VMware Octant Community Meetings, 10 Sep 2019

Previous Meeting

⏯

youtube image

►

From YouTube: Octant Community Meeting - Sept 10, 2019

Description

The first ever Octant Community Meeting!

The Octant team talks about work done in the previous month, and what needs to be completed for the 0.7 release.

A

Said he did that.

B

Not in test it out hmm does it look like it's recording right now. Yes,.

A

B

Bright, we were recording Sam.

A

Saves a day, thank you all right. So today is September 10th 2019 I'm Baron miles, and this is the the bi-weekly achtung community meeting and for everyone paying attention. We do have an agenda doc and we will make sure that gets shared for your viewing later, but let's get started. So this is our first. This is our first octant public community meeting and and really for everyone watching. What octant is is a it's a web-based application that allows you to visualize.

A

What's going on in your kubernetes clusters, we are putting in special work to make sure that developers have the information they need so that we can provide more information around. What should we get with tube control, get describe, delete, apply off and anything else we can. Think of so really. This is more of a shortcut for anyone using kubernetes, so I want to get started today with the review of our next upcoming release.

A

We had originally planning or stated that we were going to do this release either yesterday or today, but that's not going to happen so we'll probably move it out towards into the month, but I wanted to go over the high-level features that will be in the release, so one second here and I will actually share my screen. So everyone can see this.

A

A

There we go that made me go so we we tracking all of the work for octant on github and specifically we're using Zen hub, and this view right here is just the Zen hub plugin for github. So what we've done is we've actually broken the work down into right now into multiple releases and I need to clear this filter, and so we have multiple swim lanes here and we have a sprint backlog and things that are going on, but I just want to focus on 0.7 and what will and what will not be in there.

A

So right now, 0-7 has 10 about 10 items that have yet to be taken up, but let me just go through them from the top. So what we've been trying to do is focus on a mix of bug ins with enhancements with batches buggins bug the bugs and enhancements. So a lot of this release was bugs that we found when we released 0.6.

A

So let me go through the top, so we have this.

A

We have an issue with tabs and sharing of state, so we don't want the state to be shared between two different.

A

Two different tabs and that's happening right now, so this is number 179 number 43 is we have this idea for a nokton status for you. This will actually get moved out of this. This 0.7 release, but we wanted to show the status of octant, so you didn't have to rely on the the command line out plate as much, and we want to be able to move those things here, but with time and focusing on other features, this one is not quite yet all thought out, so we won't be doing this one.

A

Actually, let me remove this right now.

A

A

Next one is number 49 and it's handled time outs when contacting master and the issue here is that sometimes the master can go away or it can be slow and we want to be able to give a good experience right now. We just complain loudly that there's issues contacting the master, but if you're not looking at the web are looking at the command line output, you might miss it.

A

So we want to get rid of that, and number 149 is documenting how verbosity works inside of octant, because you can actually launch it with different values for logging, debug output, and it's not everyone thought about it in the same way.

A

So we're going to make sure you have documentation around that this next, one 56 is make it possible to make the resource viewer the default view, and this is actually part of a large pot, a larger project where we want to be able to control the initial view just in general or or you could, when you go to opt-in right now, when you go to octant here in this dashboard view, and it might not be the best view for everyone.

A

So we would like to make it so that the resource for you could be the default, and this next 146 is plugins. Ready shirt with the same gbk is duplicated I'm standing firm on this one. This is not a it's, not a bug, it's actually a feature. The real problem is, is that we can. We don't have an answer for what happens whenever you name plugins the same thing and coming from the community.

A

The plugins could be the same thing so I don't know, I'm actually still looking for more input on this one of what we should do, because we can do a tooltip like I said before here.

A

Maybe we should just say: hey, that's just how it works, and if you don't want it to be duplicate, they come up with a better name, and you know this is the low code, no code solution, so I'll actually be pinging, our our public slack with this one. This get some feedback because I'm still not sure what we should do here and so.

C

In this case, the customer upload the plugin that just used to be for a name for it, but it was exact same code of the plugin, so the UI integration just create a new the same one right well,.

A

No, so whenever they it what will happen? The reason this happens is the copy, our sample plug-in and then not change the plug-in name. So, two of the same name.

C

The plugins company unique differentiator like a unique name beyond what the UI shows mm they. We could add that right and disallowed two people who have the exact so, for example, the plugin have a unique name and beyond the UI name, right whatever, with a visible name that we display and then we can basically force their unique name to be unique. So.

