing I find interesting like if you look at line 77, uh if you scroll up a little bit from there, uh yeah so line 74 to line 80 they're, doing things like making chunks of a docker file and like pinning them together, which is interesting.

C

It's it's. What makes it faster because you are excluding things that for development purposes, you don't really gonna care about.

E

C

And and then of course, if there are things that we would care about, and I just didn't realize we can always add here, but I left a comment on the pr saying. Basically, I might have been misunderstanding what you were trying to say on that pr, but we will circle on that later.

F

Yeah I saw that.

F

And the the thing with valero plugins yeah since you're modifying the container, whether it's in a container or I'm sorry since you're modifying the deployment, whether it's in the net container or the container itself, that's going to cause it to restart.

C

Yeah yeah, it has to happen um yeah.

F

That's just a cube thing.

C

B

B

There we go and now we're back.

B

Sorry about that folks, uh the host unfortunately quit so now, I'm the host again so we'll get started again. um Yeah you were talking about tilt you're. Having a discussion. Sorry go ahead.

C

Yeah, so if anybody has questions about that's the setup that I was showing any more questions.

C

Oh, by the way you can just switch contacts to an entirely different cluster like could be aws and run tilt up in boom. You got the whole velour setup installed.

F

Does it does it use like cube, cube, ctl's context.

C

Yeah uh yeah yeah, whatever your config, is pointing to as the context. So it doesn't.

E

Care, it's really cool thanks carly's here. I think the thing that I'm probably most excited is maybe not about like the speed of it, but one thing that I tend to have issues with is sometimes I'll make changes and then I'll rerun things, but maybe I haven't rebuilt. It.

C

And then I'm like, why are we doing this so many times.

E

C

E

Having that just all automatically just happening behind the scenes is really cool, so thanks.

C

For painting all the time together, I've been relying on accent a lot for like eighty percent of the reason why I rely on octane is just to go to the velour deployment and see what image is there and then I'm like? Oh yeah, okay, so that's not the image version that I mix that I should be using because it's in building or it's insane.

F

Yeah, I think this is a nice middle ground between running it completely locally um and having it in cluster, and it includes something that I think web development folks have had for a while, which is there's a a depth line, which I think, if I'm understanding what it's doing correctly, it's watching for changed files yeah and if those files change, then it will do the build and shove the executable up, which means you don't have to think about it uh tilt. Does it yeah, which.

C

F

C

And you can make changes to other things that will not, of course, if it's not the things that are not being washed, not gonna are built, but things that you wanna you want to trigger build definitely will be included there and it will happen that way but yeah, but I think for for me the speed is, uh is huge and not having to like. If I do have to run things in as in the cluster as opposed to locally, you know you have to make container which takes five minutes.

C

Then you have to set the imp set the image, so the image can go to the cluster and be updated, and that's like. Okay and now I'm like checking twitter.

C

So anyway, and how easy it is to onboard people too so you're, just like hey drop, these files change two things, and you have the letter up and running in a development environment that you can start changing. Pokemon.

C

Yeah, it's a huge improvement.

C

Anyway, please review my pr, because I want to merge this and start using it I'll do that this afternoon.

B

This is awesome. Thank you, carlisa sure uh right do we have any other discussion topics anyone want to bring up for today.

B

All right, then, we'll dive into communities. Shout outs now yeah, so.

F

This is, um hopefully I got everything um I went through and got got some of these manually. um So um first, uh we've got a design dock from scott ceo over at red hat, for updating our restore item, action plug-in interface to have a weight functionality.

F

Basically, we have these additional items that can uh be returned by restorative actions and there are situations where it's useful to wait for those additional items to be restored. Before continuing with the current item, um I won't dive too deeply into that, but um that got merged uh almost two weeks now two weeks ago. Now, so thank you, scott for your patience and uh that proposal, because there's stuff we can use that core valero um next um pranav, also from red hat, um proposed a design for reporting restore progress.

F

um This is for the kubernetes objects, not necessarily uh volumes, but it essentially introduces two passes on the restore, so it will say here's a list of the objects we expect to restore and here's where we're at and um yeah. That will be super awesome so again, also a couple couple weeks ago.

F

Next we have um a collaboration. uh Two two folks. Let's put this in, uh there was a bug um that the scheduler, when creating a backup, didn't pass annotations down to the backup correctly.

F

um So this um this change fixes that and the schedule schedules will now pass any annotations applied to them onto the backups that they create. So thank you to funkycode and michael for making this.

F

um This is this next one is pretty small, but still it's very, very helpful um from tick kim. I think um I apologize- I probably pronounced that incorrectly, um but basically a typo correction on a an internal function name. um So thank you very much for cleaning that up.

F

um We have some um documentation instructions uh from madav um that basically allow let people know what they need to do to clone the repo for uh downloading examples. Thank you very much. Budoff.

F

This next one- ah yes, so this is from genting.

