VMware Octant Community Meetings, 26 Aug 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Octant Community Meeting - August 26th, 2020

Description

August 26, 2020
Agenda
- Revisting RBAC and restricted access.
- Need for speed updates.
- https://github.com/vmware-tanzu/octant/pull/1242
- Moving to Angular 10
- JavaScript plugin updates.
- Generating TS
- New repo coming soon (octant-ts-lib)
- Yeoman generator updated, but still wrong.
- KinD plugin demo

A

uh Hello and welcome everybody to the august 26th uh octant community meeting uh today on the agenda. We have a few items and get that shared. um We're going to be talking about um some updates to the kind of restricted our back and and and revisiting our back and restricted access. uh We're going to be looking at some of the updates that have recently happened with our need for speed changes and uh there's. There's gonna be some talks about moving to angular, 10 and and clarity.

A

Four, um I think we're just announcing that we're doing it um and then we'll get into some javascript plugin updates and then there's a demo on a kind plugin that I believe sam's been working on so uh I'll start with the revisiting are back and restricted access.

A

This has been a long standing issue within octan, uh currently the the current implementation of octane. What it does is, there's a has access cache and when it attempts to access a resource it, it makes an entry in that cache for the the the kind, the version and the name space, and uh then it keeps track of any time that is received in error trying to access that resource. uh It will then uh make attempts to continually try to access that resource with uh kind of a logarithmic back off.

A

um This works pretty well, uh if you have like the right set of access, basically like if you have very, very restricted access where you so like your your scope to say a specific name space with only the ability to list deployment replica set pod uh like list get watch for those items, uh then this starts to become very slow uh as you're waiting for a lot of things to kind of back off and and uh get those errors from the api server, uh and it's it's mostly slow in the initial launch of octane and and usage like as the system settles into its back off patterns and it kind of knows what it can get to uh then there's less uh slowness, but the the initial experience when you first launch octane in a are back restricted clusters, is pretty bad.

A

There's a lot of errors that just get spammed right out into the console and uh there's a lot of loading indicators and waiting and waiting and and it's it's kind of this unsure state of. Are things going to work correctly? uh So uh I'm I'm considering some new ways to approach this, and one of them is to kind of remove the the what we do now, which is like this backing off and logging approach and kind of switching to a to a try once uh and forget approach.

A

But by forget, I mean put an entry into a page within octant so that you can see what uh was attempted and failed and then be able to go ahead and say like retry.

A

These things like I have a new uh you know I I my cluster role has been changed and and the bindings are updated and I just want to refresh this context uh so go ahead and try that stuff again um we're trying to be like really smart about it and and be like oh we'll watch for it and we'll automatically refresh and we'll do all this, and it just turns out that that's uh that's a lot of work for very little gain.

A

uh If we just surface this information, uh let you know that it's happening and then give you the ability to kind of just retry doing that manually is much better than doing doing. That. Well, manually is much better than doing it poor automatically, so we um I'm gonna, be exploring this area and hopefully put up some pr's later this week and thursday and friday that will attempt to uh use this new model and- and we can solicit some feedback and see what people think about it.

A

Questions comments concerns there.

A

All right uh moving on to the need for speed updates, so this came out of the octant developer summit that we did. um You know as I clicked this. I just thought man I should really inspect the before. I click on an open document like I should really inspect where that link goes, I just assumed it went to the, but um so this is uh the dynamic components, pr that uh brian lyles did and and what this does.

A

Is it there's a whole suite of changes in here that that kind of move us towards um dynamically loading, the content of a component versus trying to do them both at the same time uh previously, especially with data grids? We were getting this getting into this thing.

A

Where you had many data grids on a page, they would take a very long time to render, especially if you're, putting content into them, but an empty data grid will render onto the page very quickly so by rendering them all empty and then using like a kind of a set timeout to go and then almost immediately fetch the same data that we would want to load in there.

A

The experience on the on the user side is a much faster feel of the information being loaded and presented on the page, and that's because all of the containers that are going to have this information in them get presented almost immediately uh and then, as you cycle between new tabs and and different views, you'll, the the content in those containers is what is what's being updated. So um this is a great improvement. We're still going to be doing work in this area. There's some fixes that that need to come down from this.

