Jenkins Cloud Native SIG, 30 Jul 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Cloud Native SIG meeting 30.07.2021

Description

We focused this Cloud Native SIG meeting on the CloudEvents plugin for Jenkins and Shruti Chaturvedi's GSoC work.

Mauricio Salantino spoke about his work on the CDF SIG Events Protocol Proof of Concept, https://github.com/cdfoundation/sig-events/blob/main/poc/README.md, with further discussion on the CloudEvents plugin.

A

If there are like some very precise, like specific questions about these topics, or should I just try to give an overview about what we did and uh the events that we have in there and uh then we can talk about like how can that be kind of, like included into the into the jenkins cloud events plugin as well?

A

I don't know kara that you might have more context.

B

No, I think that sounds great. um Thank you all for being here and really we just want to ask you more questions about the event protocol that you worked on. It's come up quite a bit in the last couple, g-stock mentoring meetings that we've had, so we really appreciate you giving up that discussion here today.

C

B

You know shruti has a lot of very good questions, so.

A

Good stuff, good stuff. So let me let me just share my screen quickly um and uh let's go here.

A

First right, like I'm guessing that you folks know kind of like this repository already like the sig events, that's kind of where we're doing the the work and what you will find here is this kind of directory called vocabulary draft, where you will find kind of like the events that we are just aiming to have, and there are like different four different categories here uh and here, just its basic description of the events and again, the idea of the specification here is not to define the cloud events per se, but kind of like the semantics of the events and the terminology around the events, because that's like usually not enough and people usually wants to see something implemented.

A

We created kind of like this poc using uh tecton and um and captain, and they are just exchanging these cloud events that are part of the cd working room right, and I think that this is kind of like this image summarize something similar to what I think that shruti is trying to achieve with the with the jenkins cloud event uh um plugin there, uh because it is related because we are using tekton pretty much in the same way as we would use like jenkins in general.

A

Right and again, the idea here is that we will have event producers and event consumers, and in this case of tecton and in the case of jenkins, kinda like they are both right, like they are producing cloud events and they are also consuming cloud events right. So we created a demo that basically is showing interoperability between tecton and captain for doing different things, and we are sending cloud events and we wanted to avoid sending cloud events from one service to the other.

A

One right like within one like tekton pipeline, sending cloud events directly to captain for a couple of reasons. But the main reason is that when you send events, you want to make sure that you use some infrastructure, that it's ready to handle events and, for example, 100, delivery and sequencing, and for that reason we uh use for the demo here, like the k native event broker, which is already you know designed, and it works with cloud events like out of the box right.

A

So it allows you to filter cloud events and it allows you to um yeah just move clouds even around between different systems, and you can just define how to so. You can subscribe to these events and send these events to whatever system you want.

A

uh I am expecting to be able to do something with jenkins, so if we have the cloud events plugin like ready at some point, we can do something pretty pretty similar with jenkins or even like add jenkins into this demonstration right. We are running two pipelines here with tecton uh and we can easily swap one of these pipelines to be run in jenkins, showing like again like now, three projects interacting uh and working together, trying to to do something right.

A

The the demo wasn't that difficult to build the difficulty here. It's always understanding each of the projects and knowing exactly what are the things that you need to change or what are the things that you need to adapt in order to consume and emit the right cloud events, and when you cannot make the tools to emit and consume the right cloud events, then you usually create like ended up, creating a translation layer right and for tecton.

A

I think that this was kind of like a translation layer that we built and for captain we need two translation, ledges, the inbound kind of plugin or service, how they call it and the outbound service as well and again this. The idea here is that if we start having kind of like uh plugins like in jenkins, we start removing those translation layers and we make like in some way we make the projects to adopt the cloud events that we are defining as kind of like the standard cloud events to interoper interoperate with other tools.

A

um If you look at a little bit more like uh native eventing, which is a project that I'm getting more deep into it like into how it works and how it was architected, um you will see that it is not difficult to use. But you again it's one of those things that you need to understand. Why you're using it and and why and in which situations right and for for kind of what purposes, if you keep it to the most basic concepts.

A

Do what you do is. Basically you install k native uh eventing in a kubernetes cluster, and then you basically create a broker, and then what you can do from your applications is send http request to that broker and forget about the cloud event. Basically just send it there, and then you said: okay, whoever who wants to consume it.

