OpenStack Crowbar, 11 Jun 2014

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: OpenCrowbar Design Meeting 2014 06 11

Description

http://pad.opencrowbar.org/p/community-design-meeting-2014-06-11

Discuss hardware configuration design and approach.

A

Ok call recording started, oh and someone just joined good.

B

Like Mike morning, everybody and welcome to the jun 11, that's the curb park community meeting. Our topic today is like design cycle and we queued up a whole bunch topic. Talk about our design meeting like I, find the right users passwords.

B

You you, oh the way well, with the page load I have been driving. Crowbar is just enough orchestration. Pretty much using 100 word great someone things. Now we can start. There are things that we were supposed to talk about. With this, the crowbar elevator pitch and I've been describing a crowbar using the phrase just enough for castration two provisions, physical infrastructure and that get a very positive response.

B

People need to understand my pants.

B

In ready, state also continues to be a very effective tool to say crowbar get through ready state do other stuff. On top of it, that's one thing: I get to do, get to do quite a bit of on these trips.

C

Yeah I kind of like that I might move ready straight up in that phrase, just enough for castration to revision and maintain ready state for any brand of yeah. Just kind of put that first.

B

Oh I didn't even put ready state into this.

C

Further down right, he had ready safe for any brand of physical in, for sharp sure, I just move it up got.

D

B

Oh, you know what and it actually should be an open ops framework.

B

B

Yeah I've been talking to people with the claw foundry summit. If we were actually going to have ready state, they kick up top boundary that would be cool or even a lab Bosh to interface to the crowbar ApS.

B

That would also be interesting.

D

B

Other comments on this see I'm going to read it just this short, an open ops framework with just enough orchestration for vision to maintain ready state for any brand of physical infrastructure, with raid bios and networking fully configured for scale out application deployments like OpenStack 7z along them. It works alongside tools like chef and puppet to integrate standard operational services into an integrated system.

B

C

You kind of works for me, it is a long sentence, might be tempted to stop after raid bios and networking fully considered make that one sentence, because the trick, the elevator pitches, you always have to have a short version of the elevator pitch yeah and then the second sentence, would you know yeah crowbar is used for scale out application deployments like list of examples and works modified.

B

Wait and instead system.

B

C

Works side tools like chef and puppet integrate standard operation, I.

B

Was the work, so I.

C

Like that, it works alongside tools like chef and puppet yeah, he's question which one we want to put first I do think we want to mention the scale out deployments, because that's really the canonical use case.

B

B

Actually, just take out the word.

B

They got the permit application.

B

D

B

That are cool-looking. We should keep tweaking it.

D

C

Got too many integrates these integrated in there, but yeah I like the per sentence that really summarizes your right to integrate standard operational services.

B

The thing the thing that I was trying to get and went, we went back and forth dr. Milton, steady state.

B

H1 is now was that it was very important to describe it to the system school, not as a thought striving to get the word system in there.

C

C

I think that's pretty good for a current cut. That would be nice to see that on the homepage of the website,.

B

This I'm, not hotel internet is.

D

B

41 all right back to the actual meeting.

B

It looks like about this. We did yay.

D

B

Or the other thing I was interested doing we've actually talking about the spec wheat. If we have time to do that.

B

D

The victors, have you: did you look at this design document at all, good, no.

D

B

What I'd like to do is probably just summarize it a little bit review it briefly and I'm looking so some of this was received making this as he learned what what open crowbar did and more things I liked was that he actually called out this using these commands would be the way to do the initial discovery.

B

Out-Of-Band 4ws man.

B

And this is all it is all about going after the ws man. No more calls use for that. So these free down down down deep, including sample, workflows and stuff like that I think we need to take a step higher, for this is the thing I. The thing I don't know is: would we design or hardware interface? It seems like we would write.

D

D

A

Know it's the sort of thing I would want to do. Some exploratory coating around.

B

What one approach would you take to do out of fan configuration? Actually, let's start, we have limitless documented design goal right so hit a design goal, which was a single.

