Skip to content

perf(infra): size the EU ingest fleet to its traffic - #1215

Merged
Makisuo merged 1 commit into
perf/ingest-self-telemetry-samplingfrom
perf/infra-eu-ingest-right-size
Oct 2, 2026
Merged

perf(infra): size the EU ingest fleet to its traffic#1215
Makisuo merged 1 commit into
perf/ingest-self-telemetry-samplingfrom
perf/infra-eu-ingest-right-size

perf(infra): size the EU ingest fleet to its traffic

1392dab
Select commit
Loading
Failed to load commit list.
Maple Review Bot / Maple / review succeeded Oct 2, 2026 in 2m 57s

Confidence 4/5 · No issues found

🟢 Confidence 4/5 · likely safe to merge
EU prd's ingest drops to one task on one host: I verified the ASG ceiling (6, from scaling.max * 2) still allows a two-host roll, but that is the choice to accept here.
quality 100/100 · no findings · tests covered · risk medium

Sizes only the EU prd ingest fleet to its traffic: one c7gd.medium with a 1024/1536 task, autoscaling 1-3, and the non-prd collector size, via a private isEuPrd(stage, region). US prd, previews and Electric are unchanged.

  • isEuPrd(stage, region) gates every EU prd branch
  • resolveIngestEc2InstanceType/resolveIngestEc2TaskSize replace the two constants and take (stage, region)
  • resolveIngestDesiredCount, resolveIngestScaling, resolveCollectorTaskSize take region; EU prd is 1 task, 1-3 scaling, 256 CPU collector
What was checked
  • US prd and previews are byte-identical: every resolver returns the old value for ("prd", "us"), pr-12, dev-alice (stage.ts:94-207, 310)
  • ASG ceiling still rolls one host: maxTasks = 3 for EU, so maxSize = 6 (alchemy.run.ts:383-389)
  • Disk math holds: 59 GB NVMe is ~55 GiB usable against a 48 GiB cap, and orphan-recovered frames also increment live_bytes (telemetry.rs:1506)

1392dab · Updated on every push. Reply "won't fix" to dismiss a finding, or mention @maple-review-bot to ask about one.

View more details on Maple Review Bot