Cloud Foundry Community Advisory Board, 23 Sep 2017

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Cloud Foundry Community Advisory Board [September 2017]

Description

Video from the monthly Cloud Foundry Community Advisory Board (CAB) meeting. Learn more at https://www.cloudfoundry.org/.

Agenda:
* Calendar and tooling
- Now every 3rd Wed of the month (see above)
- See Zoom info above ^^^ in and the #CAB Slack channel

CFF update
* About 650+ registration. Aiming above 1000. Please register and help spread the words and join us in Basel, Switzerland
* CF Hackathon part deux! Cool prizes and code. Join us.

Projects updates
* Runtime PMC
- Diego team has a test environment running sans Consul! All good so far.
* BOSH
- Pushing teams to test and try BOSH DNS
- Various CLI v2 clean ups
- Check out new “vendor-package” command. Docs and info soon
- BPM == BOSH Process Manager with goals to make it easier to manage lifecycle of jobs and allow them to be --containerized; also simplify process management (replaces CTL scripts with YAML). Still in beta
* Extensions
- New proposals: Service Fabrik and CF Local
- CF-Abacus v1.0 released and v2 inception planned. SAP and IBM proposing to move as Core project
- CF-Extensions bot running. Please tag (add topic) “cf-extensions” to your repos to be listed
- Clean up, process updates, inquiries, and CF-Summit office hours Thursday October 12th @ 11:15a
* Community projects and presentations
- SAP’s Service Fabrik’s proposal by Ashish Jain
- Pivotal’s OSS Scheduler by Scott L’Hommedieu

A

That's what I'm saying it's not very clear, so if you can, you can use the call numbers if that helps so that way, your voice will I click will at least come clearly.

A

You might want to try that and then we can use your your data connection for sharing the screen.

A

Ok I have 8 a.m. will give people a few more minutes to do it.

B

A

I can hear you actually she's just that it's it's it's breaking, so it's gonna be difficult if you're gonna be representing so try to use, try to call the numbers.

A

It's the same, so you should you should try to yeah, you don't drop stay where you are and just disconnect your audio and then call on a phone line. So that way your voice will go to the phone line, which would be.

B

Number, oh you.

A

Are using oh, that's, gonna be the best. We can do all right. I know you calling from bingo so.

A

Do what we can?

A

Thank you all right. So a bunch of people are joining I see about 12 right now will give people a few more minutes as I said, and then we'll get started. So as I always say,.

B

What we will do it we are getting back in in a moment, we use other number. Okay.

A

Sounds good sounds good, just let me know when you're there in an hour yeah we can see okay.

A

All right, I'm, guessing Wayne, has started to recording alright excellent, thank you and we have Eric from Diego coming.

A

So we'll still give people a few more minutes and just a reminder if you're a pivotal come join us in Scorpius downstairs football I have Danny and an Eric Whitney, so we're feeling kind of low mate you easier, but she's she's at our desk.

A

So we got about 15 a couple more minutes and then we'll get started. Dk just ping me he's gonna be calling before what we called it already all right. So we got all the key people. I don't know if.

B

Dr. Mac, you are back in, can you hear it? Oh.

A

It's much better, it's much better good yeah, and then you still need to you still need to Dow with your data. So you can share your screen at some point right. Okay, and let me check.

B

What they have be happy right, so we are already dialed in using blue meeting as well. So we will share our yeah exon exon.

A

Okay, let me check to see if Scott wills here on pivotal, this is gonna, be the other presenter.

A

Dude I love his name, his French name for men and God.

A

Alright Scott is not here yet that's, okay, he's the second presentation: ah Scott is here which you'll um June. How are you.

A

Can you hear your Scott? Can you say something just to check? Had a mic check. Can you hear me yes, excellent? Thank you, Scott, okay, cool, so I think we got everybody. We have about twenty three parts. Events I said 8:05, it's 804 in my clock, but with ten seconds to go and I will get started, is a chip here or anybody from CFF like Chris I.

C

Don't see yeah.

A

I, don't see anybody here, okay, all right, so we get started. I think we have two great presentations. So let's try to give them as much time as possible, so we'll go through the normal things, which is the FF update and then kind of like the big pmc projects updates. I think people usually take notes. So if you can take notes that would be. That would be helpful. So for c FF. Let me ask again anybody from the foundation here: no okay, so I think you know I wasn't a call recently with I.

A

Think some folks and we have some data. So first thing I think I as I recalled this was on Tuesday. They said they had about 650 or so registrations, and that's pretty good, because it's only about three more week. Well we're two more two three more weeks for the summit and usually a lot of these registrations happen again, but still, if you haven't registered, you should register. There is a cop code for people that contribute to Cloud Foundry and you're, trying to reach a thousand.