B

It was a parent super class. Well.

A

No they're all jigs anything that does something to a node is a jig.

D

B

Chef or puppy, you can't use the chef for puppet jigs because those actually operate differently, where the essay should see Union you jig right, I would.

A

Probably need a few new jig.

B

Would you create a super Jeep, they would have some shared capable.

A

Not stop Rob stop crack. This is a thing of this is jacob super Jake I.

C

D

C

Know in a perfect world the the notion, the the notion of having a single jig to do all this stuff would be nice, but um yeah I.

A

C

Ends up a perfect world yeah exactly live in the real world. We live in the real world and I I, don't see a clean abstraction, because every hardware vendor is different. um It you know and I'm, not writing the code. It could be that there there would end up being an ipm I, Jake more so than a vendor. Specific Judd was following up on this. He sent an email to the list he he was looking for.

C

You know any references on what's out there for various bios versions, for different hard words and tools available and I've never been able to find anything. So I think it plays into this conversation a little.

B

It is there a generic base that that we were that was sort of wherever g was going and so I feel like. We should try and answer that question a little bit now.

A

There's no generic base, there's just an interface, but they all adhere to okay.

D

C

Yeah I mean I agree, I mean I think, there's an interface that they all share at the lower levels. There are some api's they're all calling, but so.

B

What does that interface? Look like I have.

A

B

D

B

No just just brainstorm, no.

A

No, let's not brainstorm, let's not do anything along those lines until we actually start coding, it yeah.

C

One of the things I'm a little concerned about I'm, not sure if this is a recurring pattern, but in the past some of the configurations for say an OpenStack were different, two configurations for a Hadoop per SF when you get down at this level and I'm trying to understand that's, why I think these are jigs, because I think there will have to be bar clamps that do the hardware settings for those individual applications whatever they are. Does that make sense.

B

It makes perfect sense well.

A

The architecture for that that I added in mind the whole time you know, is that some roles get notified whenever a new whenever a node is added or it has changed, and one of the many things that the roles can do whenever a notice added or changed is classified as to how desirable it is, and we would want to build a mechanism that would whatever you say. I want this node to be a Hadoop node to make sure that that configuration or make sure that the desired configuration that's attached to.

A

That role is reflected in the in the desirability of the role and for thee and for whatever Hardware magaling jig. We wind up using on the node to be able to go through and be able to pick that configuration up when it comes time to run.

C

That makes sense, um dance.

A

That always are like a dedicated that a bag of you know just a standard setting. We would just study for how the notion look. We would be up to that I do for the openstack or the whatever role to have to you know give their exact configuration of the hardware jigs whenever the note is being processed yeah.

C

That that makes a lot of sense to me. That also starts getting us towards one of the problems we deal with in Hadoop probably also say set for OpenStack, but what's the right way to handle a handle of event like a disk edition or a disc replacement.

D

C

Know and if the rolls are getting notified when something changes, that's a good way for us to trigger the code to handle those changes.

C

That's not really hardware stuff, but it falls. It falls into that same category.

D

B

Interesting question because we wouldn't necessarily get notified automatically you'd have to reapply oh yeah.

C

And that's us I'll get a separate thread going on that, because I actually think that's better. Today we don't get notified, but part of the one of the things about a disc replacement is there's a manual intervention. Anyhow, somebody has to physically replace the drive, and when that happens, there are a sequence of events that need to take place. So as long as there's a hook, you know when we make a change on the crowbar side if the roles get notified.

C

That gives us hooks to start triggering the appropriate process is to do whatever is needed for that drive.

B

No, that's definitely doable.

C

But I'm a little further up the stack, I'm formatting file systems. We need to make sure the controllers are set up, but in some cases, controllers might need to be tweet.

C

So that it overlap a little bit with the with the raid concept.

A

D

The cyst this is.