A

They need to create a trigger which is subscription just to filter the events that are going to the broker and just and just get only notified when there is a an event that it's you know for a certain type that I'm interested in right.

A

So in general, that's kind of like how people tend to think about like creative eventing, you just install it and then what you do is you create brokers, and you might be wondering like okay, but what's implementing the broker, and then you can go to k native like the documentation, and you see that they're like different implementations, so you can put kafka in there.

A

You can put ravidmq, you can use the cloud provider, specific implementation and that's definitely offering a great amount of flexibility that you can start simple, for example, in a local kubernetes cluster with like an in-memory broker, and then, if you want to make things more serious like for production environments, you swap the implementation and put a different broker without changing any of the applications and any of the producers or consumers right.

A

uh So in general I can like that's the the you know. The main objective of the project is to abstract these brokers, which are basically in charge of receiving and filtering events and sending to different systems.

A

I'm not entirely sure I mean I would love to see like a demo with the jenkins cloud event.

A

Plugin using the k native eventing stuff- uh and I can definitely guide people on on on how you can configure that and what kind of infrastructure do you need in order to run it, but I'm not entirely sure at what stage the plugin is and- uh and you know- and I would consider maybe more interesting to start working on the on the adoption of this client like new cloud events, kind of like formats defined by the group, instead of like worrying too much about the the infrastructure, and how do we connect systems?

A

Can you folks can like share a bit more about like? Where are we right now and what kind of like questions did you have about like these projects and the poc or the cloud events that we have in the group.

D

uh Yes, I will go ahead with what we are doing here right now and also thank you so much mauricio for um sharing all of this and going through this.

D

uh So in, like the last two meetings, we we were discussing about an architecture where you know and if we had like a jenkins sink, it would basically be the sort of fall would have fall.

D

Tolerance would have recent capacity and all of that stuff, and then um we were in meeting with the unsec team and andrea was there, and then we were going through the k native and the broker and trigger infrastructure and, and then the last meeting we were um talking about this and cara, and everybody else suggested that we, let's design a system similar to what's there in the poc with captain and techton, and what we're going to do for now is replace captain with tech um captain with jenkins as a jenkins, outbound, so jenkins would be sending cloud events to um the k native broker, and then we would have tacked on listening on to the events coming in from um from the broker for k native.

D

So what I have right now is actually I've been working on it and I have developed the um sort of the core components for it like separately, and I've been trying to connect it in the morning like today, but I'm having like a a bit of a trouble connecting um like the tact on event listener. So it's everything is there and it's designed, but when I'm sending anything to the event listener, it's not triggering a task run. So we were just testing with a simple test run.

D

It's not really specific of you know extracting parameters or specific information from cloud events per se, but more or less first establishing that structure of jenkins will send an event to k native broker, and then there is a trigger which is basically have the um the subscriber url for the the techdon event listener.

D

uh So that's that's! That's an issue that I've been having with k native or with um tecton, actually not triggering anything when it's receiving an event on the event listener. I can let's see if I can, I'm like, I was working a while earlier uh just having an issue spinning up the um listener. Again, it's giving me a timeout on the on the port.

A

How are you, how are you running tecton and where are you running it.

D

So it's running on um kubernetes, it's actually running on um aks, so amazon's provision kubernetes, and I like the clusters. They are of sufficient size if there was an issue with volume or anything like that, I think that wouldn't be an issue I can so. I can also share my um screen with what I had up for tact on running.

D

A

So you can create and the other question is: can you create like a task using kind of like the tecton command line, tool and seed run, for example? Yes,.

D

Yes, so let me actually go ahead and think it's prism now: okay,.

D

Okay, all right um so so here was my like my um kept on running, and so I had to change um the yaml quite a bit from. Basically it was so it's the same. It's most of it is the same. I just changed the triggers to be more specific to you know, just like simple triggers, so I just like change it for one to have everything together had like a trigger template, just a very simple of whenever it's receiving something on the um on the event listener.

D

Just maybe like echo, something you know hello, ishiti, hello, world whatever uh and and yeah, so the name space and service account is the same and then here's the event listener and I have the url for it. So, right now, a while ago it was working but there's a connection, um socket error, so I cannot exactly like trigger it, but it gets so when I'm when I'm like, when I post a curl request to it, let's see if I can see yeah. So this is a point where it was working.

