OpenStack Crowbar, 27 Aug 2014

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: OpenCrowbar Community 2014 08 27

Description

Review progress, OpenStack, Hardware Work, Salt, Chef-Metal
Notes: http://pad.opencrowbar.org/p/community-planning-meeting-2014-08-27

A

A

B

Everybody and welcome to the August twenty seventh up and crow of our meeting, ah not the essence I'm Rob hirschfeld, and it's been a couple weeks since we've had these calls. It's been even longer than that, since I've been able to participate. So we're going to end up being very statice and and open for this and then we'll see where it goes and.

A

Then hopefully,.

B

We're going to start getting back into a regular cadence.

A

B

On these, the regular cadence would be that we talk about work and planning in this.

B

In this call, and then on the other week, we actually talk about design, topics and items that need to be need to be addressed from a design perspective and where things are going so it gives us a chance to sort of go deep technical on the off weeks and then trying to figure out what what the, what the priorities are in the project and moving things forward, and so with that in mind, I've put together this an agenda, so we can sort of guide some of this conversation. Some of these are just hot items.

B

I've talked about I think the first thing on our agenda is most likely going to be review of issues and get because this is where we're going to talk about our burn down.

B

Okay, so this is a bug I found in the in the UI around the flash interfaces which surface from an email threat that we had going on. Victor do you know, have you seen this? Also, you don't usually use UI for this to you.

C

um No I pretty much, never used the UI okay. What.

A

C

Talk about cannot add interface.

B

And I don't have a demo system running basically, there's a goes.

C

Green and use go into the bug. Doesn't the bug in.

B

C

B

I should have put a screen shot in it, but I don't think I did yeah. No sorry I'll make out a screenshot.

C

The oxide no you're late talking about there.

A

B

Me ask: let me see if I actually have a crowbar dev server running.

A

B

I might not nope I, don't.

A

B

I'm going to try and I'm going to launch one in the background. Just give me a moment: ok,.

D

B

Alright, so I'll come back to that, but basically the there's a there's, a screen that shows you the interface mappings per server and that it stores it in an attribute. It's that network interfaces role and, and that wasn't that from the UI it wasn't wasn't operating correctly. So it was tied into this. This bug that was found about the 10 gig in her face patterns.

B

There's some doc issue, I think which might have been addressed. I think sky was looking at that.

B

This is just default behavior that we wanted to add. So when you create a network with the ready state, my assumption was that it would add that network in for the deployment for you samba system, behavior pieces of as we're I'm using ready state more and more.

B

This I think it would be done. We can decide this is done and then that there are bugs against it or we can fix the bugs and then take. This is done.

B

I'm, okay, with the limit, with waiting until those those bugs are fixed.

B

You I think this actually work is done.

A

B

This is a pull request that you, you accept it so basically this this means that, when you're on the when you're on a snapshot, it shows you it was that we had that discussion about. I call the milestones when you accepted the milestones bullet it actually took. This has changed so I'm going to.

A

E

B

Because it was actually emerged.

A

B

C

A

B

Half these are ready state we're going to talk about ice house up streams which I think we've been changing a little bit. I started on that haven't finished it. This is actually you've been making progress on the ws man, jig I, don't know what the what the end state is on that Victor.

C

B

Ya think that I've actually have a user story.

B

Right so for how just to the network, we know that it works, serves out-of-band image. We did that I. Think.

C

We know we didn't, we did not do. We did not the no boot image and machine does not have to boot thing. We did not do.

B

Okay, oh I, see.

C

Yeah because that's not really relevant to adding support for open WS man, that's a different feature entirely.

B

What is it configuring? The servers, what out of band.

B

Oh like this is this figures.

C

From what I, what I assume when I read that is you're, saying we don't use sledgehammer any more on Syria on systems that are using WS man, yeah.

B

C

B

That was the goal: okay,.

D

B

That's why I'm questioning how I wrote that story, because that's a whole nother thing to do it.

A

B

A buddhist sledgehammer is not the goal.

A

C

Just for adding the ws man jig, this is not that that whole bullet points tisn't relevant to it right.

A

C

The whole point was for the open, WS man for the double yes, man, jig and logs. It's the ws manhammer now is that we can talk to the syrup. We can control server using WS man, it doesn't say anything about, you know actually being able to do anything with it. So I would zap that.

B

This whole line yeah.

C

Because that's a different task.