A

I think loading indicators uh have some walkiness to them now and um there's also. There was something else that that this uh kind of broke or interrupted. I can't remember.

B

One story book: one story, book story is broken. That's.

A

Right yeah there is, uh it can't fix that yeah right here, storybook for the dynamic components, so so yeah. So that's uh that's where that's at we're going to continue to make speed improvements.

A

um We had mentioned in the last community call that that we are going to be going down this road and making some uh performance enhancements and- and um most of it will be you'll see most of the stuff happening on the front end more so than the back end, because our the back end is uh well, it's already pretty fast, uh and then you know the main limiting factor for the back-end is how fast is the api server, uh because once we consume that information, we render it out to the stream relatively quickly.

A

um I'm sure there are some optimizations we can make there, but most of our gains are going to be seen on the on the front end. So uh this will be in the next release, which, hopefully will be, I think, we'll be doing one relatively soon. There's there's a lot of good stuff that it's landing in and master right now, so uh likely we'll be doing uh uh 0.

A

16., oh dot, 16 soon, yeah sam's nod says that I I picked the right version number. So that's good uh any things I missed there or stuff anything. People want to add.

A

Okay, the next one I'll let someone else talk about this- is the moving to angular 10 that was put in there.

B

Being on that, that can give a quick update. um So, yes, it's happening, we're moving to angular 10 and clarity 4.0.

B

um I had some issues with the monaco editor, but I just dissolved half an hour ago, so I'm just in a final phase of testing making sure everything still works and hopefully that pr is coming later today.

A

All right is there anything uh with moving to 10 and uh angular 10 and clarity 4 that we um look at like. Are there any because I don't? I haven't really looked at like the release, notes and highlights so what's kind of like what's the the high level take of like what do we get.

B

Right so so for angular, uh it's it's they they describe it as a d minor update uh because they released nine. I think four months ago, something like that. So it's a there's not that much stuff there.

B

um Clarity 4.0, um is um upgrading to to the new model where they they have a core and then angular and react on top of that, and so that's the big change there. It should not. I think they have some new components too, but it should not affect us at this point much uh along with those changes. We are upgrading the versions of other dependent libraries. um I forgot all of them now, but I'll list them in a pr.

B

uh So it's a it's. It's basically just bringing everything up to the most recent.

A

uh And so, with with the current version of clarity that we're using- which I think is three one, eight three one, three one eight, I think uh are we so so that doesn't use this new clarity, core model.

B

Right, oh actually, it does. Oh it does it's the only use, but uh I think they they changed it a little bit. So it's it's a little more. uh I think at some point you'll have to the change you'll have to do is uh the clarity.

B

Ui module will will go away, go away, got it yeah and clarity, icons too, so uh we should probably look into that right after upgrading just to make sure I don't want to do too much at once, but we may need to remove those and to eliminate some of the libraries we don't need. Hopefully.

A

Yeah great thanks for that.

A

Update so moving on to some javascript plug-in updates. uh I just want to go over these quickly. So, um as we mentioned, the javascript uh api for plugins is gonna, be continuing to get some significant investment.

A

um We uh it's still very early for the javascript plug-in uh api and and general like the supporting libraries and documentation around it, but we are we're working to actively improve those we're building out some plug-ins of our own, using it so we're kind of dog fooding it so that we can make those improvements and and kind of, uh I think, as scott said earlier, you know sand down some of the rough edges around around building javascript plugins.

A

So one of the things that's that's there now is is we're generating some of the typescript code uh out of octant from our component models. So we have these components that are defined and go um we're and we we've annotated those and now we're generating uh some factories and helpers off of those component models, um so that that that's something that we will be placing into a new repo soon, which would be the octan ts lib repo.

A

It's going to contain those generated models, as well as a set of helpers for creating components working with the dashboard client working with um you know, sending alerts and and and sending events up into octane uh dealing with uh you know, routing and path, handling and navigation and stuff, like that, uh we've we've slowly been kind of building out this, this set of helpers and and extending them and trying to figure out what all those patterns are going to look like and kind of what we want.

