ci: bit_pr on xlarge instead of 2xlarge

[luvkapur]

Proposed Changes

• Run the bit_pr job on xlarge instead of 2xlarge, halving its credit rate (80 to 40 credits per minute).
• Cap the Node heap at 12GB to fit the 16GB machine (was 30GB).

Validated on live runs at both extremes. A small 2-component lane used 318 credits instead of 661 and finished slightly faster (jobs 438274 vs 438241). A worst-case 98-component auto-snap cascade completed in 25 minutes with no OOM (job 438467). The build is CPU-bound on a single env, so the extra 8 cores were idle at double the rate.

Rollback: if the job ever dies with "Killed", revert to 2xlarge and restore max-old-space-size=30000.

Extracted from #10597 so it can land independently. That PR will rebase.

[qodo-free-for-open-source-projects]

PR Summary by Qodo

CI: run bit_pr on xlarge with reduced Node heap

⚙️ Configuration changes 🕐 Less than 10 minutes

AI Description

• Switch bit_pr CircleCI job from 2xlarge to xlarge to cut CI credit usage.
• Reduce Node max-old-space-size from 30GB to 12GB based on validated runs.
• Document rollback guidance if the job fails with "Killed".

Diagram

graph TD
  A[".circleci/config.yml"] --> B["bit_pr job"] --> C["CircleCI runner (xlarge)"] --> D["bit ci pr"] --> E["Node process"]
  B --> F["NODE_OPTIONS: 12GB heap"]

  subgraph Legend
    direction LR
    _cfg["Config"] ~~~ _job["CI Job"] ~~~ _rt["Runner"]
  end

Loading

High-Level Assessment

The following are alternative approaches to this PR:

1. Parameterize runner/heap by workload size

• ➕ Lets small PRs run cheaper while allowing fallback for unusually large cascades without reverting config
• ➕ Can be gated by pipeline parameter or scheduled canary runs
• ➖ More CI complexity and branching logic in config
• ➖ Harder to reason about baseline cost and performance across runs

2. Keep xlarge but add automatic retry on OOM/Killed with 2xlarge

• ➕ Maintains cost savings for the common case while self-healing rare failures
• ➕ Avoids manual revert when intermittent memory pressure occurs
• ➖ Retries can mask real regressions and increase total wall time when failures happen
• ➖ Requires careful failure detection to avoid retrying on non-resource failures

Recommendation: Current change is a good default: the job is documented as CPU-bound and has been validated on worst-case runs, so xlarge + 12GB heap is a reasonable baseline for cost reduction. If "Killed" events start appearing intermittently, consider an automatic retry/escalation path (xlarge → 2xlarge) before reverting permanently.

Files changed (1) +5 / -5

Other (1) +5 / -5

config.ymlDownsize bit_pr runner and reduce Node heap with rollback notes +5/-5

Downsize bit_pr runner and reduce Node heap with rollback notes

• Switches the bit_pr job from 2xlarge to xlarge and lowers NODE_OPTIONS max-old-space-size from 30000 to 12288 (both at job and step level). Updates comments to document validation results and a clear rollback path if the job fails with "Killed".

.circleci/config.yml

[qodo-free-for-open-source-projects]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0)

Great, no issues found!

Qodo reviewed your code and found no material issues that require review

Tip of the day

💡 Did you know, you can enable the Remediation agent and Qodo fixes findings in a dedicated fix PR

More tips ↗ | Customize Qodo ↗ | Qodo docs ↗