A

C

Rest of that is done.

B

And then let me edit this to call it.

B

Yeah I can vouch those things are those things are done? Okay, so we need a new issue for the actual configuration. This can be a.

C

But it won't be here.

B

That's right, it won't be in that story. So this is the in-band jig.

B

C

It that even a series.

B

This is this as, if you're going to do without the WS man you're into English,.

C

B

C

That's not a jig of its own really at this point right.

C

B

B

Okay, that's reasonable and useful. Yeah is if we had somebody who's interested in the poweredge figure. This is the way you'd have to do it. You think using this, this approach and it spins in the way we used to do it. So it's really a code. Migration.

C

There's more to it than that, but yeah there's.

B

All I know always.

A

B

Hadoop, ready state has been moving very slowly, that song that's Mike and it looks like Mike's on the on the call.

F

He started working or not, I've been doing dog stuff and I.

B

The docs f is excellent, so I'm happy to see that I don't think we're quite ready for the new British yay. Consider what.

A

Is, although, if you're going to sign.

F

It to me you cook, I, don't think I'm actually listed in the project here.

F

B

Nope there's a couple people who should be listed in the project, but aren't let me just add that to the list.

B

Because we have Mike, we have a misshaped decoy.

B

F

Like okay yeah for me, you want to add lhr c dash. Mikey p is the github login. Don't.

A

F

Mikey p got it, you.

B

And I talk to Isaac. He doesn't want to work.

A

B

It more instead of Alex just a matter of getting them. I Isaac needs some good I need to teach him enough, get that he can fix those merges yeah I'm jumping ahead. Although I'm going to talk about like this, I should actually create a story for him to working it so that what what Isaac's been working against is convert you I, page refreshes to Ajax, which I know I would like to see very much.

B

B

Notes1 is one the snapshots.

B

B

And the annealer all use.

B

Page refresh, instead of a.

B

Instead of just updating the individual items- and they really should they really should be working within the items.

A

B

Cool so that'll give him something to commit against and there's already pull request outstanding for this I thought I'm, not sure why they're failing Travis, so we need we I've been traveling.

E

D

B

To work with Isaac, we'll figure it out, I have some work that I have stashed that I want to bring back about that. The one outstanding bug too cool, so Mike, Victor I, know Judds working on chef metal, which we don't have a specific story, for we probably need to start getting a Trello.

B

Randy Randy was investigating doing product management with Trello, and so what I might do if he has time off srp Randy if you're on the phone I was going to might ask you to actually do Trello, but also help ask people specifically for status to do the drum master role. Okay,.

E

That's fine, but my only concern is rob you and I are the only ones who have talked on decade. They were responded in that Google, Group and so are people if you're in you and I are the only ones who are concerned with it. I would say we leave just using it to get up and leave it at that for now, until we have enough people interested in moving over if the two tool being there doesn't mean they will be used. That's.

B

True I, why don't we do this, then? As a start, I will. I will get you in as new in being able to manage these issues and we can keep working through the issues list. Ok,.

E

B

Like from that perspective, then is let's actually burn down the issues list.

B

Get them assigned to the right people, and then we can. We can actually start managing it right.

A

B

Agree, I agree with you. It makes it makes a lot of sense to do that. So, let's let's get them in there and then we can label the get them in get them labeled.

A

E

Then if people start saying this is very unwieldy, is there any other ways then we can bring this back up again. I, like.

B

It I'm in favor of that ok, we could also move to shorter milestones by the way the I haven't been willing to manage it, but if we, if we switch to sprint by Sprint milestones Randy, that would give us a lot of the same features so we'd know what was in this in this coming sprint and we'd be able to start work well.

E

How long is the sprint right now, a week two weeks a year.

B

Technically it's two weeks, although and if we were going to start discussing using these planning, calls to actually figure out two weeks at work that would allow us to still things down like we just did with Victor.

B

E

B

Weeks because I think that would be longer than that and it starts getting hard to manage.

B

E

B

You and I can coordinate offline, actually start building the Sprint's based on the broom released and then we'll start working through that working those things through and then.

E

The other question rich and I have is at the end of sprint: do we have a work product tote to be done.

B

Say that so that again Beth so.

E

The one thing that I've always really liked about the agile process is you're supposed to have at the end of each sprint a shippable product. Do we are we at a state right now we're at the end of each sprint? We can say this is shippable.

