text
stringlengths
0
1.8k
[2549.64 --> 2553.70] implementation paradigmas so when you develop
[2553.70 --> 2556.30] your delivery system you would rather split that
[2556.30 --> 2559.16] so for example if you create a script you
[2559.16 --> 2561.18] shouldn't write a built-in deploy or make
[2561.18 --> 2563.86] file target you just create two ones with
[2563.86 --> 2566.22] separate implementations and that you can
[2566.22 --> 2568.94] maintain them separately and modify and test
[2568.94 --> 2571.20] them separately if needed this is the main
[2571.20 --> 2574.86] difference if you talk about ci cds systems i
[2574.86 --> 2578.40] would rather say it's implementation detail because what we want is that
[2578.40 --> 2583.24] systems work for our use case if they work it's perfectly fine i know that
[2583.24 --> 2589.78] in a previous episode we talk about using something like get have actions for the ci part which builds
[2589.78 --> 2594.80] guess dependencies runs the tests and then something like argocd for the
[2594.80 --> 2599.64] deployment part where you have the artifacts and then argocd just reconciles
[2599.64 --> 2604.84] whatever runs in kubernetes with the artifacts that were produced by our ci system and i thought that was a good idea
[2604.86 --> 2609.34] what do you think cyril something that comes to my mind here is that
[2609.34 --> 2615.58] we are in a world where we want to automate more and more the deployment of what we produce
[2615.58 --> 2621.58] so even if we decide to use two tools or maybe to put some boundaries for security
[2621.58 --> 2629.18] constraints security of the supply chain process we still need a very automated way to trigger the deployment
[2629.18 --> 2630.02] work through the previous mission and Processes Suzuki
[2633.96 --> 2637.46] And in this sense I'm wondering if it's more alienation of tools for
[2637.46 --> 2641.90] some reasons like the best tool for the job or security but your two
[2641.90 --> 2645.62] processes remains completely connected together, maybe with a kind of github githubs approach
[2645.62 --> 2650.12] where a github manifest is sitting between the two processes.
[2650.12 --> 2653.88] Git YAML manifest is sitting between the two processes,
[2654.72 --> 2656.62] but the processes would remain
[2656.62 --> 2658.70] integrated and connected together.
[2658.88 --> 2662.02] Well, I can tell you what we changed about the whole changelog setup
[2662.02 --> 2664.94] a couple of years back, where we decoupled, we used
[2664.94 --> 2667.94] Concourse, by the way, to run the build, run the tests,
[2667.98 --> 2670.66] and even deploy. That's what we used in the past. And we used Ansible
[2670.66 --> 2674.14] in Concourse. That's what the setup was. And then I think
[2674.14 --> 2677.08] 2019, if I remember correctly, we
[2677.08 --> 2680.16] went to Manage CI. So we started using CircleCI
[2680.16 --> 2683.10] for the build, steps, build,
[2683.20 --> 2685.54] and test. And
[2685.54 --> 2688.84] it stops currently today, depending on the branch.
[2688.92 --> 2692.06] So the master branch is the one that produces a container image,
[2692.14 --> 2695.30] which gets pushed to Docker Hub. And that's where the CI part
[2695.30 --> 2698.10] stops. As for the CD part, we use
[2698.10 --> 2701.28] something called KeylessH. And we're meant to replace
[2701.28 --> 2704.28] it, but that's what we even today, we make use of KeylessH
[2704.28 --> 2707.60] to watch the image. And when there are changes to the image,
[2707.64 --> 2709.98] it will pull down the latest version automatically.
[2710.24 --> 2712.34] There's nothing to be done. And it's because you always want
[2712.34 --> 2714.40] to run the latest version. So in that
[2714.40 --> 2716.52] world, we can have multiple copies
[2716.52 --> 2718.36] of production, whatever that means.
[2718.68 --> 2720.48] And all we have to do is tell it
[2720.48 --> 2722.52] this is the artifact or artifacts
[2722.52 --> 2724.32] that we want you to run.
[2724.64 --> 2726.42] Whenever there's an update, run the latest.
[2727.08 --> 2728.38] So we decouple the
[2728.38 --> 2730.52] deployment concerns from the integration
[2730.52 --> 2732.48] concerns, and we can change
[2732.48 --> 2734.78] CI. We can produce those build artifacts
[2734.78 --> 2736.60] whichever way we want, even locally
[2736.60 --> 2738.70] if we really want to. Not a good idea,
[2738.78 --> 2739.62] but it could be done.
[2740.44 --> 2742.72] And it works. I'm not saying it's the best
[2742.72 --> 2744.34] way, but it's what works for us.
[2744.46 --> 2746.36] Yes, a good approach because CD system
[2746.36 --> 2748.58] will be eventually more complex than
[2748.58 --> 2750.10] CI, even in this case.
[2750.26 --> 2752.38] Because it's nice to say that we just download
[2752.38 --> 2754.58] our artifact, but when it comes, let's
[2754.58 --> 2756.46] say, to fall over, fall over is a must
[2756.46 --> 2758.40] for CD. Then, of course, various kinds
[2758.40 --> 2759.64] of scalability concerns.
[2759.64 --> 2762.56] Then you get a huge CD system.
[2762.86 --> 2764.22] Having proper tools for that
[2764.22 --> 2765.14] is definitely nice.
[2765.40 --> 2766.44] This is a question for you, Oleg.
[2766.74 --> 2768.40] What does your process of developing
[2768.40 --> 2770.30] a CI CD pipeline look like?
[2770.46 --> 2772.02] So in my case, I develop
[2772.02 --> 2773.04] pipelines locally.
[2773.36 --> 2774.78] I mostly use Jenkins.
[2775.12 --> 2775.68] Surprise, surprise.
[2776.14 --> 2778.20] I also use GitHub Actions quite a lot.
[2778.70 --> 2780.48] In both cases, I run pipelines
[2780.48 --> 2781.86] locally, verify them.
[2782.50 --> 2784.34] And in both cases, I try to
[2784.34 --> 2786.08] minimize the amount of code
[2786.08 --> 2788.08] and business logic that goes to my
[2788.08 --> 2789.78] user definitions, whether it's
[2789.78 --> 2791.68] YAML or whether it's Jenkins file.
[2791.92 --> 2794.22] Because I want to have a library
[2794.22 --> 2795.04] of common steps.
[2795.34 --> 2797.28] For example, if I deploy my application
[2797.28 --> 2799.12] like published Docker Hub,
[2799.30 --> 2800.60] it's just common step.
[2801.10 --> 2803.00] Or if I build a Maven project,
[2803.00 --> 2804.14] it's still a common step.
[2804.72 --> 2806.54] So what happens usually is that
[2806.54 --> 2808.08] there is a pipeline library that
[2808.08 --> 2809.46] implements these steps.
[2809.92 --> 2810.94] For these pipeline libraries,
[2811.18 --> 2812.50] especially in Jenkins, you can create
[2812.50 --> 2813.32] a test framework.