A

The question Michael would be, then: what would that unique name be they.

C

Can choose whatever they want right? It's just something: that's associated a plug-in. They couldn't choose that and as long as that's not duplicated, then if someone goes in intentionally, the UI creates two separate plugins with the same name in the UI, then that that's their fault, oh, that.

A

The next point that I'm making is that if you do create C plugins with the same name, that's on you, we can't differentiate between plugins but we're just it's basically garbage in garbage out for us. So if your given name we're gonna, take two of the same name: okay, yeah, so that's that's where I would like to leave it and just say yeah, you know rename your plug-in rather than because the alternative is that we do something, and now it's not I mean we can make it we can.

A

We can make it idempotent, but the problem is: is that the the resolution that we could give up wouldn't be satisfactory? Well, if we just said, if we found another plug-in name, we just wouldn't load it. No I! Don't like that at option. Ball I mean.

D

I could see a scenario where, if you're developing a plug-in, maybe you have like a one version of it running and they did another version of it because continuing to work on have them both side by side. You probably want the same. You know plug-in name for both right I mean.

A

Yeah you could, but the thing is and here's the interesting piece, the path for that link is not the same. So it's a unique path, just the same text so I actually don't think it's a problem. I just think yeah. Just don't do that. Okay, so.

E

I put it in there is my comment: yeah well, you know just go ahead and just resolve it. Just won't fix. Lets me know. Yeah.

A

E

A

Going to I'm going to put in there and I, don't really think this is a problem and then resolve it, and this was internal I know Sam, flew created this one months ago. Okay, hopefully his feelings will not be hurt all right moving along the list. Next one is 139 in the and just as a little sidebar I will only go through this one time. This is the first time and I want people to see what we've been working on and what comprises a an comprises, a release.

A

This will be changing that I just wanted to go through it for our new people on the call, so the next one is 139. The feature requests for showing our back and much more useful form. This is this still in process and once again will not be a 0.7, but the gist of this is that I think we can.

A

We can do some really good work around showing in our back showing how to edit our back so managing cluster roles and cluster role bindings and showing it in a way that is easy to understand for a user, and we haven't put any design work into it. The person who submitted this floor request actually had some ideas of some neat things here, but the one issue that we run into is that we can't do it just like tube control off.

A

Can I list I want to have it just a little bit more interactive than that tool, so just this actually needs some, some backend design thoughts and some UX design thoughts of where we would like to be on this, but consider some pretty important, so I'm going to remove this one to the next release.

A

All right, the next one here is enhancements and it's a show all namespaces and cluster overview.

A

I actually don't know how this one got in here and the gist of this one is: if you have a smaller cluster and like the cube control like the kubernetes dashboard, you can actually set it up to be all namespaces and get all your names faces in one place, so we could actually just have all namespaces octant from the start was set up to be namespace specific and the way that we work, we can be either or we can be, or we actually can't support both, but they will not work in the same way, because the way that we look at namespaces and because namespaces can have different permissions on them, we actually can do a list across every single namespace.

A

If we don't have cluster wide permissions, if we do not to have cluster right position permissions, what we actually have to do is iterate through every namespace that we can talk to and then combine all the results and then show that it's possible, but probably should think about what this means before attempting to do it, and the next one is to 35 out of hacking MD, we probably should rethink our readme just in general, it's getting kind of big, but the hacking MB will allow people to have a complete set of documentation if they want to work on optin.

A

That's something that we I, don't think. We've done a great enough job with yet, and this last one is the wrong URL. There was a case where the URL would change to the wrong item. Actually, let me see what their example was.

A

Alright, so this is inside the resource viewer whenever you clicked on it, it gave you the wrong URL. This is bug. This will be actually worked on and fixed, so this is what we have to go. This is this is about the work that we need to do before. We can release a 0.7, but in the meantime, we're going to start planning for 0.8, we'll make some noise about that and in our chat and on our mailing list, but that has not been worked out yet any questions.

A

All right, not no questions from you, Michael nope, okay, so the next thing here is I would like to get us to a point where we can do demos at least once a month, but it'd be nice. If you could do demos all the time but and Lua that I actually do have a demo or some exercise, do you have a demo of something that have some work?

A

That's been going on of things that something will be in 0.7 aren't moved to WebSockets and the reason we moved to WebSockets is because, with we were using HTTP event streaming and HP event streaming is you basically can long pull on on a URL and it will send events which you can then apply javascript handlers to to do work when things come in the problem that we kept on running into was that the front end of the application and the back end of the application were both holding business logic and because they are both holding business logic?