A

So if you can help by pushing some advertisement to your channels, that would be very helpful again. If you go to the website, you can see the agenda. It's jam-packed tons of things happening, including again there will be a hackathon if you were not in Santa Clara. Let me tell you: it was really cool, not just because I was healthy with it, but also because of the what came out of it.

A

If you want, we can discuss that vast lack, but there were really really cool projects coming out of it, including one between pivotal and I. Guess, one of IBM Jules, where they added service to CF in a really cool way. They actually won and there were really cool stuff like, for instance, adding a repo for boss releases.

A

Excellent projects- and it was one day you're doing it again- try to put it oh and by the way, the prize, everybody that run got a drone I think it was a DJI maverick I want one, but if you want one you can put it I can't put it, but but definitely there'll be something significant that you can take off all right, I, don't know you want to add anything to kind of like updates with GFF. If you have I, don't know anything else, no.

C

I think that was a good capture of the things that were talked about in the last call. Oh.

A

Cool, so yes, so register come join us and basil. Tons of people here, I know at pivotal, will be there starting from IBM as well, and I am guessing every other company so come join us if anything for swiss chocolates, pretty good all right. So with that, let me see if there any questions about summit or anything, no.

A

Okay, cool so we'll do kind of project highlights we usually start with Yui for runtime. If she can mention anything, that's sort of interesting that people should pay attention to you're new.

C

So we in the runtime PMC there was a good amount of project lifecycle activity. Recently, The Container networking team, the the work went from incubation to active in the runtime PMC. We also have created a new project services, API team based in based in London. um That will be where, with the cap, you team, but with a focus on open-source broker, API and helping with with services. One of the first things they're taking on is persistent sharing, which I did for them to take on.

C

It's been challenging and kind of scary whenever we think about it, the additionally, we also have another team incubating in the runtime PMC the CF permissions team and they're. Looking at how to first tackle.

C

Identity provider group mappings for cloud controller roles and then at some point after that kind of custom roles, fine-grained authorization for cloud controller and potentially other components in the Cloud Foundry ecosystem to you to make you safe. So those are kind of the broad strokes for the runtime PMC.

A

Excellent, thank you Yui. You want to add anything Eric for Segel yeah.

C

For Diego we're continuing with our decoupling from dependence on console. In fact, we have an environment in CI and it's totally console running mas DNS and doing any active console service registrations or locks anymore. So that's really exciting and then one other thing we've been working with another team: that's developing a new project called a Bosch process manager, that's designed to containerize box jobs, and so we just finished some opt-in integration with their release. That allows you to run all the diego jobs in run.

C

Three containers so just moving incrementally towards better security posture for box releases. I know integration of that, as an opt-in feature is also continuing across some of the other panelists.

A

Thank you, Eric very cool. Any questions for Eric for youi know.

A

People want to get to the clocks all right, fair enough, I! Guess if you have questions you know after the fact put it, either an agenda or an the sack I'm monitoring that okay Dimitri. This is your turn. Your need.

D

Not not many new things have been happening in terms of any new tracks, but we've been continuing to finish up all of the DNS related things. As Eric just mentioned, we've been trying to push other teams to use DNS in in their environments as well and trying to push PWAs to continue adopt. Some of those changes we've been recently implementing caching in our genus server and.

D

Also trying to connect it up with all kinds of more customer healthiness information and whatnot to make it more viable to be used with software like MySQL.

D

So if anyone is interesting interested in more details about genus, do drop by in slack, and we can definitely chat more about I.

D

Think, aside from that, we've been doing all kinds of different tiny eli v2 cleanups over last month, or so so. Some of the newer features should be quite interesting. We recently added a command called vendor package that allows you to vendor the software from a different release without necessarily having to copy paste all the dependencies and try to figure out to manage that.

D

So this stuff is documented on borgia, oh, but I haven't had a chance to email out about this new command for to a large community. So that's going to be an email up coming soon. You.

D

Know, that's quite interesting, I think actually Eric didn't mention yet another thing. That's interesting, BPM boards process manager.

D

So one of the you know one of the goals for this project was to make it easier to you know manage the lifecycle of the job starts top right and we've historically had people try to write some kind of a common utilities to to manage the pitfalls and what that so, BPM, aside from you know as an implementation detail, trying to throw in your process into a run, see container at the goal of BPM also was as well to simplify and hopefully improve the UX of you.

A

Know, starting.