F

um This is part of um some cli reworks that we were doing- um and this is this- is kind of one that probably should have been around for a little while, but um genting added the ability to delete a backup, storage location via a valero command, rather than doing a cube, ctl delete, so you can do it via valero, backup, location, delete or valero delete, backup location.

F

um So thank you very much uh genting for adding this feature really appreciate it um next, uh because my script wasn't updated to add dave as a contributor, uh dave and ashish have been working on getting our end-to-end test framework integrated. um So I'll still call this out because it's it's really good progress. um Ashish has a has a pr for getting actual tests out out there, but this is the framework for getting it spun up and run against clusters of any cloud provider. Basically,.

D

Yep um sorry, I was running a little late from one previous meeting. uh I was hoping to demo the end-to-end test uh or demo, my pr today, uh I'm sure, but we may. uh I will do that if we have time at the end, I don't want to hijack the rest of the agenda yeah. We certainly could oh.

C

um With that in mind, I have to look. I have to dig through the documentation to see when I ran when I run go test after I emerge the tests. It fails so.

C

D

I can I can look at that offline.

C

D

Think there are two more contributor shout outs: uh nolan, do you want to run through them.

F

Sure yeah, we can finish those up. um These two are from the helm chart. um So I have not looked at these as closely um the first one is from the moon. uh Let me open this up.

F

Looks like there's some dns issues.

F

So um the dns policy was not being passed down correctly, looks like, in both cases.

F

See yeah, so um this fixes it so that a dns policy can be passed down.

F

The moons actually uh relies on uh jared dollar jared allards, um so these changes allow for a dns policy for the cluster to be passed down into both the rustic daemon set and the valero deployment.

F

So thank you both for uh yeah. These two are dependent. It looks like um so. Thank you both for uh contributing this um setting. This dns policy is, is actually pretty nice. So thank you very much for your contributions.

B

Awesome um so ashish uh do you want to demo your your work.

D

Awesome, so let me just share my screen. First.

D

Now, why is the share button and leave button like so closely located? So while I was waiting, I just started this e2e tests that I have running um so what I have so far. Actually let me share my uh pr, because that is more easy to talk.

D

D

D

D

So this is the end-to-end tests that I started more into interests that I started, adding uh basically some some refactoring and how we want to install valero. So previously we we were expecting or, as part of the previous uh foundation work that we had that dave had put up.

D

We were expecting weather to be installed in the cluster before we are able to run tests, but the change that I have so far: installs valero before running the test and cleans up valero afterwards. uh This will allow us to like uh customize uh valero installations for the test that we want to run.

D

So I'm still continuing to use the kibishi workload that dave had set up initially to backup and store and still using rustic, uh not made progress into the wall into using the volume snapshot plugins, but that switch should be pretty straightforward. I expect famous last words, so that's uh that's what I have so far accepting uh this. This is how uh somebody would run the test with these configuration files or these configs uh paths to credentials file and the bucket and the prefix and stuff uh the bsl and the vsl config.

D

So this is the part which installs valero before every test, and this is the part that cleans up valero and here in this case, in this test, I'm I happen to be running. The kibishi tests in aws, using restic using aws, is a backup, storage, location, interesting. So the way I have.

D

Kind of organized the tests is these are this is a common test function and this is a file that is specific to aws tests. So all aws, I'm expecting all aws tests uh to go into this file, and this kiwish run kbc test happens to be the first or one of the tests that we want to run similarly to kind of draw comparison or to compare and show. This is a similar test for azure again uh running the same tests, but installing uh valero for azure, using the cloud provider related configuration and credentials.

D

So that is what I have is going so far. uh This is pr three one, one one. So take a look uh to give you on a demo of uh how this thing works. I have a kind cluster here.

D

And there's something on my laptop and I don't have valero installed.

D

D

Actually, these don't matter, but I can clean them up. My install only deletes the name space, but it should also delete the crds.

D

Okay, now I also shouldn't have any crds uh now. This is how I invoke the uh I have an environment variable setting all of these, so that I don't disclose my subscription account and all the other secrets. um So I am in the e2e test uh directory yep. Are you meaning to share your terminal.

F

That's all right.

D

D

So uh this is the command that I ran earlier and I realized the crds were still hanging around. So I deleted them and here and then now I don't have any valero crds. So that's all you've missed so far from my terminal session. So now I am in the e2 test, e2e directory uh and I'm running this make target with all of these configurations.

D

I have environment variables set up for all of these and that way I don't disclose secrets on on my demo and it is using these credentials or credentials in this file uh and I run make run what this is going to do is it is going to build the tests and then first it's going to fetch gingko and then it is going to run oops.

D

I just did disclose my storage account then subscription id did I didn't I. uh uh I figured how to clean that up later on. So what this did is it is running this test um and then uh it is doing claw.

D

It is installing the crds, the regular weather installation and all that, and then it is waiting for the installation to be ready and once it is ready, it is going to create this workload, which is again like the um kibishi test or using the distributed data generator or the kbc and setup that we have or the dave setup, and it is going to generate some data by running this script. uh These magic numbers or what seems as magic numbers is how much data it want.

