|
| 1 | +--- |
| 2 | +title: Week 12 |
| 3 | +author: Muhammad Salman |
| 4 | +author_url: https://github.com/SalmanDeveloperz |
| 5 | +tags: [gsoc25, scheduler, infrastructure, devops] |
| 6 | +--- |
| 7 | + |
| 8 | +<!-- |
| 9 | +SPDX-License-Identifier: CC-BY-SA-4.0 |
| 10 | +
|
| 11 | +SPDX-FileCopyright Text: 2025 Muhammad Salman <[email protected]> |
| 12 | +--> |
| 13 | + |
| 14 | +# Week 12 |
| 15 | + |
| 16 | +*(August 19, 2025 – August 25, 2025)* |
| 17 | + |
| 18 | +## Scheduler Agent Work |
| 19 | + |
| 20 | +This week, despite exploring multiple approaches, the scheduler is **still not fully functioning as expected**. This has been the main challenge and blocker in pushing other dependent services forward. |
| 21 | + |
| 22 | +### Key Challenges Faced: |
| 23 | +- Path mismatches related to **`fo_scheduler` install paths**, which created uncertainty in the setup. |
| 24 | +- Deciding between continuing with the **existing scheduler implementation** or starting **fresh from scratch** to avoid legacy issues. |
| 25 | +- Lack of clarity on **evaluation metrics**, e.g., whether to focus on *job queue performance*, *agent execution times* or *resource usage* in a containerized setup. |
| 26 | + |
| 27 | +At this point, I am leaning towards **rebuilding the scheduler from scratch** as it seems like the more sustainable path forward, but I am awaiting mentor guidance to ensure alignment. |
| 28 | + |
| 29 | +Overall, while the scheduler has not yet reached a running state, the time invested has helped surface **important blockers and open questions** that must be addressed before progress can be made. |
| 30 | + |
| 31 | +## Open Questions / Confusions |
| 32 | + |
| 33 | +During my work this week, I noted down a couple of key confusions that need mentor input: |
| 34 | + |
| 35 | +> 📌 Based on my research in the wiki and older methods for scheduler, what is the correct handling of **install path mismatches** for **`fo_scheduler`**? |
| 36 | +
|
| 37 | +> 📌 What **metrics** should we prioritize (e.g., job queue performance, task throughput, resource usage) to evaluate the scheduler in a containerized microservices setup? |
| 38 | +
|
| 39 | +Resolving these will help me focus my efforts in the upcoming weeks and avoid wasting time in the wrong direction. |
| 40 | + |
| 41 | +## Kustomization & Infrastructure |
| 42 | + |
| 43 | +While the bulk of my time was on the scheduler, I continued revisiting my **Kustomization setup** from Week 11. This structure will be useful once the scheduler stabilizes, as it ensures we can maintain both **dev** and **prod** overlays with cleaner configurations. |
| 44 | + |
| 45 | +I also reviewed some Dockerfiles, but this work will be resumed in later weeks once the scheduler issue is clarified. |
| 46 | + |
| 47 | +## Collaboration and Mentorship |
| 48 | + |
| 49 | +- I actively discussed progress and blockers with [Avinal](https://github.com/avinal). We explored different approaches for the scheduler (**Reuse vs. Rebuild**). The consensus so far is to carefully evaluate before committing to a rebuild. |
| 50 | +- [Omar](https://github.com/OmarAbdelSamea) is still pending to respond, but his past GSoC experience could bring valuable input. |
| 51 | + |
| 52 | +## Meeting 1 |
| 53 | + |
| 54 | +*(August 20, 2025)* |
| 55 | + |
| 56 | +The call did not take place due to the unavailability of [Avinal](https://github.com/avinal). |
| 57 | + |
| 58 | +## Meeting 2 |
| 59 | + |
| 60 | +*(August 21, 2025)* |
| 61 | + |
| 62 | +In the weekly community meeting call, I shared my updates: |
| 63 | + |
| 64 | +- Tried different approaches but it is still not working as expected |
| 65 | +- Dedicated time to make the **Scheduler pod** run |
| 66 | +- Raised confusions regarding **install paths** and **evaluation metrics** |
| 67 | +- Suggested that a **fresh rebuild of the scheduler** might be more reasonable moving forward |
| 68 | +- Highlighted that I am still seeking guidance to refine my direction |
| 69 | + |
| 70 | +I concluded my update by handing over to [Avinal](https://github.com/avinal) for further discussion. |
| 71 | + |
0 commit comments