A

It was sometimes hard to know who owned this business logic and right now, all the business logic should be in the back end. So move the whole app to WebSockets and now because it happens WebSockets, what we can do is make sure that the front end is basically just a shell for showing output and the back end actually is responsible for everything all the way down to when the front end, when the front end has a URL change, it'll actually say hey back in my URL. Is this the back end?

A

Can then synchronously respond back and say no, your URL should be this and the front end should adjust it and take the content that it's going to get, and with this we have management of context and namespaces and the navigation and and query params, and let me show you how it all works or I'll show you a demo of it.

A

So, let's see here.

A

So this is octant running locally. For me, and this is the WebSocket view- looks really no different than the end of you. That's on purpose, but man'll all these sections, so you have this sidebar section and the namespace selections and the filters and the context in the top right corner and the overview and I mean like all the component views. This is all served by WebSockets and like I was saying before whenever we change paths, we're actually telling the backend what our path is and our back-end is telling us what our content should be.

A

So it seems kind of it's not synchronous at all. It's actually it's very a synchronous and just to show you exactly what I'm talking about here. Let me reload this page.

A

There it's a WebSocket and there is traffic going back and forth all the time. So what this gives us is now it's way easier to test octant in its parts. So the front end, because there's no business logic. Now what we can do our testing for is basically when an action happens, upfront say: I click, a link or I select, uh I select a new name space. We can now just test on those, but we don't have to be concerned about what the we don't hear about the business logic.

A

We just be concerned that we're making a WebSocket call and you'll notice that that was blank. There's there's two tiny little bugs in here that we're still working out right now of making sure that we're always updating content but you'll notice that that this thing did change to its new namespace and a neat thing is: it was kind of hard to do. I'll close this window is actually taking the reading in the the path whenever the page is loaded and making all the necessary adjustments.

A

This is all pretty a big deal, but the neat part about this is now it's easier to add new functionality to octave. So let's say if we wanted to add a new some kind of new bar up here or some kind of thing.

A

Now we just create events for it and- and we can unit test it very well on the back end, and we can make the components and have those unit test it very well and we can marry those together rather than having a lot of complexity in the front end for being able to display disparate items. So I told Sam yesterday, I would do a walk-through, but I want to know.

A

If you, what kind of what level walkthrough you were interested in and if you want to actually see the websocket stuff a little bit more I can I can actually show you a little bit more about that.

B

Yeah I mean the type of walkthrough I'm really looking at is I say you know: how do we? How do we make sure that the websocket is actually returning the information we want for a printer generally? What debugging looks like with the tools that you're using currently okay yeah, all.

A

Right, so let me show you a quick debugging type thing that we can do with with octant and WebSockets. So what I'm doing here is me. Let me just change a path here, real quick and the reason I'm doing that is so I can get.

A

This I can copy and paste this little Jason piece right here and what we'll do now is I have I was telling you about this browser web socket client, it's a Chrome extension and it just does what sockets so I'll cut and paste this message in here for a second, but now the URL is localhost, sometimes 7/8 di /v, 1, slash strain.

A

So now we can connect to this thing, and now we can actually see the messages come in in real time from the backend and you notice as soon as I connect it and I didn't do anything. It starts sending navigation context and the content and for page because it's gonna make some assumptions about what we are giving given some defaults. So right now we haven't sent any message to it and this is what will happen on the front end and the front end digest this stuff to present the view that we see right here.

A

So we were to make a call and I'm just going to send this JSON message right here, which is a set content path to a different namespace and once I do that? What we'll see quickly is that now, when I set the content path, it actually came back and told me update your content path and, and now it's sending content, which would be basically this page. So this is actually this is the this one would replace things like curl.

A

So now you could either use this or you could find the command line web socket tool to do that and like I was saying before, there's um there's WS cat and I'm sure we can find more curl might even do it to tell you the truth.

A

Cuz curl does everything, but this is it's easier to understand now, for the one reason it's easier. So let me clear these messages and send this again is now, instead of just having an opaque, a path with a query parameter now what we have is actual function, so we can actually say set content path. I know what the payload looks exactly like this, and on the back end, when the backend sending message to the front end, we can now have it the front end, all of it based so you'll notice it in the first column.

A

Here it says type. So, whenever the front end gets a namespaces messages, it looks at this. Current name faces that it has in an update, similar or there's also a namespace one too so it'll change the namespace that it's using and we see the one for cube config, and this is actually very similar to what we were doing before. With the event handlers.