B

Shy of having breaking bugs yes, all the work, all the work that's been coming through, it has incremental pieces. The thing that we don't have right now is the full hardware lifecycle and we're working towards that. That's.

D

E

Long as we head low it out.

B

Hardware profile yeah: that's.

E

Fine, as long as we can qualify what the shippable is and what we change between shippable unit, one and chip flows sprint to then that that means we can show progress right.

B

Make sense to me so what let's do? Let's? Let's continue this as part of this email thread and start and start putting putting in some of these expectations in the Quarren and process. Randy done all right. That's.

E

Right yeah does.

B

Anybody on the call have concerns about continuing the dialogue on an email and at least giving us some plus ones on the dial ugh I knew we need that commitment, because we don't want to do it in a vacuum.

F

Now it sounds good to me.

A

Cool: let's do that.

B

All right I like that, because I that's a good, visible place to do it, so we have action.

B

A

A

B

Hey all right, so that's good! I'm watching the clock, salt integration, so this hasn't. This hasn't surfaced through the list yet. But I had some inquiries about doing. Salt integration and I'm talking to the salt team.

B

I think tomorrow, maybe this afternoon haha this afternoon, if I canoeing this afternoon, to see about what actually building a bridge in with the salt pieces, so that we would basically set up a salt minion and tie it into the salt master as part of that process. So I know Greg and I have been talking about how that would work, and so I I'll give you I just want to warn you that it's coming.

B

If people are interested in the salt integration, please let me know for the list I'll send out an update after I've had the meeting, and we have an initial design idea and we'll send something out from the list so that people can come in on it and see what their interest level is.

B

Ok, good to the salt Victor, you want to talk about the the hardware hardware pieces. I know you have a pull pending that I'm trying to I'm trying to just make sure it does no harm in that. You want to talk about that briefly.

C

Unfortunately, I haven't done as much work as I had planned on due to taking my family to the beach the week before last and birthdays most of last week and my air conditioning going out. So it's been a lot of fun, but I am at the point where I have started port.

C

I ported over all of the I ported over V, at least the code for the dell raid in the dell bios bar clamps and engine eyes them and made them able to be able to run able to at least not break anything else inside of open crowbar. All the recipes are there, the cookbooks, the recipes and the data bags I have pull requests out there that make them all import properly and right now, I am working on the discovery, pieces and I had to make a detour to UM add some code.

C

That will make it much easier to actually you do discovery and picking what set of quirks do you use for any given ipmi controller or raid set or BIOS configuration for stuff in such a way that it done won't result in a massive, crazy spaghetti code. So that's been fun. What.

B

Is what is what is discovery mean from that perspective? Is that is that, like the inventory piece of discovering.

C

Depends on exactly what you're talking about, for instance, for Mike, ipmi and BIOS discovery means figuring out who made it what model it is and figuring out how properly to talk to it.

C

That's going to be the bulk of discovery for configuration and inventory.

C

It depend in Vittori as part of discovery. It really depends on whether or not they always whether or not there is going to be a consistent enough inventory method that we can do during discovery. Instead of that having it be its own role and I'm, not certain about that right now, okay,.

B

If you did it as own role, we'd be able to add different roles and which is which is probably necessary anyway, but yeah.

C

We'd be able to do it, we'd be able to well. We wouldn't necessarily have to add separate roles, but the disc up, but the inventory piece would be able to use the information found during discovery, including what the best way to talk to that particular controller is in order to do its work.

B

Well, would that change the role graph.

C

Ideally, not what it would change would be blood information, we're passing back and forth to the jig's to do their work.

C

For instance, of what I'm doing right now is I'm, just as a test of things, I'm abstracting out right now, everything for IP my discovery configuration is hard-coded to assume that we're talking to essentially something where the ipmi controller works perfectly. We don't need any special delays. We can do whatever we want to whatever users week we feel like, and you know we don't have to and be, and everything falls out in the order of operations that we expect to, and we already know that that doesn't work for supermicro.

C

So I'm abstracting that out so we can easily plug in other we can easily. Instead of turning our ipmi configure roll into a bunch of spaghetti, we can set it up so that it just pulls in the libraries that it needs to do proper configuration.

B

To be interesting to review that on next week, that's why we can talk. We can. We can talk to her.

C

Yeah I'll probably help I God willing I'll have full requests out there that demo, how it all works.