D

So if I have like a post request sent to the um event listener for tacton, here's like what I'd get so this is, I also have components ready for the cloud made for the k native broker and the um the trigger with the subscriber url for tekton.

D

So this was me just like testing if it is actually receiving an event and it's actually triggering something on um like tucked on, but I don't know I'm not really sure, because I tried a lot of like changes inside of inside of my triggers or yaml to make sure that okay, I don't know what might be wrong. So I did try. You know like having simple tasks here rather than like a trigger template.

D

Instead of a task run or like putting this together, I also tried running like taskref, uh but again, I'm not sure why it's not working, but if I'm on the ui right, if I'm on the ui and if I create a task run, it has the definition here.

C

D

See it it doesn't, doesn't have it because I'm pretty sure that, like something went wrong in like an hour so I'll test it again, but before that it had like the task here, which I had defined inside of um the task graph or inside of like the template, the template binding.

D

It had that here, if I would like, go and create so it was working from the ui, but as soon as I would send a post request, nothing really would happen.

A

uh Uh-Huh, okay, okay, so there are a couple of things there. You can just check the one that is receiving the event in the cluster, so you have access to the cluster right and.

D

You have access.

A

To the pods that are running in the cluster.

D

A

D

A

So looking at the logs, there might be the first thing. The second thing is that you need to send a cloud event right. You don't need to send like only a post. I don't know if you're sending a cloud event.

D

um Yeah, I also tried that so that wasn't working either so then I I had like tried both like a simple json post and I also like tried sending a cloud events from the jenkins um event, uh plugin that we have running, so I also tried that so that wasn't working either. So, let's see where is our gent cam.

D

So at this stage we pretty much have for jenkins. We pretty much have that part of that contending cloud events to an external system and we tried playing around with sockeye earlier, so it was working. It was receiving events.

D

So, if I go on to configure.

D

This was the url for the event listener at the time, and it was working uh time when it was um like it was receiving requests and it didn't really happen there. So I did try doing this as well. This wasn't working either. All of these events are basically like cloud event, compliant events, so.

A

Where did you get that url from.

D

So right now, actually, when I'm trying it's right, okay, so I basically had like a okay. Let me go back to the url and I just copy and paste it so right now, this time I'm getting a service unavailable, and this might be like a port issue, so I might have another port running, I'm not really sure. But before all of that, when I was sending cloud events, I like basically would get the same sort of uh reply from the server which I was getting when I'm sending like a simple.

D

So this was the reply that I was getting even for cloud event, sort of like how to win compliant event and also for like the post event, but it wasn't like triggering a task run.

A

Yeah, because, in order to trigger so like that's the other thing like, how did you install picked on there in the cluster? Did you follow kind of like the poc instructions.

D

Yeah so like first like going on to tact on pipeline uh okay, so you know like capcom pipeline trigger. I have the dashboard on um then applying. Obviously, what was inside of you know like service rule and bindings, and all of that stuff I mean I do have the controller but like again as we're going to we're not going to need that here. So we can move.

C

D

From that, because we don't need that here and then in place of you know, like um obviously like install the whatever resources we are going to need, but replaced the triggers with just specifically um relating to like a very simple trigger of an event or a very simple trigger of a task run.

A

Can you, but that's the thing that you maybe can do it's like list the triggers with cube, ctl right and see if they are like applied and they do not have any errors in the in the resources? That's something that I will try first, and also. I would check that the broker is is running. For example,.

D

um Yeah, I think I I might have I might have like tried looking at um like if the resources inside were running specifically like even when I had all of those things you know like. I would have tasks mentioned here when it was like a different format of a task rather than like a task spec inside of trigger template. So I had the task mentioned here.

D

I had the event listener trigger binding triggers, but it just I'm not really sure why that was happening, but I think what I might do is set it up again because it's like it is um an aws like eks running cluster, and there are a couple of other things which are running on it already. So I might try running it on another cluster because it could be an issue of I'm not sure but like cross port binding or something so okay. I might try doing that, but.

A

In general, if you have like the components, if you have the tecton component that it's waiting for cloud events, it should work right like that, shouldn't be a problem that you should try to solve and if it's not working, of course, we can reach to andrea, or I can try to help debugging in general.