B

Depending on the flow of roles, which is one of things I want to talk about, it would be the logic would be able to come in and say: alright I've noticed you're, not you don't match the desired configuration and then make the required changes right. That's what the network bar clamp does that's pretty common for what we've done and it's what all the chef's logic does that we have.

D

B

The one of the things that, in victory, you know I love, naming things.

D

B

Of the things that, on my mind about this is what do we call the rolls at least the no opera that we're trying to target and do we split and corporals for raid bios or those orders? I don't know.

B

Have you thought about that?.

A

B

So let me let me back up a step because I was I'm, always bugged, until I've resolved this. Do we call it hardware? Do we call it metal? Do we call it Rajeev had how in there as a thing.

A

What for the names of the role.

B

Name so the hardware bar clamp, which seems very clear to me it's long, but who cares if it's long.

B

And that's equally covering of switches. So let's say one of things that, for me, my thought is: is the hardener bar clamp ultimately include switch configuration if we're going to do it, or would that end up being a network workload that that's.

A

A different thing: if.

B

We're going to wind up doing.

A

Configurations we're going to represent, which is his nose.

B

Right, so is this: should we call this workload, the node we're a metal mode or so like that or the bear node I.

A

Still think that I'm want to worry about names right now. Why.

B

I know but you're going to have to name it at some point and then we're going to we're going to have we're going to be stuck with whatever names we pick so cbeebies.

C

B

That's does with what I what I don't mind doing is having a discussion of some of the options and then, when we get to it, do it right now it's hardware and that's the way the bar clamp is named and I would, unless there's a real issue with it. I would stick to that.

B

The other options to call it server.

C

Now, I don't want to call it server I. His server is just so overloaded in so many contexts. Yeah.

A

Don't call it server, please yeah. You.

C

B

Other alternative, I would consider, is metal just because that's you know, the concept of bare metal is.

B

It's so widely used. Looking back to our short interruption.

C

Maybe we could call it the mental abstraction layer.

D

C

Of the hardware abstraction layer.

B

Like she works very well for me, really nice 100 birthday.

B

B

Could also call the physical think of Jane Fonda.

B

Others, if there's a part of me that like metal and that, actually you know what Victor, just from the simple aspect of what is searchable and easy to find inside of the code base, the metal might actually be a very good choice, because I don't think we used the term metal anywhere specifically. Do we.

A

B

Cool, so that makes sense and I think it's easy. It's capable enough. It fits in with the tool theme three week.

B

So from that perspective, if we were doing a set of roles to turn the end it with a no-op roll, something like metal ready would probably be the likely no op candidates that matches what we've done with other with other no-op.

D

B

Would that be radicchio logical? So if I was actually I like ready, is nice because it lines up with the readystate words, we've been using and so metal? If some, if you said oh I've, got I've got to a metal, many state people would understand what that meant pretty intuitively um and that differentiate she from being a less ready or even hardware.

D

B

Nuts perspective.

D

B

Any issues with the direction this is going.

C

Now it makes sense to me Victor.

D

You cool with it.

A

B

Let me know if you get if you get in there and or whoever gets, and it doesn't make sense I wish we can change it, I I'm tempted to leave. The thing is probably confusing to leave the glowed repo called hardware and everything inside of its called metal, but maybe that doesn't doesn't matter that much.

B

Oh probably doesn't matter because the jig and the rolls are separate from the repo name aren't.

D

B

Pretty pretty distinctly separate yeah.

A

B

So if we in the hardware repo we might end up, including figurations or something like that, just added into that repo without that being any complex office, yes,.

B

There's watching the time.

B

B

And then, from a design perspective on some of this.

D

B

The victor, how right now we use oh hi to inventory the the system. This calls out. We could do out of band and use WS man to enumerate systems and data in here.

A

D

A

It depends on exact on how much stuff would you get from the enumeration.

D

B

And it can exist side by side with the Ojai data I. Don't that ends up not being a problem. It would be interesting to well.

A

What we have to do is to whatever data we get. We, we need to put it into a format that we can fight them through, with that with using a trib right.

B

I was just I was just going to add ribs right, so we have the.