D

And stopping and managing the process through to wash it still works with Monnett, so the money isn't gone or anything like that, but instead of writing a CTL script, you just write a Yama file and hopefully that makes a makes it more easier to package in it's not quite done yet in just yesterday. They had the first final release, but we are definitely trying to incorporate it into a few other releases. Did he you know how it goes?

D

That's all I got for now.

A

Whoo man- that's pretty cool, so this this I guess llamo language. Is it something that the boss team is creating or is it based on some other existing I guess I.

D

Would call the language it's it's really just you know, maybe five or so keys right. They just specify how to start start the process it's based on what we did for Windows integration were quite similar, at least so you know as the executable path arguments, environment with labels and there's there's like two or three tiny other things. But let's finish it just seemed the what was commonly already being asked in other places: yeah, okay,.

A

Cool I did a him haircuts when, when I said thing with, but one more thing is I guess it simplifies because now you're not writing directly form on it or depending on sort of the underlying, and then you have a translation layer that takes that llamo and does the right thing is it? Is it also less things to worry about like it's? It's the the resulting control scripts are tiny or smaller. Well,.

D

So there is no more control script after this one right, I mean unless you're packaging. Your software to you know to do maybe super fancy things you might have to have a more complex starting bash script or whatever, but for 99% of cases you do end up without any CTL script. I see.

A

So cool, so you simplify basically your release with just a couple: Yama files versus controls, but correct all right, Thank You, DK I mean this. Isn't. This is major I'm asking on purpose, because I AK I know a Dmitry and I discussed this way back, and he showed me some example of it I'm pretty amazed that this I mean this is gonna, be super helpful to know borsch adoption, but also to sort of simplify the current releases. So people you should take a look at that. I, don't know if people have questions comments yeah.

A

So this is Ruben.

C

A

C

And Wayne I have a question, so would it be something you vent? I used a new Bosch fender functionality, for so you vendor the release of BPM in your existing release to use this or how would you consume this one.

D

Is to actually do something similar to what we do. We're planning to do is bores DNS, the one of the you know some time ago. We will introduce the functionality to have multiple runtime configs, that functionality in a Cell isn't baked right now, because there's no way to list runtime configs and things like that.

D

So there's actually a work to address that that's happening right now, but what's gonna happen is we expect that you know we're gonna end up with two runtime configs, one, let's say would be for DNS and I'll be saying Pooh genus everywhere and the other one would be put BPM everywhere and when you do put BPM everywhere, your other jobs should be able to just say you know reference ppm binary at this location where we kept jobs BPM in DB n.

D

So the expectation is that this is actually just it's like a it's almost like dynamic, linking versus static linking right, because the one of the notions is that BPM as a very stable interface- and you know, even if it is changing, it- will have a version approach to it. It is okay to actually depend on it without rendering it and given that we don't want everyone to be rendering in the same thing, you know this is the kind of approach that were going after I.

C

Should add Diego released opt-in? Does it justice seems that there's a BPM executable at a conventional location? So that's the case. If you manually deploy the BPM release and it happens to be at the expected location and Burma kept Travis ppm. That would be the case if it arrives here. Runtime.

A

All right, thank you, Eric and Dimitri, any other question Reuben or somebody else. Yeah.

B

One follow-up question:.

C

You meant, or to have mentioned, that it also allows you to run stuff in garden, run see. How is that insulting? Not the garden part just runs.

A

C

Oh just from sea all right, that's also a dependency right. I was at time version.

D

That that comes as a package from the BPM release.

A

Thank you, okay, any other question and one last question all right sweet. So let me go to extensions very quickly, so we can give a Shi Shan Scott some time to present, so we have two new proposal, one of which will be discussed today. Service fabric, the SSAT team, will be discussing this next and then Stephen the line CF local, which he presented last week, I as I mentioned to people, you should definitely check it out, really really cool.

A

So one more thing is that abacus version 1.0 was released and I want to acknowledge sa peas contribution for making this happen in particular, here still in his team android area. They they are really, you know doing a game game of job stacking up a lot of people and they also have a separate project called abacus poker. So, if you're looking to charge people for your CF insulation and you wanna track, you know what people are using and you want to do it at scale.

A

You should take a look at abacus. It's deployed right now in production, in both PSAP cloud and, of course, the IBM cloud where it started and with that I want to mention that we are proposing to move it as a core project. So I'm gonna start the process for that to happen, and then hopefully was a bullet coming with Jeff summit coming up, it probably will have will probably suggest a food after CF summit, but just be aware that this is coming.

C

As an active project max.

A

Yes, yeah. Thank you yeah couple more things. The bot that I discussed I think maybe last time it's running. So if you tag your project with CF extensions, it should find it if it doesn't find it ping me because I'm not looking at all repo all organizations that cloud foundry has repos on, but it should find a major ones.

