keptn Developer, 25 Feb 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Keptn Community & Developer Meeting - February 25th, 2021

Description

Meeting notes: https://docs.google.com/document/d/1y7a6uaN8fwFJ7IRnvtxSfgz-OGFq6u7bKN6F7NDxKPg/edit

Learn more: https://keptn.sh
Get started with tutorials: https://tutorials.keptn.sh
Join us in Slack: https://slack.keptn.sh
Star us on Github: https://github.com/keptn/keptn
Follow us on Twitter: https://twitter.com/keptnProject

A

A

And also share my screen. First.

A

Yeah hi everyone to this uh captain dev meeting uh it's february, 25th and uh yeah. The team is working on on getting um 0.8, um finalized and out the door all right yep today we uh jurgen can't join. This is why I will kick it off and give you an intro about the meeting.

A

um First, there are no discussions or questions from the other uh meeting or on github um yeah, so we we will skip that one, the lfx mentor mentoring ship program. Here we got quite a lot of interest and also folks that want to participate in this program thanks a lot to all of them who submitted their interests and submitted a four or the.

A

um Yeah submitted the application to be part of the program. The selection of the mentees has already be conducted and assumed. They will then be get informed from the program manager and we are happy and looking forward to to work with them all right. Yeah. As I said, uh captain autodate is around the corner as we are currently talking. I think the last pull request is getting built by our ci pipeline, and when this is done, we will create a release branch, the release candidate and this one.

A

We will then test finally test out and then also wanna release it by next week, the official one. Last week I pulled together all the the issues and the features that we implemented into the release notes. I will quickly open them to show you or what you can expect from the 0.8 0.8 release and it's quite impressive, the main key announcements.

A

They are that, with the new implementation of of shipyard, um you now get the chance to really separate the the sequences that you want to execute with captain. This can be a delivery sequence or remediation sequence. They are now separated, which is pretty handy and also makes it more explicit to you uh what's going to be executed.

A

Then you can also add individual tasks to your sequence, meaning that when you want to add, for example, a migration task before you deploy, then you are now able to do so, and the shipyard controller, which is now the the core and the heart of captain, will then take over the action of sending out the proper event which will then be covered by an execution service.

A

Then it's also possible that you can define the triggers that should execute a sequence and with this capability we are also now able to have parallel stages, um because when you have, for example, two production environments that should get their deployment from a hardening.

A

These two uh production environments, listen just when the hardening stage is finished and then they get automatically uh the images already. The artifacts from the hardening stage are deployed to production.

A

Then we also streamlined all the events um yeah yeah, because we had some legacy uh code or naming in there and now the all the events they follow, the same naming scheme, which is also very useful or cool, because now you just have to know that the task starts with sh.cap.event.

A

Then there comes the name of the task that gets triggered and at the end there is an always um an an add-on. So to say that tells you whether the task was triggered started whether the start was changed or uh it finished, and this is now the pattern that we have for all of our events and yeah. That helps a lot.

A

With the ability to implement this new event thing, we are also now capable of splitting the control plane, which is kept in its core from the execution plane with where the execution service is lifted. Do, for example, the deployment, the testing or executor remediation action, and this execution plane can live on another cluster, which we now support. The the multi cluster setup, and I want to point out the last key announcement, which is the sequence screen, which is also very nice, as erin showed us last time.

A

It allows you now to filter for sequences that are executed by captain and it also splits the sequence into the stages, which is also very handy for getting a better overview. What is actually executed.

A

Where all right, thanks thanks robert quite a lot of um users, kept users were asking for uh multi-cluster support and yeah we're looking forward to to announce that and and making it now available.

A

I don't want to go through all the other issues. um I think we plan to given 0.8 um quick run through by next week, and then we also can then take a couple of those features and want to show you to them. Show you them, but when you want please scroll through this release, notes it's pretty impressive. What have been implemented and all the changes also all the changes that are now coming up.

A

All right, then, um actually kudos to christian who did a really great job on providing guidance for migrating. Your captain contrib services to the new to support the new captain. um I added here the link that leads you to the to the preview of this migration path.