C

Sorry but hopefully, I'll have demos out there that show how it all works. I.

B

Think I think that'll be really helpful because I there's a lot of bits and pieces that have to come together to make all this stuff go, and so it's starting to explain it all will help me be able to exploit it to other people better I'm, following what you're saying but I think it will be helpful for me to have some illustrations.

B

C

B

I know I'm building I'm trying to build visualizations about this in the orchestration so I might I. Might it might be worthwhile for you and I to find time to go through it once you, especially once you have a pull request? I can look at it than I dental I'll. Be able to ask ask questions, because this is exactly. This is the heart of the orchestration problem in this physical.

D

B

Problem: okay, actually.

C

I'm going to note that, until what I'm trying to invent here is a way that we can account for whatever arbitrarily weird configuration or command sequences, that Hardware expects an account, for you know glitches that we find you know when we test a specific like a specific IP. My controller, there's a guy through my controllers that you know, may exhibit quarks that only appear at certain firmware revisions and bias issues that appear the same way. That sort of thing and I know we're going to encounter them, because we have encountered them until I. Think crowbar.

C

One and I want.

D

C

Way that the accounts for all of them in a way that accounts for all of them, without generating a ton of spaghetti.

B

It makes sense, let's use the knowledge we've got to improve this and I know. I know you from experience aren't going to over architect it it's just a question of knowing where the wedge points are, that they caused the systems to break well.

C

What I'm doing right now will probably look like over architecting but having written if you having written in quark database in the past and yeah I, want to like something.

D

C

Don't have to I want I want to and I think I've gotten to the point where we have something that would you just enough of what we need that we won't have to go and radically or tear it apart. You know a few revisions down the road. Ok.

B

Let's look after you've made some headway on it, or even even while you're talking about let's, let's chat on it, because I.

A

B

It, but if I can re, articulate the court database to people and how we're solving the problem that will it will help me explain the value there, because right now, every other thing on the market assumes that there they don't even deal with quark right. It's it's a they've choose to avoid the problem.

C

Most of what they do like for ipmi in particular, they assume you have to sleep five seconds in between every command and they don't try to account for controller specific difference or most of them, don't even try to account for controller specific quarks.

C

And kind of the same thing goes for w's man and I lo stuff.

D

B

Sense- and some of it will be unavoidable, but some of it if we can capture mixed lessons, yeah.

C

I mean that's: there's a gym: I've been working on called structure maps that basically incorporates the core of what I'm trying to do, which is we get a bunch of data back from ojai or some other discovery mechanism, and then, once we get that data, we can run a bunch of matches against it. That will go in and pick the one that matches most closely and use it.

C

So it can be kind of fuzzy but still get reliable results.

B

Looking forward to seeing it and then we'll we'll, try and articulate it tomorrow, if I can I might try and draw a picture, but I've been working on an on what orchestration looks like picture. It's not been easy. Well,.

C

This is what I'm doing right now is kind of uh it's right now is something that's pretty down in the weeds. If you're trying to do anything, maybe sort of high level of presentation.

B

I understand it's amazing when I talked up in traveling the country for those for those of you aren't aware, I've been all over the country, les plus six weeks talking to people about all sorts of things, including up in crowbar and it's it's amazing. How fast people come to the question of well I thought that Braden bias is not a big deal.

B

Don't people just rack servers and forget about it, and, uh and so it's important for me to be able to explain the quirks and how, when somebody doing something as simple as a bios patch can actually cause a cascading system. Failure which we know happens from experience, but most.

C

Ending there done that.

B

Hi-Yah across 593.

C

Fortune 100 servers it dealt with the angry fortune, 100 companies, so.

B

It's just I what I want to be able to do is usually if I can pull together in a picture, then people are like. Oh, that looks scary. Okay, that makes sense, and so we will get. We make progress there. All right, we'll talk about this next week, it'll be the major major topic. I think that probably is the right major topic unless we can get further with Chef metal. In the meantime, OpenStack integrations.

C

B

C

But all that's been pre-configured.

B

Chef chef, metal, I actually wrote a spec about this and then have trouble actually making chef metal itself work against our chef server. So basically, this is a pool is that the strategy is to create a pool. I sound of I send an email around. We should formalize that into a story, so one of the things Randy for us to do is to actually create the chef metal story.

B

Brandi had to drop off.

B

Because I actually wrote up a nice I thought it was night.

