[iris] Fix scheduling loop: filter reservation jobs at SQL level

[rjpower]

Add has_reservation column to jobs table so _claim_workers_for_reservations
can filter at SQL level instead of deserializing all active job protobufs.
On production DB (1,340 active jobs, 218MB of request_proto): p50 drops
from 156ms to 0.0ms, p95 from 2,669ms to 0.1ms. Adds before/after
comparison to the benchmark script.

[claude]

Claude finished @rjpower's task in 1m 37s —— View job

---

PR Review: [iris] Fix scheduling loop: filter reservation jobs at SQL level

• Read and analyze all changed files
• Check migration correctness
• Check query and controller logic
• Post review

---

Adds a denormalized has_reservation column to the jobs table so _claim_workers_for_reservations can filter at SQL level instead of deserializing all active job protobufs. The implementation matches the stated objective and the performance claim is credible.

Findings:

• Reservation holder child job INSERT (transitions.py:667-686) omits has_reservation: The child "reservation holder" job INSERT at line 667 doesn't include has_reservation in its column list. This is correct — the holder job's LaunchJobRequest is constructed without a reservation field, so has_reservation=0 (the DEFAULT) is the right value. Just noting this for reviewers since the column is explicitly included in the parent INSERT but silently defaulted here.

• _jobs_with_reservations uses snapshot._fetchall (private method): The new function calls snapshot._fetchall() directly. This is consistent with how _tasks_with_attempts (already imported in controller.py) operates — it also builds raw SQL against snapshot internals. No issue, just noting the pattern.

• Migration backfill is idempotent: The migration checks for existing has_reservation=1 rows before backfilling, and uses IF NOT EXISTS for the index. Restart-safe.

• Partial index is well-designed: CREATE INDEX ... ON jobs(has_reservation, state) WHERE has_reservation = 1 will keep the index small since most jobs won't have reservations.

No bugs, correctness issues, or guideline violations found. Clean PR.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a18b6333ed

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".