B

How are we going to know the IP address.

A

The IP address of what.

B

The ws main interface.

A

We want to begin.

D

With the dish shoot you.

B

So from an architectural perspective, do we assign it an IP address out of DHCP, assuming that sometimes our network there on the admin Network they're going to get it.

A

We would, if it's configured to do you dhcp, okay, which is there's no guarantee that it will be.

A

D

B

So today we can figure the ws man interface from ipmi. Well,.

A

So W Spanish base is the ipmi interface.

B

How do we say have a same wouldn't.

A

A

With ipmi tool from sledgehammer is the answer you're looking for okay.

A

Yes, I wouldn't go gung-ho long, you're moving at me or sledgehammer. It.

D

A

Sakra mental alarm, acidic that still our method of configuring ipmi and therefore the ws manager face.

B

No fun so once so once that interface is set, we can go out of, and so you still go in band.

A

Well, we go out of bands for things that it's useful to go out of, and florin and dance for things that useful together in van for sure,.

B

Not argue, but does that mean that we don't need the? I guess I'm trying to eliminate the proxy and we need to set up the proxy on the cover server. Maybe I shouldn't be so you won't be so father the router. So we.

A

Won't need the router anymore. If we're doing it right you it's. The only reason the router exists was because our previous architecture, it didn't, have a WS man jig, so it was doing all of the actual works from the chef recipes that were running on the so we're running on the know that we were trying to configure.

A

That's the only reason the proxy existed for the, not the proxy, but the semester rating rule.

A

B

So from an actual story, work perspective you.

B

There's actually a story for set. The WS means that the.

B

You have to do that as part of them.

D

A

So it's a shame just on the call I wanted app about the colo spacek eat, acquired.

B

He'll, probably you might join us after the cloud stand up which unfortunately flakes.

B

I know I, know he's also working to ask ask for hardware donations to see if people in the community wanna, which we have space to see if people community want to help donate donate gear, to get that space running well.

A

I mean the first piece of gear that we would need would be: uh oh we're going to have KTM access, which is good. We need some sort of just local know to run built to you know, to build and run an admin note from right.

D

B

As soon as soon as we have a spec for that I'll get back here and there I.

A

Mean the note that we can write an admin note could be any old box, it doesn't have a video still hardware or a piece of it could be pretty much anything like an old crappy laptop okay I mean we're going to be doing the doctor or the kvm based admin. Do it anyways.

B

True, it doesn't even need an ID rack on it. Does it nope.

C

B

But rated draws at least okay, let's just let's figure out what one of them and somebody pronounced.

A

Hanging up taking up like a t320 district pulling around at home, but I think.

C

A

Actor tolo that he did do I wanna.

B

I want to do the same thing so I'm ready to if we have aspect that I can buy, I will order one or two one yeah.

A

Researching online, it looks like I'll put the h3 10 at all cops because it has sucky right performance.

B

It has web performance.

A

Sucky right performance, Oh.

D

B

What's the dress, huh what what is it a drive issue? What's? Why would enough bad right performance? No cash? Oh ok! Now we call it.

B

You that was I was actually looking in here to help build the spec. You.

B

P, so we did, we move far enough, that I think we can actually start building. We've actually do a little bit of work on towards this jig at least build. The no op rolls essentially create some other other roles. You know we at least need WS my credentials.

C

I yeah I mean I think we could start doing that and we could probably start getting a feel for what that consistent interface starts. Looking like.

C

One comment I had on on raid setup. Is we have a history of calling this raid setup, but in a lot of cases it's not just rate. You know it's, this controller, I, don't know if we want to change that, but you know I mean in most lizard. Observers will potentially have a raid for the operating system, but in general we're dealing with plain hbas.

D

C

Know I think we got in the habit of calling a trade because we potentially started configuring raid controllers, but it really just disk controllers.

D

B

Why don't we? Why don't? We then think that's a really good thing to put in here as a change alright. So, instead of calling it raid we're calling a disc.