B

Scription of what would happen, yeah chef metal is really a poorly named project. It really is chef cluster and it just uses a recipe to build a cluster sort of like crevo or one did where you create a cluster. You say: I have this many machines with these rolls on it.

B

What the way that what it does is. It makes the assumption that the chef server can call out to a system and get back a machine that has an OS installed chef, SSH, key chef client installed on it and.

A

B

No, it just eshg, it solves the client itself, so you you say, give me a machine, that's installed and ready, and it gives you and and then chef goes and does stuff on it. It's actually exactly what we described with ready state.

B

So if chef metal expects you to have access to ready state infrastructure, and so we can basically create a pool of servers that are available for chef metal and then get allocated into that pool and just use deployments as a way to manage that is that it seemed I actually wrote about fifty percent of the required code on a flight. It's once chef metal itself is working. It's not based on what we've got. It's not a big deal. That's my assessment.

B

The judge does actually looking into it because I sat down and couldn't get chef metal to work.

A

B

Wasn't I never got down to the crowbar piece, openstack integration, so there's two there's two paths here that I've been talking to people about the first path is to do a foreman integration and I. Believe dave was interested in that.

B

Rdio, take a pack stack.

B

And so I think this is the work that we're doing in salt is going to be exactly the same work that we need to do performance. So it's effectively using the script jig to link into the form and infrastructure that already exists.

B

Potentially, we could use the script Jake to build a foreman infrastructure and so I'm. Ideally, this work, we'd have somebody in the community say: hey we want to take this on or fund it I'd be delighted to have somebody fund. It I.

A

B

Several companies who were interested in participating and could conceivably even have multiple multiple parties participate in the funding of it.

D

B

Think it's a crazy amount of work.

B

Salt work will help investigate that the salt work we actually, we actually have a specific person asking for it and Dave if you're listening and want to see this move forward, just contact me- and we can. We can talk more about that.

B

D

B

Good then, let's talk offline and see what we can do to kick-start this I think I know at least one other candidate for it, the other. The other approach is to take, and this is a longer shot, so sue says: cookbooks are released for ice house.

B

Unfortunately, they did a v1 integration and if you haven't looked at what susa did.

B

B

They call it susa cloud for.

B

And so this is all crow Wow, a nice picture and a rainbow over the tour. They have a really strong OpenStack deployment based on crowbar one I excited as I am about that.

B

I'm not particularly excited anymore about crowbar one, because I know the scale limitations and so there's a there's, a hope and a potential person in the community who wants to use chef to do OpenStack deployments, but on centos, and actually I think the the mantra here is sent acento centos, everybody won't do it on sent us there's some people doing a bun too, but a lot of people are looking at Santos pieces.

B

So if I would I would be happy to coordinate, helping and also coordinate with Susa their next steps on those cookbooks, whether they move them into the OpenStack community cookbooks, or they continue to maintain those cookbooks as a community set of resources. Remember, though, they're actually in the crowbar github, the original, crowbar github. Sorry.

B

So it's OpenStack star is really the in not going to help anybody. It's not really a link, but they're cookbooks are here by design.

A

B

Cookbooks could be ported to open crowbar, it's not zero effort by any stretch, but we could put port those over and take advantage of a lot of the features of open, crowbar, and.and up and tilt you up until June those were being maintained, I think junior, April April those are being maintained on centos of their sent us parody. That was June, so there are actually a lot of sentence capabilities and that's a full H, a deployment and included trove and neutron and salamat ER and heat and.

A

B

A sophisticated deployment so I, don't know. My preference frankly is actually to do the form in pieces right now, because it aligns really well with already state story if Susan wants to come in or somebody wants to help encourage Susa to come over.

D

B

Another way to do it and then the other. The other way to do it is with the salt integration.

B

This end, this, if you like the foreman work, I, don't have as much experience with the with salt.

B

Ok, so that's an update for that I think the time is right for for us to go back to that. It's just a matter of resourcing chef Mehta. We talked about I'm, going to run out of time Mike if you're still on hold on a minute and I'm going to. I have to go mobile and switch this to a voice. We talked about process management planning. We talked about you I issues and Ajax polls, so we're pretty good.

B

So this is, this is excellent.

B

Mike can I turn it over to you for a minute to talk about the documentation and.

A

F

B

For next steps, okay,.