A

We were actually using the exact same name to tell you the truth, and we were just using them as event handlers, but now we're processing them as WebSocket messages and one of the biggest benefits that we're going to get out of this is that before, when we are using the event streamer, we had to rely on a client being connected to create this polar where we could send messages. The problem is: is that any particular case? There was no way to broadcast to all the Polar's a message.

A

Let's say: we've had an error on the back end or be able to communicate things that we don't know about now in real-time. So now, with this WebSocket and moving the way that we're managing this state, you can actually just talk to one website, a client or you can talk to all of them and you can start sending them messages.

A

So now we can start distributing the generation of these messages throughout the application, rather than just having them in one place, which means ultimately that it should be easier to get these messages in there, and we can put this code closer to where it where it should be.

A

So that's that's the reason why we talked about this months ago and I said it was gonna, be hard and I didn't want to do it and what brought this on was we had an issue with context. Switching and I realized that it was very hard to let the front end know all the state without having to send the state. So that's whenever I said well now it's time to move WebSockets because I don't want to share all this data if we don't have to share all this data.

A

So that's that's where this comes from, so I'd be I I think we have just two tiny little bugs that need to be fixed, but I would love for everyone when to start trying it out and seeing if they can break it, because that's a it's. It's not actually a big change, but it is a different way of what we're doing before so still very new is.

C

That a cluster somewhere where someone can go and target octant against it. So it has a lot of data populated in apps, or should everybody just create their own cluster and try this out so.

A

One thing that we we're just saying use your own cluster: we actually have data inside of octant, not a large amount, but I'll show you where we have it so under examples. Under the Occident project itself under example, and resources we actually do have some.

A

We don't have a large cluster, but we have so. We can do a lot of different types here. Okay,.

C

So you couldn't deploy all of these and basically get them a good sample view in there got it yeah.

A

Yeah, you definitely could and we can apply them. We can apply them right out of here. If you have it and just use two right now, you keep control apply in the future. We will, it will change like.

B

The fashion store or whatnot one of the examples in there no.

A

So there are other examples too, and this is another project that I'm working on outside of this group getting more community things. But if you wanted to try an application itself, Google Cloud has a project. Oh it's some kind of store, we'll get that into the notes of what that store was, and you can apply that to because we definitely use that during development. So we could have a lot of web services running at one time and being able to see them all work together.

A

I think that would be super helpful, especially to show off resource viewer yeah. That's a that's actually a good point, so you put that in there the reason I didn't before cuz I, just actually just didn't. Do it but I think the whole team used it. We just didn't put it in the docs Sam just put the link in zoom chat and you can put it in the hack MD all right. So that's um that's that. So let me stop sharing here and actually I'll share. I'll go back. I'll pick this thing over here.

A

So we're still sharing things so.

A

Tim, could you find that and put that in this link.

A

Yes, absolutely all right! Thank you, alright! So um now that we have half an hour left or less I wanted to actually solicit more feedback topics or more discussion topics from everyone else on here, I'm talking for half an hour. So what else do you want to talk about.

B

We didn't drop a quick note about the website octave, so we're going through final changes with the designers and I. Think most of the content has already ran through editing and legal. So hopefully, once the final approval comes out, they'll actually give me the code for the website, so I could start doing things to it, because they're not going to release that until the contractors get paid but I think we're looking out at a timeline for probably a month for the website.

C

All right, yeah.

B

I heard on the website.

C

Asked Amanda and since Jonas is out to make sure I get to view and review I haven't seen a what it looks like so I'd love to actually check it out before it goes, live.

A

A

Anything else we want to talk about any feature. She would like to see.

D

Late, at least what a question: yes, so you know I demo octant at VMworld and using fusion running, mini cube and everything worked. Great was a great demo and the one thing that I noticed, though so I you know, has the clusters into the ocean and then I've got my mini coop stuff is the stuff in digital ocean I can do the port forward, but the stuff inside of mini cube, I'm not able to is there is that a known issue haven't filed a bug yet but know.

A

That it's not an issue, that's actually the the one I don't use. So yes, it's funny I, don't use many cute I've used, I, tested, asan, digital ocean and AWS and like something we stood up, something / stood up a cluster and then I haven't running locally. What kind but I don't use mini cube. So maybe that is something that that could be above yeah.