A

This means, when you are responsible for or when you're responsible, captain contrib service, then please take a look at this preview which or where you can see what to change what you have to adapt and how you have which changes you have to implement, to get it to get it working with autodate.

A

All right, let me just look if we have.

A

Okay, um any questions so far.

A

When not um wanna hand over to airman aaron, do you have something to show us today.

B

um Yeah, actually, I haven't planned to present anything since uh there haven't been any um yeah uh user facing changes in the bridge. um Just bug fixes and uh yeah preparations for the release happened this week, um so yeah nothing to present from my side.

A

All right yep, then um I will hand over to christian who pulled in a couple of issues to show us.

C

Yeah, um although same as with I mean uh the release, preparations took most of the time, so what I'm presenting is rather small changes.

C

The first one is what you already mentioned: the migration path for captain services from captain 0.7 to captain 0.8, I'm just going to open the preview link. I think that's easier, um so uh there's quite a lot to go through, so we're not gonna go for all of it here, just uh the primary headers. So we are explaining a little bit.

C

What has changed in our cloud events, or at least what has changed for you as a developer, how these will be implemented in the workflow itself, by example, of a helm service which meter service and an sli provider, then we go through several changes of go utils.

C

These are really code changes. So this is where the yeah, the coding starts.

C

The distributor itself has changed a little bit, but that's a more of a manifest change in kubernetes and the rest is really just some smaller summaries, so the go utils part is where you really want to take a look. This is what you will need to change when you migrate. An old service- and there is also a youtube video coming up soon- that uh I have recorded today, where I migrated the litmus service from our captain sandbox to captain 0.8.

C

So I used my own guide to do that and I hope others can do it as well.

C

Okay, then, the next one is actually an enhancement. We have our api gateway engine x deployment and I noticed especially when upgrading or overriding the deployment it has never restarted. The api gateway engine x, which is kind of bad.

C

If you think about our nginx configuration because it has like paths in it- mostly four butted paths via a proxy, and so what happened to me is I upgraded from from zero, seven three to zero, eight zero alpha or something, and there was a path changed, and I wasn't able to access the path because the deployment didn't restart automatically and the easiest fix actually was to add an annotation with a random alphanumeric string into the template and that basically fixed it.

C

So now, whenever we reapply the helm chart for api gateway, nginx, the port is restarted.

C

I also double checked the resource requests. While I was at modifying the manifest- and I saw that the memory requirements and the cpu requirements could be slightly slightly tweaked, based on what I measured for this pot.

C

Any comments on that: okay,.

C

Cool um the last one is uh something that I kind of already um teased, so I was doing some things with the litmus service in captain sandbox. That's a really cool service by the way. So thanks for the original contribution, um so I added three pull requests. Two of them are already merged, so one of them was upgrading all the github actions and other things based on what we have in our captain service template.

C

So litmus service should be up to date now, and um this is actually something that I want to show when you create examples uh or when you onboard services with cap, please follow our guidelines that we have in our docs and implement all the values and not just in the value sample, but also in the chart. In this case, the replica count was missing in the help chart or any deployment for the example service in here which led to a problem when you want to scale up the service.

C

So yeah and last but not least, here is the pr with all the changes for captain 080. uh It looks rather big with 27 files changed, but in reality it's uh two code files, um event, handlers.go and main.go that really have changed and the rest is markdown or manifest files and some example files that have been changed. But if you're interested in the code changes, then it's really just the main.go and the event handlers that go that have changed, and I do have a question to the original author of the service.

C

Who didn't reply back to me yet? But I don't think he's in the meeting here. I'm just gonna ask it anyway, because maybe any one of you has an opinion on it.

C

So the litmus service is running or is running a litmus chaos experiment in parallel to our chemita tests or any other performance tests, and when chem meter finishes or any other external testing service finishes.

C

This is handled in this test, finish method or function. Then we need to stop the tests or the chaos experiment in particular, which is done with. uh Let me see if the code is here yeah, which is done with a cube, ctl delete of the lit bus experiment file name. So it's a very, very simple step, but it's something we need to do else. The experiment runs forever, but I'm not sure if that's the best way to actually do it, because, right now, what is happening is uh chemita service finishes by the way.