A

The shape of this thing to be so um with that said, the the the yeoman generator that is out there for generating octane plugins has been updated. It's going to have these new um automatically generated libraries baked into it. It's not the best model, because you know once you once you generate your plug-in with the uh like. You do yo octane plug-in uh well now you're pinned to this version that came with the generator you used.

A

So that's kind of the the motivation between behind getting that octant uh ts, lib uh library out there is that now npm and yarn can be, can use that to install those files and then, even though you've used the the yeoman generator to scaffold your plugin, you can easily update those libraries as we continue to make improvements and you don't have to go and fetch a whole new generator and regenerate a project and then copy things over.

A

You could just npm, update or yarn update your your octa-ts lib to get the the latest set of changes. So that's that's the direction that's going in uh we've recently merged uh some prs that now allow the javascript plugins to send events. uh This was leverage the work that sam did for enabling send alert in the uh go plugin api.

A

uh We took that same approach for the javascript plugins, um but being that it was on the javascript side, we were able to actually uh be a little less specific, which was nice and so on the javascript side. It's send event and sending an event can be any events that the front end currently supports, which right now is only alerts anyway. um So so yeah there is a pr up from scott to make the front end uh support the content path event, so a plug-in will be able to send a content path, change.

A

Think of that, just like you would do a redirect after posting form data or something like that. This will be very useful for a lot of folks. I know the there was a helm plugin that wanted to do something like that. There was a jx plug-in that I wanted to be able to redirect uh we're creating a plug-in that we want to be able to um do redirects with so a very common pattern that people want to be able to do so. That's that that'll be there soon uh and likely in the next release.

A

Anything I missed around javascript plugins.

A

I think so they're moving quickly they're getting better. uh So if you are writing one and you run into issues, please open them, uh reach out to us in the slack channel um and that's it so now there is a kind plug-in demo. I believe this.

C

Is sam, yep that'll? Be me uh this one should be.

D

Fairly quick because it's still fairly in rough shape, but I think it'd be kind of fun to share. um So this is a plug-in that I've been working on.

D

It's in, go not javascript, sorry wayne, um and the reason for that is because, um when you have, I guess the kind I'm importing kind here to create a cluster, and uh you can have this really nifty uh v1 alpha 4 config and the idea was that we'd have this form that would populate this object and then uh kind has this very nifty just provider create, and we can hook that into a plug-in and effectively uh have information that you would have otherwise be doing with a kind cli and do a ui so effectively.

D

This form here does nothing so far. What we actually get so far is an overview of all your client clusters, and this is really the analog of. uh I can go back to the docker psl.

D

If you kind of run docker, you can kind of see that your client clusters are really just these uh node images of a particular kubernetes version, so I'm also importing um docker to run these commands in the back, and it's looking for the version of kubernetes that you're running and the motivation here is that we'd eventually have this form, which will either look up or download the image you need, and it can show you what kind of what images are available and you can control how many workers or or nodes that you'd want in this particular cluster, and you can kind of see these configurations in the uh kind documentation.

D

So this will construct um this effectively. This uh cluster config file for you behind the scenes and uh currently right now. This does nothing because we haven't really. We haven't really implemented how to pass data in our stepper component. Yet, but let's, uh but let's say this is filled out- and uh the second step here is feature gates and feature gates are interesting.

D

I think it's also documented in here somewhere um yeah, uh you, you have a particular feature, gate name and it could be true or false, and then you would also have, uh depending on uh what kind of feature gate it is. You'd also actually have to insert uh the name of the feature, gate and other parts of the config as well, and there's a kind issue open for that.

D

um But these would be you check them off and I guess the interesting piece of this is uh what if a particular feature, gate is not enabled for a particular kubernetes version, so these some of these might actually have to be grayed out or removed entirely. So this will be a dynamic stepper, uh but let's say we actually hit submit and we create a cluster. uh This will start spinning up and ideally the vision is that we'd have this logger, that's going to be provided right. Now, it's a cluster.