A

If I can access the cluster, if I can take a look at what's running in the cluster, it's way much easier than looking at the ui, because the ui will not tell you, I mean if something is failing it, you are not going to see there like the entry right right.

A

So it might be interesting to see that and also see if the rocker is running, for example, that that might be the the next step to check, and um if you have the broker running, you can always use you remember that in in k, if you have the broker running, you can send events, and then you you can create a trigger for sockeye right. So in that way we usually debug kinda like these applications.

A

If we don't know if the event is arriving to the broker, then first we send can like a request and we expect to get, for example, a 200 in response, that's something I would do and then, if I do not know, if a trigger or like a subscription is working, I will create one to sockeye and I would just check if in soca that's arriving.

A

You are thinking that you're trying the most complex use case, which is a tecton trigger, which is going to create a tecton object based on an event. So I will just try to split up the problem on checking that the cloud events are being moved around correctly and then we can basically just create the task definition and trigger it manually, and then we just trigger it by the cloud event.

D

Yeah, I think the problem is with time with like techdon communication to tecton, specifically because I also have sockeye on the k like the cluster for the kubernetes, and then I did try first like directly um triggering from jenkins to sockeye. Then we get we're getting events there um and I was also like doing like a cross, maybe having a middleware, which would rock events to sockeye.

D

um Maybe like one thing we can try doing, is shifting um like the k native broker in between jenkins and sockeye, to make sure that that part is working. But but even if, even if we take away k native for now, I'm not really sure why the tecton specifically is not working because, as you said, you know like triggered a definition manually and then try and triggering that from either like a like a post request or a cloud event. I did try it doing that as well like I did.

D

Try first, triggering it manually both on through the command line, then also went on to the ui and did that there and then tried doing like a post from jenkins cloud events, making sure that you know it's receiving something but yeah that didn't work. I'm not really again, not really sure, but I feel that it probably is might be some issue with the the trigger style yaml, because there are different, like alpha version, beta version and inside of that too. There's like difference in how they're defining their um like yaml definitions.

D

C

D

Lot of uh like a lot.

A

Yeah, I feel that you are in the on the right track. You just need to spend more time just figuring out the details in tecton. You will see that there are like a lot of like namespaces filtering and you have like the event listener.

A

That needs to be configured correctly, plus also the triggers right, like you need to make sure that the triggers and the event listener are in the same name, space they are pointing to the right place and that you have kind of, like you know everything in place to make sure that the things are connecting that's.

A

Why, like looking at the resources, for example and looking at the trigger, for example, trigger status and to see that's okay and also looking at the event listener status and making sure that that's up and running and in the right name space, I think that's super important. So again, it's like checking your configuration, and maybe I think that's something that will definitely help is creating something like uh what andrea created like that uh script. That basically does install all these things, uh maybe in a kind cluster right like you're running on aws.

A

That's a pretty complicated environment on its own right, so maybe what you can do is just put jenkins. You know just start jenkins in a container with the with the cloud events plugin.

A

You already have can like that script, that you know bootstrap, tecton and uh and kennedy broker in a kind cluster, so this will kind of enable you to just do it faster and like just to reproduce that very complex environment just with the single script, so it might be kind of like one idea to to look into it. If you are in windows, that's a different thing, that's a more complicated stuff, but it can be done in the same way.

A

Right like you, just need to create like a script from from scratch, but I guess that creating kind of like that sense of oh yeah, I can reproduce this in a local environment, or I can reproduce this in aws will help you a lot to just to avoid making mistakes of namespaces and stuff like that, all right.

D

Yeah um yeah, that's a good idea and I might go on to like an ec2 instance or something because I don't have like linux on my machine or a vm or somewhere, because I feel like it might be easier.

C

For windows and you can spin up linux containers on windows.

A

C

uh To start off, I think you should just run tecton by itself and play around with eventlessness without doing anything with the poc in the beginning, because I have a feeling because of the poc some things here and there might be confusing right now, so just kind of like get started with you know, running a simple task run and then turning uh even an event listener which triggers or triggers something like a basic task from. um I think you should start with that.

C

If you're on the tech town, side and like rest of the stuff, I think it will just fall into place after that.

