text
stringlengths
0
1.8k
[312.40 --> 316.88] Yeah. So, here we have initiated an effort to provide visibility
[316.88 --> 319.46] in the execution of the jobs themselves,
[319.98 --> 326.46] where we were able to break down the duration of jobs on pipelines
[326.46 --> 328.70] in the different steps of these pipelines.
[328.84 --> 334.20] And also, we were able to track the time spent to allocate build agents.
[334.82 --> 340.38] But then, we didn't have detailed visibility in the steps to allocate build agents.
[340.38 --> 345.82] And so, we also had limited visibility to explain what kind of problems
[345.82 --> 348.70] could have been happening when allocating a build agent,
[349.20 --> 351.46] like cloud resources being unavailable,
[351.96 --> 356.94] or maybe the Docker image you want to use being unavailable or broken, and so on.
[357.16 --> 360.12] And this was an important focus of AkiHero,
[360.46 --> 363.24] which was to complement the existing traces we had,
[363.32 --> 366.60] the existing visibility we had on the CD pipelines,
[366.60 --> 371.60] detailing the agent allocation, which is on the agent communication,
[371.80 --> 374.84] which are some fragile areas.
[375.26 --> 375.72] That's a good summary.
[375.82 --> 377.56] Okay. So, the talk is available online.
[377.62 --> 378.46] We can go and watch it.
[378.64 --> 381.26] I haven't watched it yet, but I will do right after this,
[381.38 --> 383.68] because that's basically what started this conversation.
[384.18 --> 388.12] And that made me actually think specifically about open telemetry
[388.12 --> 389.74] in our CI-CD systems,
[390.28 --> 393.38] and how open telemetry is like a nice unifier
[393.38 --> 396.18] of all different CI and CDs that we have,
[396.18 --> 400.78] because sometimes people recommend that CI is split from CD,
[401.28 --> 403.26] but you still need to understand the unit as a whole.
[403.62 --> 406.98] And then what happens when you switch between CDs or CI-CD systems?
[407.26 --> 410.06] One day you use one, and six months later you switch.
[410.48 --> 411.86] Do you lose all that visibility?
[412.40 --> 414.86] Because the things that happen in your CI-CD,
[415.12 --> 416.82] they kind of tend to stay the same.
[416.96 --> 419.84] I mean, they may expand in the future and become more sophisticated,
[420.26 --> 422.14] but the building blocks tend to be the same.
[422.14 --> 426.16] So, with this context, why would you say that it's important
[426.16 --> 429.34] that we use open telemetry in our CI-CD systems?
[429.68 --> 430.52] Oleg, do you want to go first?
[430.64 --> 430.78] Yeah.
[431.04 --> 435.12] So, first of all, I would rather disagree with CI and CD statement.
[435.40 --> 437.00] It's a subject for Holy War.
[437.24 --> 440.54] Personally, I use quite old-style automation term,
[440.54 --> 444.46] because CI-CD is a methodology, it might be culture,
[444.80 --> 446.88] but when it comes to automation to tools,
[447.48 --> 450.72] then actually CI and CD borders are quite blurry,
[450.96 --> 452.94] and there are many other use cases, for example,
[453.04 --> 456.36] for operations, for organization automation.
[456.68 --> 457.92] All of that needs accessibility
[457.92 --> 461.24] if you want to have your software delivery in place.
[461.38 --> 462.40] It's not just CI-CD.
[462.40 --> 466.42] And this is exactly where we can talk about open telemetry
[466.42 --> 467.62] and other open standards,
[467.76 --> 470.52] because if any system independently
[470.52 --> 473.12] creates its own monitoring and observability,
[473.42 --> 474.56] you basically get lost.
[474.98 --> 477.54] So, when we talk about modern cloud native deployment
[477.54 --> 478.46] with Kubernetes,
[479.06 --> 481.88] you usually build your CI or CD system
[481.88 --> 483.76] from dozens of different tools.
[484.30 --> 486.48] Each of them may have different applications,
[486.76 --> 488.34] them may have different interfaces,
[488.66 --> 490.10] and then basically you end up
[490.10 --> 492.38] with just trying to understand what happens.
[492.40 --> 495.42] So, similarly to why Jaeger was introduced
[495.42 --> 496.88] for cloud native applications,
[497.40 --> 500.16] we need the same for CI, CD,
[500.56 --> 502.36] and automation in the cloud,
[502.44 --> 504.80] because we also need to glue information
[504.80 --> 506.88] from these tools on multiple levels.
[507.26 --> 508.74] So, it might be a CI server,
[508.92 --> 509.76] it might be agent,
[510.00 --> 512.10] it might be just a build tool like Maven,
[512.64 --> 514.26] but we need all this information
[514.26 --> 516.26] to understand how is our pipeline going.
[516.92 --> 519.00] And now it's also important for audit,
[519.30 --> 521.12] for supply chain security,
[521.12 --> 523.52] and many other buzzwords that are emerging.
[524.04 --> 526.18] But overall, you need data
[526.18 --> 527.98] to verify what happens.
[528.50 --> 531.20] And open telemetry is one of great opportunities
[531.20 --> 533.88] to provide this data across the system.
[534.12 --> 535.56] You said there's something really interesting
[535.56 --> 537.52] about you disagreeing that CI and CD
[537.52 --> 539.10] should be two separate systems.
[539.34 --> 541.16] And I will want to come back to that.
[541.36 --> 542.38] So, that's really important.
[542.60 --> 543.86] Like, I've taken a mental note.
[544.36 --> 545.96] But Cyril, why do you think
[545.96 --> 547.62] that open telemetry is important
[547.62 --> 549.40] for CI and CD systems?
[549.40 --> 551.04] I will break down the point
[551.04 --> 552.58] in two different themes.
[552.72 --> 553.70] The first theme is,
[554.20 --> 554.88] as you have said,
[554.92 --> 556.12] there is a lot of visibility
[556.12 --> 558.82] in having an end-to-end view
[558.82 --> 562.14] of the execution of the CI and CD processes,
[562.50 --> 564.38] where distributed traces
[564.38 --> 566.12] is very valuable.
[566.12 --> 569.16] And we see that distributed traces
[569.16 --> 571.14] is a very good data structure
[571.14 --> 573.38] to model the execution