C

This is a screenshot of the latest version of captain's bridge in case you're wondering um so chem meter service finishes, but litmus service is still running. We get in. We run into this function here uh handle test finished.

C

We check that the service name is not equal litmus service in this case, so we then kill the experiment and then litmus service itself, although fails because it fails to retrieve data from the chaos engine and I'm not really wondering if that's the best way to do it, like deleting the experiment, then just saying goodbye or if in that case we should, if we do that, if we should also just say: okay, it has gracefully finished and we're done.

A

This is what I would have expected uh that at this point, um when you kill the the chaos you as a slit service, you take the decision that I'm now shutting down everything, but this is part of my job and I'm now finished in the sense that I have successfully executed my chaos.

A

I would now send back test dot finished with result pass because, okay, the behavior was then as expected,.

C

A

C

I I was thinking the same but yeah. I then I'll I'll actually add this also as a comment here and then kartik can decide which way to go eventually but yeah. The service itself is migrated and I've tested it. It worked nicely.

C

Okay, any more comments or questions regarding my changes.

C

If not, I am done and I'm handing back to you, johannes.

A

All right, um no more people who signed up in the agenda burned. Do you want to show us something today.

D

um Unfortunately, I had no time to go over the tickets, but if you want, we can take a look at hawke if something is worth to discuss. Otherwise,.

A

D

Let's have a look.

A

Maybe pick one or the other, maybe uh maybe interesting for.

D

E

E

So we are put a little bit bigger.

D

Is it okay, like this or too small.

A

For me, it works.

D

Okay: okay, um let's see what I have here on the board, um the first one auto generate json schema. I think that one was already presented last week, so actually we can put it here.

D

um This one kept update project, so that was a small bug really where, when you wanted to update the git repository upstream, then something has crashed. A small bug fix not really worth the discussion. The discussion this one extract. The rollback task is maybe something we could discuss, um let's open it or if I have it here. Yes, um as you probably know, the rollback functionality previously was kind of baked in into the.

D

Service, I think- and now with this change, we wanted to make it more explicit so that we can even model the rollback sequence or task inside the shipyard file.

D

So actually now the han service will subscribe to event to an event and then do its rollback logic. We can take a look at the examples. I already opened it a couple of minutes ago here. Yes, how does this look like? um So we have your roll back, yes, um which is now yeah. That's that's the thing. Now it's uh visible in the shipyard that the rollback can be done uh by introducing us a sequence here. It's called callback and this sequence is triggered when and will event staging delivery.

D

Finished is uh um yeah fired, um staging delivery finished. Is this one? So when the delivery sequence is finished, then this circuits seek sequence uh will be triggered uh here we have the selector on the result field, so the result must be fail, because only then we we will do a roll back, and then we have the task in this sequence, which is called rollback, and this task is exactly the task where the uh home service is now subscribed to, and it will do its logic. So it's more explicit here in the shipyard file.

D

What actually happened before it was just in the logic of the shipyard of the helm service itself. Now you can model it with a separate task here. That's maybe something worth mentioning any questions to this.

D

So maybe not all people are familiar already with this kind of uh layout in the shipyards. It's a new shipyard layout. We will, I think, hear much more about it in a maybe next week when it releases out or something like that, but that's how it will work in the next upcoming release zero to date.

D

um What else do we have kept? The support default sequence for delivery, evaluation.

D

I don't know what's that, but let's, let's take a look.

D

Oh yes, maybe also something which is worth mentioning, because um there are two commands now uh which are marked as deprecated in the upcoming release. So we still work, but they are deprecated and we will represent by these two here. So we had the previous releases captain sent event new artifact and captain send event start evil evaluation.

D

um These two are still working, of course, but you will get the message like uh this command is dedicated. Please please use one of these two uh in especially for send event new artifact. The new command will look like this captain trigger delivery.

D

um It has pretty much the same flex, um one difference, though the minus minus sequence flag is now optional and will get the default value. So if you leave it out, the default value will be delivery and you can override it by providing the flag.

D

The captain send event start evaluation command is then replaced with capital trigger evaluation.