D

Yeah, that's like a fair point and um again like the thing with that. On that I feel it's probably. The task run definition itself rather than like how the event listener is configured, because the things with event listeners which could be wrong is yes, you know. Maybe the wrong port has been exposed for the url that I'm using or it's in a different name space or something. So I did check all of those things and it is receiving the like the request and it like it doesn't sort of.

D

You know like give me an error right away, or it doesn't give me an error in general, even when it was running earlier. It's just the the triggering part of the task run from the event listener, uh so, which makes me wonder if the event like the task run definition itself, which I have might might be incorrect somewhere.

C

So so the thing is, uh I think there might be ui bug with the task run and the tasks the task stuff, because in a task run you can even uh define the task inside the task run or reference it. So.

A

C

Might be an issue there with the ui so try using the cli for that, because it will become more clear that way like what exactly you're doing so whenever you do like tkn uh task run ls or something like you can see, the status of the task runs or even the task has been triggered or not. So I think that that's probably like probably move to the cli for now, because the dashboard um yeah, I think I think just uh using cli, is more is kind of better. At this point.

D

Right so um I like, when you say that the dashboard might not reflect, do you mean that it does like there are bugs with whenever we are trying to push it from external systems or they're, just bugs.

C

Just uh just kind of like formatting stuff, I mean I'm not a ui guy. As such, I prefer to use cli and in that way uh like right now I saw the ui I couldn't see. I could probably only see task reference over there, so I I couldn't see a place where you could mention the task in the task run like the whole of a task.

C

D

C

I caught my attention and uh I think what you're doing is you're, giving a task with a task from the task definition inside or is it referencing a task which is already out there.

D

So I did three different things, uh so the first thing was having the ta, so I'm basically first is like using a trigger template, and inside of that we have like the one of the resource templates. The task run, the one thing. The first thing was like you: it's specifying a task ref and then having a task outside of it.

D

So, as I was saying earlier, when I had that then over the ui, it would specify like when I would click on um okay trigger a task from the ui just to make sure that you know that it's there actually. So I would go on to task um then go on to like task, then define that particular task which I had to find in the yaml. So it was present.

D

The other thing that I tried doing was inside, of the um like task run, have task spec, which is just defining all of the steps inside of inside, of the um the task run rather than like outside of it, and I also had something similar to that.

D

What the poc has, which is just defining like a task outside of like a trigger template and that's just sort of the the the trigger, so they I don't think that has the trigger template, but it's more just the trigger, which has the pipeline run so something similar to that and yeah the it like when I was on the ui was working. So I did try running that, like steps from the task run, going on creating a task inside of the inside from the definition in the in the yaml file.

D

So that worked as well, but it was just.

C

I have a question so did these tasks when you run them individually without the trigger template, did they run properly.

D

They ran even inside of the trigger template as well uh when I had outside of them. They ran when I had inside. Looking like something like, I can share my screen. Let me see um like basically inside of you have the spec. You have resource template, kind task, run and then task ref, and then it's defining a reference to that particular task, and then task is also another sort of crd inside of that same yaml file.

D

Oh, maybe I need to pull the components out into different yaml files, because right now my trigger template, my um like event listener and also my like the template, binding or the trigger binding they're in the same file. I don't know why that could be a word. It's like maybe worth trying, if that can be an issue with different, like sort of versions.

D

So maybe I'm not sure.

C

Everything should be v, one beta one.

A

Yeah, that's the other thing like uh task. References are something like it's in alpha.

A

I guess so that's why, in like in the plc we're enabling that, uh like in tecton, specifically with the flag but uh yeah anyways, I think that you're now like discussing the specifics- and I would definitely again like suggest the same as without like simplify the setup like if you're running in aws, that's complicated for us already just to keep track and just to help you to dividing that, and also like from the ui perspective, is that sometimes the ui will not show you things that are like available in the resources themselves.

A

So you definitely need to get used to to to do that and use kind of like the cube ctl tool in order to see the details and figure out. What's failing um but again, if you have any of those issues, just just let's create their threading slack and uh we can definitely help you or at least try to guide you. If you, you know, if you get stuck for for more than a couple of days,.

D

Yes, yes, yes, that sounds good, because definitely this is not creating, so that that sounds good. Thank you.

D

Well, like one another question is sort of more specific to the entire poc.