A

But if you don't, if you tagged and you're not seeing anything ping me and then a few more things is that obviously you know so I'll give a shameless plug, I, guess and then I'll explain why I'm doing cleanup so at the summit, I have a talk, I guess sponsored by IBM. So that's why I get it I get to talk, but it's an interesting one because I, it sort of explains the the philosophy behind extension, which is let a thousand flowers bloom right.

A

So come to this talk, if you can, but the reason I'm mentioning it is because while we try to attract as many projects as we can into extensions because we believe in innovation, so if you have your innovative ideas, bring them on. We also need to realize that some of those flowers will die. So that means lean on so be aware that, as you submit to extensions and as projects are building up we'll probably look at projects that are not having a lot of traction and start cleaning them up and I think in foundry.

A

Typically, that means moving them to attic or moving them out of whatever the PMC is so I'm.

A

Looking at that, including projects that I've submitted to essentially be cleaned up, and then one more thing is that there's been a few inquiries, so Windows come in out, let you know, and so we've got a nice pipeline building up if you're interested in detailed, trusting me and then final thing is and the process update in the document, which is two pages, there is a update to the last part of the document, mainly in terms of how does a project move from incubating to become an active project. So I added a paragraph on this.

A

If you care about those kind of things, take a look at it and make sure you give me any feedback. Otherwise, by the end of the month, I'll assume that nobody commented and then we'll move it to official as part of the process. It's pretty straight forward. It just says you know: people can essentially, after incubation, ask for their project to become active as long as it's in production. So that's, that's all I have, if you have any questions that you know.

A

Okay, so I have 825 fishes here and Scott is also here. Ashish has a longer presentation, so we'll start with him, so we can get give him as much time as possible. Make sure she's part of the team and si fee that did service fabric is calling us all the way from India in Bengaluru. So sheesh, please share your screen all right. Thank you. We see your screen, so can you say something.

B

Yeah all right, thank you, dr. banks, for giving us an opportunity today to present the with fabric in this meeting today and I have time. Ashish I'm the lead for Paris fabric and I have got Debbie. Who is one.

E

Of the lead architect.

B

For Thomas, happily, all right.

B

B

All right, so so what? What is the rich fabric, so service fabric is basically a Cloud Foundry broker and developed for provisioning and management of enterprise-grade backing services. While we comply to open service broker, if they specification, we also provide additional capabilities for data operations such as backup, restore and monitoring. As of today, service fabric provides couple of ways of the weekend services. One is using docker swamp as our developer, owning services and using Bosh efficient production rate services. Yeah. We move to the next right.

B

All right, so so our vision with respect to this project is.

B

Our mission is basically to try the sewer with well-defined api's, which can be utilized by service owners to develop enterprise-grade backing services and when it comes to day to operate, we look forward to applicable framework which should provide the service owners to have the freedom to choose the bonnet and backing back up and restore some sense of their troy and, of course, finally, they should be able to deploy their solution on a multi cloud infrastructure.

B

And then probably achieve this idea is to basically conform to open service broker a pH specification and then obviously to work closely with cloud foundry community to solve. So any challenges which we see forward.

B

So what is SP is gonna bring to the table, so we we bring a solid ten full-time engineers, contributing to service fabric and also we have been running surface fabric in production since May this year, where some of our services, like both grace radius, MongoDB and RabbitMQ, utilize our service every framework, and when it comes to day to operations we and then they are well supported by day to operations for backup and restore. So we have expedite of managing and developing some of these services. Yeah.

B

Going to the available features right so currently we provide two flavors of provisioning banking services. One is using docker swarm for developer services and then again watch base services for log production-grade, antonio impression, trade, Falconeri application, and we.

A

B

Open sourced generate backup and restore library which basically abstracts functionalities like volume, creation and deletion attach detach, which can be utilized by the service owners who implement their own backup and restore strategy. As of today, we support platforms like OpenStack, AWS and mature, while service fabric conforms to open service, API, open, sorry, proper, APA reciprocation. We also have provide extension API for backup and restore so these API scum with functionalities, like start backup, abort backup and similarly for your restore operation. Why? This is important.

B

You also needs to have capability to list on all the various backups which are there in the link of this file. So for that we provide API is variants. You can retrieve all the factors which are there and, according to your choice, you can restore the backup and then we also provide a generic service instance dashboard, which is created automatically for every service, which is provision on the platform on a production rate system, unique capabilities to basically automates the backup automate your backups backups right.