B

And does this stay by us, which is a switch to compute? From that perspective,.

A

Well, no, it doesn't switch chicken at all because.

C

We're going to we're.

D

Going to consider.

A

It significant Tobias differently, depending on whether we're aiming towards an I/o intensive, are no that should be I on to the broken window. That should be creative.

C

Yeah I think I'm, okay well by us I mean you know it's eventually going to be ipmi, but everybody still calls her by us so stuck with us forever. Oh.

A

Oh I know I feel an enviable see different things. Yeah.

C

But I think in terms of calling it would there be a metal, ipmi, comfort, yeah.

C

Saying I PMA I met you fee, but brainwash.

A

Cookie is also a biopsy.

C

B

Know I, would you can be different, would UVB a different role. I guess is the question.

A

So I don't see. Why would we different role at this point? It's just uh it's just a it's, just a configuration fit that you would flip in the bias. So what mode to boot into as far as how that affects our managing system.

D

A

Our series gear, it doesn't and I, don't know 40 series here because yeah we only have to a couple people see series and it would do you to begin with and we never set them up to do that. Yeah.

C

But I guess thinking out loud here again: I mean whether it's, whether it's a youth, you by us or traditional, buy offs kinda, really doesn't matter I. Think most people call it yuffie bias, although technically it's not a bios but I'm getting off on a tangent ah I I do like the the disk config as opposed to raid config I. Think that's better better! I'm.

B

In favor doesn't make sense to me.

B

They were actually going to call it disk configuration no respect.

B

Though us man can handle disk and bios eggs go yes,.

C

C

D

C

Not sure some of the controllers, like the lsi controllers, have their own BIOS settings that are only accessible in traditional bios mode. I, don't know if WS man can get to those.

A

D

A

Perfectly managed vehicle, yes, man, the bios has to be heard from where has to be written, to allow the expansion to manage it as well.

B

Can I control.

A

Ridden that way, it's not the fall back to their crappy. That.

B

We're going to need other roles or they're more intermediate states that we have to worry about. From this perspective,.

A

We'll burn those bridges when we get there.

B

Thinking about major major levels of this light, I guess we're going to have I p.m. I configured, which is going to be the first sort of basically for the workout like this.

A

Well, the way I see the role split up is that you know things like I p.m. I WS man and the other kind of the hardware management tool. They all have discovery roles that run.

A

Third, such am bridges to discover the just to discover what management capabilities and node has report that back to crowbar, and they also all have configuration roles that their entire purpose in life is to get whatever that hardware is configured so that the crowbar can talk to it, using whatever set of credentials as communication methods that we're going to need to use to talk to that and then, once both those roles have done their thing, then then the jig that does the actual work would be ready to be used and that's kind of what we do for the chef jig right now.

A

If we r you setting up a chef server, we have a role that, when the admin note starts, stands up a chef server and extract a client or extract the client that has admin privileges and injects eyes those credentials into they didn't check those credentials into crowbar. So the chef Jake and you can use those credentials directly to essentially be the admin chef client for everything else that we manage and I would be. I would reuse that pattern for ipmi and four WS man.

A

A

Am I still there or did.

B

I, hang on oh you're you're here, I'm thinking about what you said and that made me think through the if you weren't using WS man, but we're using the direct on hardware d, TK tools or with a in band stuff, you would. You would ultimately have a provides.

B

Chain an alternative provides chain for that right, yeah.

A

And we try to set it up to all the harbor of management jigs, you know had approximately the same. Let's say an API are the same interface right, so the rest of the tools that want to curve the rest of the rest of scrub are that wants to go in in you know, configure whatever, as long as there's something that sound to the nodes that exposed the capability to configure like this curve by the settings. Room, whatever the rest of chrome part, doesn't care how it's done so.

D

A

Lot of gas on to the note it does use this configuration and then you know that gets farmed out to the appropriate management. You actually do work.