C

D

Would be enabling that sort of infrastructure where um agnostic, sync and source binding can work? uh So you know like how you mentioned with captain that you had to create systems as a translation there for moving like cloud events between different systems. So uh how do you see? How do you see that agnostic binding with maybe when we have jenkins- and we really want to go into um doing a lot of filtering and also triggering based on specific parameters and stuff?

D

C

Like mauricio, this is in relation to like filtering in the uh like filtering the cloud events as well, and uh we were talking about the other day and uh you're thinking about doing a matching of cloud events. uh Cloud event like like we give a word and it kind of finds, a substring in the uh cloud and through a match, filter or something along those lines, something that there is in soccer and something similar. So uh in tecton triggers they have a cl interceptor, the common expression, language interceptor.

C

So do you have an idea? We could do something similar in java in uh this. In this cloud event, kind of scenario.

A

Okay, so in general, like the infrastructure, will allow you to filter events right uh and tecton, both like the filters there and sockeye, also like in the ui, but usually in order to filter things. Yeah I mean as soon as you can just parse the event. Then you can do whatever you want with it right.

A

So I'm wondering kind of like. What's the specific question in there, you have certain filters in canony and certain filters in in tecton right and then you can write your own filters in the in the cloud events plugin in jenkins in java, because you can parse the event and then just filter in any way. You want.

C

I would say, but the printer is something that the user will give in the configuration so like. Is there something some library or something we can use for filtering these events, because uh in the cl interceptor for tech tom, uh you can basically do something like uh header dot.

C

uh I don't know like header dot, something like header dot. uh Artifact name is equal to something something like I don't know, I'm not being very articulate here. I'm not articulating well here, but uh you kind of understand what I mean here.

A

I think that I do, but I don't know if there is any specific library to do that right like I don't I don't know if there is like any helper that will allow you to do that in java itself.

A

Besides, like the cloud events sdk that I think that you're not using right well you're using the the java sdk right, yeah yeah yeah. No, I don't think that there is anything in there to like easily filter cloud events by just sending kind of like a regular expression or something like that. I think that you would just need to do that in java. For now it might be kind of like a good feature request there for the sdk, but.

C

uh Like it doesn't have to be cloud events as such, because at the end of the day uh we can read these cloud events as json object right so.

A

C

Something that uh something to pass json objects and the user can give like a string uh which will uh help us parse. It yep.

D

Yeah and like also like super complicated json with json arrays inside of objects and other json objects inside of the array. It usually, if you know event body, can be complicated if you're talking about cloud events, but also in general, just json.

D

So if something was to exist like that, I feel it would be easier for even the user to mention, but definitely for us to also um get that information parsed from the event body. I think that's the more I'm sort of tricky part, uh because the headers are straightforward.

A

I wouldn't worry too much about like going and passing the body, because no framework is providing those filters. I think that if you want to really pass the body, then you just need to use something like you know, like a json framework to just go and parse using the path, basically so yeah. I I wouldn't worry too much about that for now, because I know that that sounds like that's something that people will use and probably they will at some point they will require that, but no framework is actually providing that for now.

A

So I made my push to kind of discover that from the demos- and I know that sometimes we are filtering based on the on the metadata right, like the cloud event type or the source or or.

C

A

Metadata level artifacts, but not going into the body, and I'm just filtering there, because that's usually that's usually a much more complicated thing to do.

A

I need no, it will not work for all the transports in cloud event and making it you know just making that you're going to support that, for all the transport makes it way much more.

C

Complicated which is coming in from different sources is going to be very different and the uh structure probably isn't going to be the same except the actual cloud administration.

A

Exactly folks, I need to drop to another meeting in in a second. So, um but definitely I do see that there is a lot of progress in here and now you are just trying to tackle much more complicated issues and I think that's great so feel free to open a thread there and just mention in slack, and I will just try to help uh to unblock if, if you keep hitting kind of like these issues with galaxy installation or just even sending cloud events to tecton that should work, it shouldn't be such a problem.

D

Thank you so much mauricio and everyone else for the time and appreciate it.

B

D

A

All being here, thank you. Thank you, claire for organizing and uh see you soon pick me there in slack. If you need me, please it's okay, it's okay! Nothing!.

C

I'm doing nothing! No! That's good! That's good! um Yeah! Well! Thank you! So much guys I'll see you outside.

E

A