D

We want kbc to jb generating uh and at what hierarchy so that could that could do some polishing. So now it generated the data. Now it is trying to take a backup of whether or what and the other interesting bit here is it uses this local install a local binary of valero to create backup and to create restores.

D

Now it did run back up. It did confirm that the backup completed successfully and then it to simulate a disaster. It removes this workload namespace and now it is restoring. Now, once the restore is complete, it will verify that it will verify that the.

D

Data that was generated before is restored correctly, while we wait for this restore to finish.

D

While we wait for this to happen, ok now it is running verify which is again like the same magic numbers. um While we wait for this to like complete the success complete successfully, are there any questions that I can.

D

D

And the other interesting bit here is when I ran this make, I said: gingko focus azure, which meant it skipped uh certain tests, it says, will run only one out of the three specs, so there was the aws related tests and also another dummy suite that I had it skipped over, because I had azure focus focus on azure. uh That is all coming from. That is filtering based on the these labels. uh These these are not. This is not any syntax. Specifically, this is just plain strings.

D

I added these box brackets around them to make them seem like labels.

D

That way, it is easier to discover and filter.

E

I think that ginkgo focus can use uh regular expressions as well. If I remember correctly, it's the same um thing that we used uh in sauna boy for targeting specific conformance tests, yep.

D

So those are regular expressions that you can use yeah, so it ran the weather over. If I sorry the kibushi verify stuff and it confirmed that the workload was backed up and restored successfully and everything passed um going to switch back to my pr real quick to show the code that I was running so if I say, run, run test. So this is the run kbc test method that it does um I'm using like timeouts um to control. So what this test does is creates the namespace installs kibishi install kbc goes to the.

D

Install sheet goes to the goes to this repo to fetch the yaml for the specific platform. This is. This is the uh platform where the cluster is that way. It knows which storage class to use and all that. So, even though I'm testing the aws plugin, I pass in kind here so that it gets like a storage class specific to my uh to the platform on which the cluster is running. So it installs kibishi generates data.

D

It backs up the namespace and simulates a disaster by deleting this namespace deleting this namespace, and then it will restore that namespace from the same backup that we took here and once it once the restore is complete. uh It will verify this and later on, it will as part of cleanup. It will also delete the namespace that it created. So this way on the same cluster, I can run aws related tests and I can also run azure related tests, so I just ran the azure related tests.

D

Now I'm going to go where to see how I can run the aws related tests.

D

More azure, more.

D

D

Yep so now, on the same cluster, I did not reinstall the cluster on the same cluster, I'm going to run aws related tests, which is going to run the same test, but on using aws as the backup storage, location.

D

So it is waiting for the valero demon say whether or deployment to be ready and then the resting demon set, and then the good thing about aws is one. I did not disclose my any of my secrets running the aws tests.

C

Did you want to be showing your terminal now again? I know.

D

That's one way not to disclose secrets, so this is uh this is how I ran the uh aws specific tests. Yeah ginkgo focus aws using these credentials, and this is the bucket that I wanted to store the backups in.

D

Yeah now it installed valero. Now it created the uh workload that we want to backup and store. Now it is generating some data. It is going to run the same tests, but on using aws as the object, storage.

D

D

As next steps for this pr uh once this, I will probably be taking this out of work in progress uh sometime today. uh After uh finishing a couple of uh after tidying up a couple of things that I had in mind and then uh work on more tests once that gets smudged, I think I cut you off hold on.

F

um I had a question about um the setup with kibishi and specifically with kibishi: um is there a um a set of methods for like before and after uh tests to set up stuff so that you don't have to install kibishi in each tests in each test? So I know some.

F

Some test frameworks have like uh uh the functions to have different names, but they have a before each test run this and after each test run. This.

D

Yeah, so so here kibishi the way we are using kibushi is: we have a sample app using kibishi, uh which has a specific uh environment that it is expecting so the way kbc works is it it needs to be. It needs to have certain scripts and uh commands or tools available on the container for it to run to verify data uh it just does uh so.

D

For that reason, uh this is uh the workload and kbc is like totally tightly coupled right now, so we will want to decouple that to make the data verification part like a separate tool uh from the actual workload.

D

So today they are like super tightly coupled and that's why the installation of kibishi is tied into off uh tied into the deployment of the workload.

D

Yeah so yeah this, this test ran on the same cluster using aws as the backup storage location and it cleaned up everything that it created. So this will kind of give us like a way to spin up valero installations, take them down and run for for each test.

D

That's uh that's how we handle no learn to answer to get back to your previous question. It is using the before and after uh methods uh that we handle installation and cleanup of valero.

D

So that's the demo that I had.

B

This is awesome. Thank you, yoshish, we're almost that time as well. um So I wanna thank everyone for joining today and uh yeah. Please try out the uh the testing.

B

Please try out the uh the tilt that kalisia uh built out and uh with that see you all on slack and uh have a great rest of the week. Everyone see you next.

B