B

So for that we provide a scheduler scheduler API, which enables you to take periodic backups. You can we can define you can use ground-based expressions or human readable expressions to subscribe, some of the 10 and also if you want to abort some of these strategies. That is also a tional supported by the people.

B

We also have written our logical volume driver for dr. Cottle, which is which is again as being opened publicly, which provides capabilities tools.

B

You will use to create multiple logical volumes in anything and give a party and then, when it comes to, and then when it comes to monitoring we, we generate lifecycle. Events like the bind and unbind and infer broker specific operations flying backup/restore, so these are generated as women and using animal interpretation integrate, and you can use women to listen to some of these events.

A

B

Then, according to your dashboard strategy, you can plug in like influx DB and then of all so that's about the available features as of terrain service. Happy moving to the next slide on our road man. Lately we have seen lot of load on our wash when it comes to update. So we are coming up with an implementation where we can define.

B

We can define an update window where the surface owners can schedule their updates as per they would strategies. We also want to enable.

E

B

Of other backup and restore solution, so currently we provide a python-based library and we look forward to providing some other functions as well as of today. We provide events for women and we look forward to in creating some of other monitoring and logging solutions, for example, drama tears and.

B

When it comes to operations management, it becomes important that we provide a UI based framework which can be used by operators to monitor some of the various service instances. It becomes extremely difficult for operators who know about various PF command, so something which is in a work plan, and then we also look forward to the implementation of the new feature which has been introduced by Bosch, who don't know something like route. Then some extra capabilities, which we provide like backup and restore. We look forward to providing the PFC alignment.

B

Ok now I will hand it to these. Who will take us to explain the are key, and then we followed by ugly.

C

B

This is the high level diagram of service fabrics vocals, as you see, as you see in the center of the component diagram in the service fabric broker, it is a node.js implementation of the APA specification, so the.

C

Northbound interfaces.

B

Are confirmed by the cloud coverage the UI and we and since in EPA, consume 50 LX plugins, as well as the normal and we service broker. It talks to swarm manager for instantiating Java based service instances, which is used purely for development purposes, and it goes where the boss or deploying boss a solution.

B

This board can grade the back open up information we stored on a three and six toes and that is blocked or the news today with the high level diagram, we can take a look at the deployment diagram as well for service fabric broker. I will from the heart of it is the director the board directors starts, the landscape diploma starts with ah deploy by the bot unit and the cloud quarter gets deployed, and then we have the service fabric deployment, which one which has two main parts of it.

B

So the semi-public deployment has the service local implementation, wherein you have the service rabbit. The Euler.

C

Who's, the moderator for this.

A

Somebody has some, can you please.

A

All right, thank you. Go ahead, go ish! Okay,.

B

So I was explaining the fabric deployment. It contains two main parts: one is the broker job, which has the broker the swarm manager on the scheduler co-located on the same VM and then based on the number of order containers that we want to browse. You can add the number of copper nodes and we also have on this fabric.

B

We also have a retainer MongoDB, which is used to store the schedule, information for the for the battles and other scheduled operations that get done via the service fabric process and each of the service instances themselves, which are deployed by a service fabric broker. They they can be where the fabric broker and the access to the service fabric broker for an admin can be via the jump-offs and there is a service fabric admin UI. So this is a high-level deployment view of the service fabric. So let's take a quick demo as well.

B

So as a service owner, so if somebody is a service owner and they want to bring in a service on-boarded onto the service fabric, so what are the things that is required? So there are certain responsibilities that we expect from our service owner, and then there are certain responsibilities that the service fabric does the service of steel. They need to define their service catalog, which which gets displayed on the marketplace in tier. They need to also define a partial manifest with just a job definition, so it is a minimal mannequin.

B

The minimal service manifest that we call which they need to define. It just has the jobs information now what service fabric does? Is it takes that minimal service manifest and the service catalog and at real time they get part of the they become part of the service fabric? Configuration and at runtime service fabric, generates the complete deployment manifest for the required service based on the service plans that is provided in the input and based on the service plan you put is a boss. Mr.

B

Hispanic can generate the manifest and provides that manifest for wash to do the provisioning. So let me just quickly go through the demo part of it. So I have three tiles here, so the bottom kind is I have logged into the jumbo and to quickly show you to start with one of the services.

B

This is an example service, so this is our open source, github ref, also where we are giving some examples of what.

E

B

Needs to do so I'm part of your onboarding. They need to define the catalogue by defining their service plan. There are certain information that are specific to us and we define the format of how it is supposed to be done, and once they define the service plan, they also need to provide a minimal manifest. So the minimal manifest it's a template. It's a template which is used by our the lodash template or to generate the actual manifest.