F

I can talk about that so at a high level. What I did first is I. Have this kind of manual process, I I check out the main crowbar repository I convert the markdown to restructure text, pass it through Sphinx and I can publish docs and that happens to work nicely with read the docs, so I actually have mirrored the my I have this process where I I pull the main crowbar repo.

F

Do the restructured text conversion mostly automatically and push into a second repo that I can then trigger read the docs do a build from so that gives us online docs and basically HTML. You also get PDF and epub kind of for free, because it's all going through the same process.

F

Issues outstanding there right now are some of the linking is a problem, because there are hard-coded references in the markdown to dot MDS. I need to figure out how to how to get that working I and make the fix work in both crowbar and read the docs and the other issue is.

F

There are a lot of formatting problems in the existing markdown, so what I'm doing generally is fixing the markdown so that it converts cleanly, which means we get fixed on both the crowbar and they read the dockside, so that's short term and I'm just going to do that, probably manually for for a month or so on the crowbar side. I need to look at how we get the actual documentation within the crowbar repo converted to restructure text. That's the easy part!

F

The hard part is going to be rendering the restructure text in the crowbar you I dynamically, because there are no libraries in Ruby. So, oh, and there are some there are libraries, but they kind of wrap around the command line utility those the conversion. Okay, what I'm a what I may end up doing after prototype. It is see if I can actually get crowbar to just call in two sinks directly and rebuild the docks when they change so the first time around.

F

We do almost a five minute doc bill to do the conversion when we first do and install and then from that point on, the docs will just render because they'll be converted to HTML in advance in the developer mode. We can fire off a make. So.

B

So, actually it it might be a reasonable thing to not try to render the docs like we are doing mm-hmm.

B

Instead of maintaining all that code. Just point people to the internet for the docs keep the docks where they are and then, if you're a developer or you need access to the docs, the restructured test is actually very, is very usable. If you're reading it on a screen, mm-hmm and I'm interested to hear other people's position on just stripping out the dock station.

D

B

Using what you've built as the sort of the docs fight, mm-hmm.

A

F

That's a possibility too, since the dockside is pure HTML, we could render that locally and have a copy of it local to the crowbar server and still have the link to the online docs. I I.

B

Don't I don't know that we need to render it locally. Could.

F

B

I know Victor hates the process because it's slow, yeah and you know it it's not the latest, and if people fix bus you don't get, you don't get the latest bugs we're. Still it still checked out. So you we actually have.

D

B

To give up the code control for it, but I've really come. You know, I'm, not sure we need to render it in the app the way we are as much time as I've spent writing that code. Maybe it's just crap and we shouldn't we shouldn't bother.

B

That what you've done is actually a much cleaner, more searchable documentation inventory, the biggest the biggest concern I have is that now we have to, you know, have a have a system that that actually updates it and renders it, but Travis could do that if it's a simple set of commands, Travis could easily do that for us, yeah Travis.

F

Can do it, we could actually put in a post commit hook to trigger, read the docs if the original.

B

F

Are in the master, crowbar repository, so that's an option and, as I said, we always have the option. If we wanted to, we could render the docs locally with no extra effort. I mean it's literally a make file, so there would be a lot of there. The.

B

One thing that we would give up is we do and the reason I did all this was so that we could deposit the docs from.

B

From multiple workloads, yeah.

F

B

That that's one of the things that originally we did with all these parklands is we needed to have each each bar clamp had its own component, the docs, and now that we've simplified and streamlined a hofstra that that's a big, a need anymore.

F

Yeah, it's a good point that we're losing that capability, so it might be possible to do that. Anyhow. I haven't really looked into it. Maybe an alternative approach to that is just make it easy for the various plugins and bar clamps to generate docks. If we give them the right framework.

B

Right, well then, but then they'd be standalone, docks, I, don't know it just it doesn't feel like we are.

B

Patching all those pieces together as much as we used to and that the workloads will be have their own do it. Maybe it's fine to have standalone docks for the workload.

A

B

Birthday than I, though so I'm carrying use cases in that, maybe we don't have to worry about right, ready state means crowbar, has its own docks. If you bring in a workload, it's it's. It's own dock, I.

B

You know if its own independent supplement.

A

D

F

We pull in things like the Hadoop infrastructure. There will be documentation there. Some of the bigger workloads are definitely going to involve extra documentation. Hi.

B

I would hope they do, but this documentation need to be integrated into it, create a single thing. Oh yeah,.