D

And I think that's over all the flags of this of this command is are, are the same, so no functional change here. Just the command is now called differently and, as I already mentioned, both variants still work, but in a upcoming release I don't know 0.9 or whatever.

D

We will then definitely only have these two flavors and not anymore. These two but 0.8 supports both.

D

um If there's no question, we can take a look at the remaining tickets. um How does harm service behave or fall to use a managed term chart? So it's just a verification task.

D

Is this worth mentioning? I don't think so. That was just verifying that we are doing the right things when the helm upgrade does not work and it turned out that everything is okay and.

D

For those who are interested, the helm library, we are using returns an error in case the deployment whatever it is, is not marked as ready. So it really covers every possibility, uh including the image, pull error when you are putting there an image with which does not exist or something like that. We just after a period of time, fail- and we are I've- checked that we are handling this area or error correctly, and that's the case and nothing to do here really.

D

Change triggers to trigger on one shipyard: um okay, that's the wrong title, because we've said that we will use now not triggers but triggered on another one here as the keyword for triggering sequences that was changed in the documentation and this one I don't know.

D

I think that was also part of the task where we excluded the rollback.

D

Rollback sequence or task, because there we needed the functionality that we can put the condition on the activation of a sequence which is done by a selector. So you can say: please only activate this sequence here. Whenever the result field matches the string failed.

D

um I think that description here is a little bit, not.

B

D

This not up to date on must be on triggered a triggered on here, and I think failed is just not business, not failed, but just fail, but anyway um that was the that was the deal of this ticket.

D

The select implementation in the shipyard controller, okay, that was already last one.

D

I think nothing to it from my side.

D

If there's no question.

A

Perfect thanks for showing us the issues.

E

A

E

A

Are there questions from from the participants? Maybe other comment, thoughts or ideas you want to share then please feel free to do so.

B

A

Okay uh in the meanwhile, I was taking a look at our ci build job and he now finished the last build for for the audit eight release uh we triggered for on master, and I think we are now ready to create the release branch, uh the release candidate now and that's pretty exciting- to get this now done and to be very close to also get the auto date, release finished and shipped. Let's try, it live, let's try! It live. Let's go christian! You have to lead me because.

C

Yeah, if you click on the actions, react button and on the left side there is the create release, branch all right and then run workflow on the right side. And then it will ask you for the version so.

C

Zeroit0-Rc1, I guess or just rc, I don't think we will need more than one. I hope.

A

I hope so, but let's still uh use an uh or let's add a one at the end uh and we are building it from the master. Then I now click run workflow.

C

Yeah and that should hopefully uh just create a new branch, and there we go it's running, um create a new branch and um that will trigger the ci and once the ci is done, we then need to manually trigger integration tests.

C

I think it takes a bit it.

A

Takes a bit more okay: now it's not working.

A

All right yeah- this is now done automatically. Let's, let's finger, let's hope and fingers crossed to get everything built correctly. I.

C

Think it's done.

C

It looks like it's done, so if you refresh, you should see the branch and I think it should also create that pr from that branch back to master the branch is there christian can.

D

We have such a button for the contrib services too,.

A

The pr is there really cool.

C

And there's no draft release, yet there is need to run the integration tests for that so yeah, but yeah I agree band. We will need something like that for for our captaincon trip services.

B

C

Because not just that button, just the whole workflow with integration tests and uh and publishing a draft release uh christian do I have to trigger the tests now manually, uh let's check, let's go to actions and select ci and let's see if ci is running.

C

Doesn't look like it started the ci okay. I need to look into that because, in my opinion it should start the continuous integration here and then once that's done, we need to trigger the integration tests manually I'll look into that. Okay, thanks.

A

All right yeah, this was now a live demo of releasing a captain.

A

All right, cool.

A

Still, um I know, did some questions come come up from the community or from the participants.

A

Okay, if not, then um yeah, I will give you back 30 minutes thanks for joining in this week. In this captain death meeting stay tuned. Also take a look at the slack channel and uh yeah. Next week we are going to announce the official release of older date looking forward to it. All right, then see ya. Bye, bye, see you bye.

A