B

So you can see the actual manifest where the low we use low and they are supposed to provide the job definitions for the jobs and the properties that are required for the job status and all the other attributes like the network's definition, the resourceful definition. So this pool and the releases of all these are injected by service fabric that ran five. So now, assuming that some of the services have provided the catalog and we have our code, it misses it once it's on board.

B

We can use them in the marketplace, and this will take a sample example. I. Will provision adopted service.

B

So I'm gonna provision a booster sequel server on a doctor base, so the doctor will service creation follows the synchronous rate we're in the services created synchronously. So now I am on the double post, so I can see that there is container that is created by the basic service fabric broker. Now I could also create a boss based service plan.

B

So this is amazing, closed operation. So since it's the boss base as well and then.

B

A

B

See that this is a dashboard URL for a service and I can login to the dashboard. Url and I can see the events as it is happening.

C

B

Right now it's creating the missing VM and I can see the same thing on the boss.

B

A

C

B

See the deployment is getting created here and on the service instance dashboard.

B

You can as well monitor this here as and when the progress of the deployment is happening, so the servicing class would use the basic information about the service, and this is a generic both that is loaded for all the onboard services by the service fabric automatically in the meantime, by this by the time this gets provision I also have a sample application, which I just want to show you.

B

So this is I just downloaded this app and open. So this app is not bound to anything.

B

So it's right now not bound to anything, and what I would do is for the sake of this demo. I already had created a service instance called a demo test, so I will just bind that to it, and- and this is the app URL.

B

So you can see on the top right hand, corner. It is not bound to any services, and all the information from this is coming in from the in-memory.

C

B

So I will bind this app.

B

The service and I will restart the app.

B

So, by the time the applets restarted, let me just show you the log, so this is I am logging into the broker. Vm and I would just want show the deployment manifest back service fabric generates.

B

Ok, so yeah here is the deployment manifest.

B

Yeah, so here is the deployment manifest for the deployment that we just created and so which fabric automatically takes care of, providing all the all the other aspects of our deployment manifest. Probably most of this I want to go into the cloud particles when we integrate it into the bottle. Oh, but all of this right now is provided by service fabric, the network definition and only the job definition. Part of it is something that the subsystem makes you define.

B

Okay, so now looks like the the app is up, so we can go back to the app and the switch refers. The browser, and you can see that the app is bound to this person and I would also like to show you. The extension EPA is that this fabric has implemented having.

C

B

Its extension et yeah.

A

Let me do a quick time check we have about because we need to get to Scott. So maybe if you can wrap up in about two or three minutes: okay, yeah sure yeah, yeah.

C

B

Sure we also have extension API, which is integrated. Why are the CFC Li? So one could just go and they can do a he'll start back up and then they can say demo tape. So this was a service instance that was created some time back and it was bound to this app and one could monitor the status of this. And then this is an asynchronous process and one can take a backup and.

B

Yeah, so what did you get or the state is currently something and then once it's done, we can actually.

B

So we also have the provision for listing of the backups, and similarly, you can I mean in the interest of time. Users can also go back and restore to a particular snapshot where they mean where the backup something and these are scheduled. So now let me go back to the demo and probably efficient alright, so so this is a summary of ATP's contributions to various cloud foundry projects and.

B

So this is all we had to whatever questions yeah yeah.

A

Excellent, thank you really cool, so we have about a minute for questions. If people I don't see anything in the sack, but if you have question you want to ask live, maybe we can take a couple questions. Remember.

D

Everybody literally.

A

You have to unmute yourself. Can.

C

I ask a question sure what sure so I thought in the footnote I should introduce myself I'm from pivotal and I'm the product manager for a Bosch, backup and restore. So thank you. This looks really interesting. I wondered if there's a footnote that says it integrates with it can integrate with VBR. Could you talk about that.

B

Ok, so I for this, hoping that there could be some questions on this so who so the boss, backup and restore, and what show is fabric, backup and restore data, so boss, backup and restore it, is focused towards backup and restore of boss, deployments weather service fabric, backup and restore. It is from a service perspective and also.

E

B

Scripting framework, and so what we are looking at, is that what we have as a what we have provided as a backup and restore library that could be plugged in as one of the into one of the hooks of the bbr framework. So at this point in time, one of the things that we see as a limitation or a challenge with regards to be beer, which probably we can discuss, is you know, probably in how the backup artefacts flow back from the deployment itself.

B

So as part of the DVR, the back of artifacts goes back into the VF, whereas.

C

B

Case of service fabric, the back of our artifacts directly, gets uploaded from the deployment VM itself. So that is one point where the point of contention and if that is a result, probably VBS, then we looked at for integration.