A

B

You know and like the hardware stuff might end up being integrated in as well, but we want to be able to run without the hardware loads, which is why it's designed like that Victor. Do you have a thought? Are you still following this.

B

A

B

Mic I I'm very tempted to take all that complexity out of the product. Yeah.

F

Because my tendency, what I was leaning towards doing is once I had the conversion working I was literally going to do one massive check in to convert all the docs to restructure text. You know through the automated conversion, which is ninety percent of the way there and I was considering. However, the docs URL just point to a local directory that had the HTML rendered in it already.

F

Oh, and that way we we would end up moving all that we'd end up removing most of that code from the crowbar server and probably probably replacing it with a row.

F

So it might be possible to keep the local copy. But if we with minimal effort, but if we lost it and just pointed to read the docs I, don't think we'd lose anything I would tend to agree with you.

B

Let's do that we always can, if we need to put the cutback in, we can put the code back in, but you know I think everybody agrees. Restructured text is more expressive that keeping up with rendering that code is as decreasing value to stuffing it into database. Yeah.

F

B

We're not we're not there's a whole bunch of things. I wanted to do that. I'm, not sure, is relevant right. I wanted to be able to have. You know, links from individual screens to the docs. We haven't gotten there and they are going to be so and we.

F

Can still potentially do links from individual docs to the individual screens, the docs accepted to be an external hyperlink. What I see is the big barrier is today it's been kind of hard to read the documentation until you had crowbar installed. So it's a chicken and egg problem, at least with the doctor, itation hosted and externally it's possible to read it before you start the install process, I.

B

Strongly agree with that. That makes a lot of sense.

A

B

F

Work on me go ahead.

A

And do that? Okay.

F

It's less work for me, so trust me. It's a lot easier to do that. That's.

B

Well, that's I, don't I don't want to create any any side effect work. That is anticipating a feature that we don't amah sure that we need any more one.

D

B

Thing that would be interesting. Is it possible to throw metadata into those files like a grid or some type of ID number that we could we could track? I was just thinking a date based quit something something simple enough to find a document. If we wanted to cross-reference them, I.

F

Think that's possible. Okay,.

B

Don't don't spend actually, you know what that that's over a roll.

F

B

F

No I mean the way the way the docs render when you're using Sphinx. It creates it's going to create anchors within the HTML for most sections and subsections and the docks so I can always linked to an you, can always linked to a tag somewhere within the documentation pretty easily. And there is an indexing feature which I'm not using because we haven't been able to put index tags in the in the markdown. But if we actually convert our format to restructure text, we also get indexes.

A

B

We've pushed back on this way too long. Let's switch it, let's pull out the code that renders it and just point people to the website, I'm delighted by the work. You've done it, it's okay,.

F

B

It's going to help everybody quite a bit so.

F

I'll move I'll generally follow that path. I, don't think it's going to be a big project. It's just going to be a little bit of time to do the conversion and see how clean it ends up being. Okay,.

B

I'm sorry I pushed back on this at all. This is the right path, very exciting, cool, so I'll, something that I'd like to move to the mailing list. As our final topic, just I'm I've been hearing, people complain a gap of a cobbler and in talking to people cobblers. Clearly the alternate choice that people seem to have, for you know the default in doing these provisioning deployments. Well, I, don't think it competes with us from a featured perspective.

B

Definitely once somebody's you can cobbler the interest in crowbarred declines, and so I want to start pulling together a crowbar versus cobbler sort of fact sheet and do some do some research and I'll pass that through the mailing list, hopefully get some contributions on it and hearing a bit of complaints about cobbler.

B

Nobody seems very happy with it. Razor is that is in a weird state.

B

The old razor founders are now doing something called hanlon, which is a fork of razor from way back and it looks like you know, razors very long. Puppet hanlon looks like it might align the chef I'm trying to talk to the self guys and see if they're, if they're looking for something. But it.

A

B

A set of use cases for crowbar to be streamlined replacement for cobbler without and we might have to look at limiting some of the complexity from that perspective. To actually be a cobbler replacement, I'm just experimenting with it. So I'll pass that through the mailing list and see if we can start a dialogue about it and a feature, a feature set to present.

A

Anything else, oh.

F

No not for me, we.

B

Have I think this was good? We heard a lot of ground today.

B

Right, everybody I will talk to you next Wednesday, if not sooner thanks.

A