B

So what you're saying is we would restore the requested configuration disk raid configuration on the metal ready. That's the only constant role in that perspective. So if you added met already or if you added Hadoop and you had the metal ready to roll in your system, then you could say I want this rating system. I want this configuration.

B

Right because maybe.

D

B

The order of the attributes doesn't matter particularly does it.

A

D

A

D

B

Don't think so because when you, when you pull together the attribute collection, it goes and finds them against their a single namespace, though it's going to pull them up stream from upstream or downstream.

B

First I believe.

A

Yeah, if you're referencing an attribute against the note it doesn't if it doesn't matter what it doesn't care about, what node will hold it from right.

B

Well, it were, or this precedence right or whether it's ahead or behind so the.

A

Character, it tries to look ahead, but it doesn't lead of it care as long as that attribute is somewhere on the node or bound to something that's kind of an owned. It'll pick it up so.

B

In this case, the Hadoop, the Hadoop or open sac or whatever bar clamp, would rolls still want the column bar clamps rolls would would add configuration to the metal ready, no update, because that's the constant things, no matter how we do things and then.

A

Wouldn't even at no no that's complicated, okay.

D

A

Would have to do that at all. You know.

B

Oh, it was close yet there's just set the friggin attribute to your right straight.

A

On that note straight on that Hadoop node well and just let us do its thing and then, whenever the E and then, whenever the you know, whatever node role, that wise up actually doing the work of configuring, that this is called and it looks up the attribute on the note it'll find it on the Hadoop config, no troll or whatever we decide to call it.

A

B

Oh, that's not what I was expecting you today, so we're going to need at ribs for the role itself. That are things like the type of a trip to do. We have.

B

Discs that you have something like disc sets, there's obviously.

A

It here to a common name, and it doesn't matter what they was there and it doesn't matter. What no troll has them.

B

Right, but what what we'd end up with is something like array 1, I'm just.

D

We're going to end up needing some just.

B

I'm, just I'm thinking through what that would look like so we'd have an attribute, probably called something like disk disk, arrays or disk sets and you'd be able to say get array. 1 array, dwarfing I, like you see where I'm typing I.

D

B

So that's what I'm, what I'm considering is Hadoop would then set the attribute to roll when it when it came in- and you know it wouldn't even worry about whether the capabilities are there just set the attribute full disc set array. One is it wants this? This set array to it, wants that and go.

B

Think of how you actually set the raid and bios, maybe I'm thinking too far ahead, but seems like a good time to talk about it.

D

B

Maybe I'm with maybe I'm too far ahead.

D

B

One of the things Mike, one things that you see as a requirement is that we have to be able to have a variety of dis configuration options.

C

Definitely I mean in the hidden space. You know the preference, the the preferences of vanilla HBA, but you know in some cases raid controllers are needed even for the data drives on the operating system side we need, we see requirements for either software raid or hardware RAID, sometimes using the chipset controllers. So there are definitely there's definitely a need for flexibility in the disk. Configs.

B

B

I, we haven't even going to think through how to enumerate this I know.

D

We already have.

B

You've already done to mark on that.

C

Pinterest yeah, the other complexity is not all of that is actually at the metal level. You know especially things like uh if you're using the linux md interface, it's really almost purely software.

C

On the other hand, it's done it. It's effectively done at the operating system level. So once you can put an OS, you work with it at that level.

B

Yeah, there's almost two passes through that: it's not the metal, it's not the metal interface. That point is it's a difference part, it's a different role that has a different dependency chain, so that so we're saying is that there's actually a.

B

There's a crowbar OS installed.

B

State no ops ii for this, and what you're saying is that there is a something jigged. It does a.

A

B

Disc off yep, that's exactly what it would look like or perfect. It would be a provision or disc.

B

Victor, would you do that in the provisioner.

D

B

A

B

Yeah I did it that's fine. What and.

D

B

What I'm saying is really confusing. So if we had a post, curb ROFL troll that handled OS, soft and thought soft rating and disk configuration would that would that be in the provisioner or the deployer bar clamp or none of the above? That has.