A

Okay, cool I think you guys are both are gonna be at the CF summit. So maybe it's a good time to discuss this. The res will be there she's in Europe and I. Think a shoe shoe coming right. So hopefully you guys can chat about that. Yeah! Okay, let's see, if there's another great question, otherwise we'll pass it to Scott any other burning question: okay! Well, thank you! Ash and I know it's 9 o'clock for you.

A

So thank you for staying late and we'll see you and I guess in Basel, and we look forward to two people commenting on the proposal write. The proposal is out so please comment so if she should have to unshare your screen. So that's quite critical. Thank you very much, yeah no ball! So Scott you're new, alright, so take it over alright.

E

Thanks for having me dr. max good morning, good afternoon good evening, I am p.m. a pivotal and I was asked to come today and talk to you about a product we were releasing into the market and we'd also like to get it out into the foundation. So we want to get out in front of everyone today. So I will start this presentation, real, quick demo, so.

E

Scheduler for CF, we currently have this we're productizing this as order for PCF, so regular disclaimers about anything I say here may or may not actually happen. So I wanted to tell you a little bit about what we're doing and why we're doing it. What our vision is for this now and in the future, so we have noticed a lot of our customers. We're asking us about the ability to execute workloads.

E

I should try to keep that term workloads as things they'd like to execute what whatever those things are in the context of cloud foundry and we kept coming up with answers of you should have an external schedule or you should have an external scheduler, and we found lots of folks who are having difficulty integrate their legacy, scheduling platforms with the constructs of foundry, so we decided we wanted to put together this sort of ecosystem that would abstract jobs or tasks that could be executed here, specifically on a schedule in a in a Cloud Foundry way right, so that the primitives of cloud foundry would apply to schedule it, and that was something we were really excited about.

E

Helping our customers move legacy edge, alene platforms and not because they're necessarily bad, but because they haven't quite kept up with the move to cloud native and some of the mechanisms that we all ask folks to participating in terms of the development lifecycle. So you might see a whole set of hardware and software deployed to support scheduled loads that existed right next to all of the apps and services that were on cloud foundry.

E

So we wanted to support our customers using those best practices in terms of cloud native micro service development, life cycles in terms of CI, CD and whatnot. So that was our short-term goal. Pretty simply is to enable that longer term we have some things we'd like to target like right now.

E

The construct of this this workload is either a CF task or a HTTP call and we'd like to expand what those jobs can represent like executing scheduled service interactions, binding or unbinding, or backing up and restoring, etc, or things on the platform like adding removing routes, blue-green, deploys, etc, so that the platform itself has a scheduler as much as the app developer as a scheduler. So those are our longer-term things, we'd like to look at so in terms of what we've got right now we have a sort of productized offering scheduler for PCF.

E

We have a beta out available on pivot and we're working on a ga here very shortly, which would support what does PCF 111 and 112, which are close to the layer. 112 is close to the latest of open source CF bits, and why that's interesting and a note about this product specifically, is that we're trying to abstract away some of the changes that may happen in the underlying Cloud Foundry. So, as you can see, there, we've mentioned 111 and 112, specifically PCF.

E

Those are very different versions of the cloud controller API and while they may not change as much in the future, as the v3 endpoints start to stabilize there's still a lot of work coming in the v3 endpoints and the way you can work with, builds and staged droplets and use different versions of those and whatnot and so to sort of cushion the user from those those types of changes in the API and specifically, the way the API works. At a deeper level, we wanted to provide this abstraction of the job layer.

E

So we see people using this for things like database migrations, sending emails or reports out or doing your standard batch job processing and again. I. Think a big thing that we're trying to express to folks is that just because you've got scheduled, workloads doesn't mean you can't run them on CF so that you have data. That's in CF you're, using these data services, you've got your micro services went through testing a lot of times.

E

Folks are following a wholly different lifecycle for doing scheduled workloads where they file IT tickets and attached jars to these tickets, and then they have a whole separate team.

A

Okay, I think Scott is I hope he comes back because this is actually very very important project. I.

A

Guess if you guys are pivotal and you can ping him on a separate channel to tell them.

C

Sure I'll try, but it looks like he's working from home today. Yeah.

A

I guess I'll try to use some of this time to mention that you know so. I went over this with Scott I mean. Obviously this is already part of PCF, but this is really good. I mean this is.

D

It is a it is kind of an interesting I wasn't.

D

But what like you, know, Sheen since his uh he's back for now, you know imagine if you have a concourse right. What else one right! Why would you want this? Okay,.