D

I caught file the issues you know so I you know, wrote a a homebrew formula for the docker machine driver. So we were testing that getting that working. So that's sort of how I got in touch with the homebrew community and whatnot, but yeah. It's just doing the demo and I'm like alright. So you install it with this doctor machine driver homebrew go and then okay, let's show it it up and I'm like look how port wait? Don't port forward yeah, Michael, I? Think overwhelmingly folks in Hans.

B

Kind right and we're use kind, exactly so yeah.

D

The where I draw the line between the two is like kind as if, like you're building stuff, that it's gonna, you know live in a real big, kubernetes environment kind is the way to go if you're, like learning it and trying to figure it out and poke around and like even just get applications and services running many kids are really easy: click, the Go button and away you go, and it is a certified kubernetes installation. So it you know, makes sense. No.

A

So um actually, I'm suggest create a create, an issue for that, because I'm not even I, was not aware that it did not work with with mini cube, because what we're doing is we're.

A

Let me see we're creating we're using a port forward, API we're not doing any of the heavy lifting ourselves, so the the the question would be if you can get it to work with cube control port forward. Why wouldn't it work with octant, because they're actually using the same code path, doping? That.

D

Make you need to support that particular API and it doesn't write down yeah.

A

I'm not quite sure, but create a create a bug and we'll definitely get on that cool.

D

One thing I will also note, is: there's a an ongoing documentation, bug I'm happy to pick it up, which is around documenting the differences from the official kubernetes dashboard I. Don't know if there's an owner on that, but if there isn't I'm happy to add that on my plate, thank you for stepping up yeah I just want to do the right thing about right. Go to you.

A

Yeah so I'm, my stance on that is that all right, I'll, let's take my stance on it, kubernetes dashboard, is a dashboard. It is nothing more than a dashboard, often actually started out as a different as a troubleshooting tool, and we realized pretty quickly that you would need a dashboard to be able to explore during troubleshooting.

A

But ultimately that is not the only things that we will be showing or hope to be showing so I didn't I didn't want to say that we're hey we're just a competing against this as a dashboard, because that was more of a byproduct of trying to solve another problem, because you know what dashboards are good ideas sometimes, so that's that's how to look at it, but I I mean if someone wants to put that in much nicer words than what I just said.

A

That would be great happy.

D

To and then the only other one I wanted to comment on was I, see that there's a dark same request if Wayne needed any help with that I've gone through the exercise of hurting my website, so that it's dark moto where there's a couple of CSS flags and not particularly hard, you know happy to Wayne gets stuck or anything like that. I did a China into you.

A

Know what you should ping Wayne wings yesterday, but you should ten, you should ping well. Yeah I will cool.

D

Well, thanks for you know, I love this tool, I think it's great I love showing it off folks that get it they're like do this thing's fantastic well,.

A

That's great, we just I, really I can't wait until we get more of the vision implemented because it's actually, if you think this is neat, wait till you see some of them some more of the vision. I mean it's not a secret things like, since one of the goals is to basically not replace but to give another look and feel too cute control get keep control. Describe, we actually do have some exploratory work for embedding a terminal inside of octant problem is where is it gonna run like? Where does that terminal thing run?

A

But you know that we can be creative with that, but that's, but trying to give you all the tools that you need to do your job right here and then the things you don't know about. You know putting those in front of your face, because that's that's really what it comes down to.

A

We also have some discussions going on with the resource viewer. That's want to make it even more useful, more tactile and really show the power of being able to do that and combined with that being able to extract more statuses from objects automatically, because right now, it's pretty manual but watching what the K NATO project did with how they're working on some of their duck typing and yeah Mel. Believe it or not. This is a thing. We figured out that what we can actually do is um we can.

A

We can actually make this duck typing with conditions for any workload generating object, because they all generate conditions somehow and being able to show those matically. So we wouldn't have to maintain those things manually and trying to figure out how to infer more logic and then infer more objects that are related without having to hard code. Those because, like right now we have the hard code a lot of those to make it work.

A

Yeah. It's a lot. I think that is pretty exciting, though I'm just I'm just happy that people really enjoyed this project as much as they have so any more discussion topics today.

A

All right, well, I'm not going to keep you all here any longer so today, I guess we'll just call this an end thanks for everyone for showing up we'll get this video up as soon as we figure out where it is, we do have a youtube site for it, so we'll show up on YouTube and we'll do this again in two weeks. Oh, what's that what's today, attempt so we'll do it again on the 24th, alright, see you all, then thanks everyone all right, all right, bye.