D

I don't know if you can see this in the screen share right now: it's cluster provider with docker. uh Ideally this would be uh some sort of logger that can be provided, and this is something you can do. It's like cluster dot. uh I think it's what blogger provider with logger yeah, so you can provide a logger and maybe there's some way to show uh what's happening uh because right now, what's gonna happen, is it's still creating this cluster, but you can't really see it.

D

It usually takes around 30 seconds, but when it does get created, what happens is that kind will actually change my current context so, right now we can see it's still on this kind, k native cluster that I've already had running.

D

But when the cluster is successfully created, uh we should see a kind, oops, yeah kind test uh pop up and that's how we know that this uh new cluster is uh good to go and once it is, uh we can kind of see uh that if I navigate back to maybe the path here, uh this is just a clean uh kind cluster that we've just made a few seconds ago and uh other than that it's uh also got a delete. I haven't tried out the delete, yet I don't think it works.

D

If it does it, he doesn't.

D

It does not, but that's all right uh yeah. So that's pretty much all I have um if uh y'all have any ideas or feedback with uh what uh you might like to see in such a plug-in uh feel free to share. I haven't really pushed the code up for this uh mainly.

B

D

I'm still blocked on the stepper go inside and block this and kind of have the form semi working. I will push this to get up somewhere.

A

Yeah, that's pretty awesome um yeah. I think I think the two things you mentioned, which were getting the logging like if we can get that logging surfaced up, so we can actually see the steps as they're happening. um That would be. That would be pretty they'll, be pretty.

A

Great but yeah, this is really that's really cool. I think that um that'll be super useful for a lot of people. um I should I'll mention this. uh I have a kind plug-in that you might be interested in um which I will link in the chat here.

A

It is it's very, very rough, but it's also written in go. This is specific around uh the kind registry it lets you load. You know when you're dealing with kind, you can load from your local docker registry. You can load modules into this, the kubernetes server uh or into kind so that they can see it so that it can access those uh images.

A

So, if you're building custom images locally and you want to then deploy them in kind, you have to cut, you have to load them over, and uh this plugin will will do that.

A

It uses the cli to perform all those actions, but um this might be something that you might want to tie into the uh to your kind plug-in once you get a little further along um and because again I didn't really. I just like, did it and was like hey, it worked, kind of and then moved on, so it would be good to have a more like a like. Yours seems like it's.

A

It's moving the direction of being a little more robust and uh having like a tab that exposes what I've done is its own plugin, but just like having a tab like registry tab um or something something like that or another menu item in there registry. um That could be really neat too.

D

Yeah I'll definitely take a look and see if we can tie this together and yeah it's. This is I'm pretty stoked on this one.

A

Yeah, that's super useful, like yeah, I I'm excited for it to get to the point where I can use it. I want to use it.

E

Can you use it if you don't have a context set like if you have no kubernetes clusters running at all, can you get to the plug-in to be able to boost up a cluster then.

D

Yeah currently, no, because you just can't even get to that view- I mean unless you actually, that might be good. I haven't tried. um My guess is probably no, because uh you just really can't click it like. Maybe if you just hit like hit to the pad directly, but you can't really get to it from the ui.

A

Yeah yeah the way well and the way we the way we uh do deal with the routing when you don't have a config, is we load in a a temporary uh handler so that handler wouldn't even have the uh module routes on it? So I wouldn't even know where to send that traffic.

E

It might be interesting to allow a plug-in to hook it in the case when there is no config.

A

Yeah, I think that would be an interesting setup like brian, and I were talking the other day about plug-in capabilities and and being able to there's a couple things.

A

That's one where being able to say expose this plug-in when there is no config, because it does set up work, uh another one is being able to say um this plug-in uh execs so like we need like, if you think, about you're, running a plug-in in cluster plug-ins in cluster, can't just randomly exec they're, not designed like so being able to have a set of capabilities beyond our right. Now we register a plugin will register like I support this stuff, but but similar to how we right now we say I am a module.

A

It'd, be really nice to have a better way to say, I'm a module. I I do local execution. I support. I have you know, no, can no config support or whatever something like that, but in a in an intuitive way. That's not what I just said.

A

Excellent, well, I think that is all we have unless there are any. Let me check make sure. There's no questions I got added looks like we are done. Thank you, everyone for attending, and I will see you next.

A