A

Interesting, that's! That's! That's so Dimitri has a nice question for you, but Scott since you're back try to finish and then let's see if we can address the case, question you're.

E

One sec: let me just finish this up. Okay, so I think I caught the tail into that question. So I'll try to get to that. But so it's a service broker. It's a service. It has a service broker, exposes the service in the marketplace and it also as a developer. You have access through an API, so it has a restful api and also a CLI which executes all the same functions.

E

We it enables you to schedule the execution of Diego tasks or Cloud Foundry tasks more specifically, that are do tasks, and it also just recently we're adding outbound HTTP calls. This will allow for integration to other work, let's sort of work load managers, and it also helps to maintain status in history.

E

There's another thing: that's a little bit difficult in terms of other tools that you might see deployed onto Cloud Foundry is that they, a lot of times, lack the ability to maintain the status of the executions and the history of the executions as they occur. So the scheduler takes care of that and again is to move workload. These schedule workloads on to CF and get you into CI DD best practices with your CF workloads.

E

So you can define a job or a call. You can execute those jobs and calls as ad-hoc executions, or you can schedule them. The scheduled expression right now is a cron expression. We fully expect that folks will want to use more natural language or another type of expression, so we have left that extensible, where you can define the expression type as well as the expression right now. We only support cron expressions again I mentioned we retain the history of all those executions.

E

The nice thing about this is you get the full support of all of the CF functions of an app. So you have all the bindings of services. You have all the sort of geometry definitions of your app. You get the build pack niceties and logging functionality and whatnot. So that's a big bonus for these that you can use the same tooling that you do for your app and the same services.

E

You do for your app same, build package of your app for your tasks, so it streamlines the development, lifecycle and process and again we have this sea lion API. So it's on pivot.

E

We have a gia as a beta right now, if gia coming, it is separated in its repos into both a bas-reliefs and this tile that functions in a PCF, ops, man and then the CF CLI again is a separate component, and the great news is that we're preparing the proposal to start incubating this in the cloud foundry foundation, so that will be coming shortly.

E

So with that I, don't know how much time we have left, but I could run through a really brief demo. I.

A

Think I think Scott, since it's already in PCF, we trust you people can take a look. Let's take some questions because the meteor each question is actually quite interesting and people might have a question and we only have about two minutes. So if you do remember, dimitra's question I.

E

Think it was related to why not just use Concours.

A

Exactly pretty much that's a good summary yeah.

E

So while concourse has the ability, obviously to.

E

Schedule schedule work, execute work on a timer. We perceived that the implementation of concourse is sort of purpose-built for pipelines of executing these continuous integration and deployment, and it wasn't as close to the those things I mentioned, like services and cloud foundry, specific primitives that we didn't want to have to force the I guess the concourse purpose into the scheduler purpose and the scheduler into the concourse purpose.

E

Being one major thing, the other is that we, the future work, we're really looking to do, and one of the reasons why we're really excited to bring this to the foundation is that we see that this could be very tightly coupled to the Cloud Foundry primitives such that you can define these things in a manifest, so that you just push with the the app manifest test and the tasks and schedules are defined. All in one. The services are understood in a Cloud Foundry way, and so that was our main reason.

E

Another reason we're bringing the foundation is to hear things like this and get that kind of really broad input into the direction of how this moves. Okay,.

A

E

A

Do you have a follow up or are you okay.

D

Well, I my internet broke off fully out before I thing. I got I got the gist of it. I'm still very much unconvinced that implementing any of that would would be operationally beneficial right because concourse provides you strong primitives, for where are these things right and and of course you know when running in concourse, you could, all you know, be kicking off. You know, CF run the task right at that time, so it's not necessarily that you know I'm advocating for running the actual.

D

You know executing the actual bits in concourse, but rather having concourse as a consolidated way of you know, managing the history managing the change managing the the the trigger right and given that I guess, one of the you know strong kind of a directions that always been pushing. Everyone at this point is to actually use concourse.

D

You know, supposedly a customer is already all in in its right, so but.

A

But, on the other hand, I mean you could agree, you could agree with Scott right like if I'm an app developer, I'm not trying to use a pipeline. Why can I have a scheduler, so the the notion of a scheduler is actually quite useful. I think Scott mentioned that and you guys are both pivot. Also got me: Dmitry Dmitry meet Scott right. You can chat about this. Let's see if anybody else has a question. I'm not trying to say Dimitri's question is not good. It's just. We gotta see if other people, in the other comments.

A

Okay rerunning over I, want to thank everybody for joining and for great questions and presentations. Scott I think this is really cool, so everybody see you in Basel Switzerland, take care bye, all right, very.

B