A

To be done as part of our less insulation, so.

B

Right, so is the 40.

D

B

D

B

During so, but the ending that the ending role in that chain is Kovar, OS installed, OS is not installed until after that's done.

B

But it's done after in the chain got to be in the chain of some in the provisioner. It's provisioner them.

A

Yeah, it has to be done in the provisioner right.

B

So there's there's an intermediate state in the provision or where it's done installing the OS, and it can do a this provision. Disk configuration yeah.

A

But that's only if we're going to worry about doing this operate.

D

B

I'm, just since Mike brought it up I'm thinking through the flow chain of where to do that, because I think it helps describe some of the flexibility of the system to do hard and soft raid that I'm going back to the agenda for a second because we're within five minutes all right, I actually think. We've done a good job with this I.

C

Think we made good progress on that. We haven't touched ducker yeah.

B

Docker I was hoping to have sat down with you, Mike and and sort of gone through. The docker for death, workflow and smart.

D

B

C

Yeah I I've actually done some of that right now: I'm, I, mom, I'm, actually, rebuilding my dev environment I need to get off that aura. It's just the issue with fedoras it just every time. There's an update, something's breaking and it's okay for playing around with stuff but I'm shifting over to centos for a for my dev env. So I'll have dr. up and running there in a few days. So.

B

What what I was hoping to work with you, and maybe we can just get on the phone and do it as a live session, but I was going to want to do this for the hardware one to just create some stub, no op roles and work through the problems. So I think I might try to do the docker for dev against the ipmi bar, clamp and I know. I know it won't do any IP I configuration, but at least I can no op it and.

D

B

The ipmi configuration hit and then I can pull back. Victor da Cunha is going to fail on on a vm too right. It's got to be on physical harm, physical gear. Do anything Scott what.

C

Did you want to accomplish in terms of using docker for dev what.

D

Thank you. What is it what.

B

I want what I want to be able to do is show how to do the configuration of load. It doesn't have to be chef, but chef is easy of a workload using docker so be able to do the OpenStack deployment or some. You know some of this, but dupe like for you. You've described a whole bunch of Hadoop things that need to be done to prep, an environment.

A

B

What I was hoping to see is build the Hadoop workload, create the no opera run. Those actually, let me take second well literally.

C

Doing we're talking about just doing the deployment and docker to test out the dev env include.

B

Any of you, okay,.

C

B

I think we actually need to walk through the process to say: okay create the top rolls show how those run create the you know, a sample chef run. That does this one thing verify that it works and then show how we build up a deployment in docker and actually think about validate the deployment.

D

B

It ends up being as much here's how you build a new work load bar clamp from scratch. Yeah.

C

This becomes more like a bar clamp tutorial using docker yeah. That's really what hoes close.

D

Yeah, that's it. Ok,.

C

I thought you meant something else, because the doctor supports there and we've been doing this, so this was really. This is just really an example of what what does the workflow look like yeah.

B

Well, we we have, we have the docker piece, but we've never shown a developer, how to take advantage of it to build a new of the OpenStack cookbooks, or anything like that, so that that's really all I was trying to account.

C

The most, let's schedule another meeting to go through that- oh my god, I'm, probably 48 hours away from actually having the new dev environment built, but well.

B

Let's try for friday because I'll be I'll be available and back on friday and lay out some playing time to get loose. Ok,.

C

Sounds good, let's aim for friday and see if we can run through it. Ok.

B

And then what I'd love to see us do is get this sample spec, because I want to buy I'm itching to spend money on hardware.

B

This Victor, if you have a speck in mind the note and know what you want to buy I'll just order the same thing.

A

B

A

B

I was hoping to have Randy to eyeball it just because I think he's always always catches errors in the missions cool all right. Everybody. Thank you for participating at this is think, really helpful. This.

C

One was useful, got some actual design work done.

D

Amazing what you can talk about in an hour, everybody! Thank you very much and I'll see y'all next week. Okay,.

C

D