CI map · public data

177m 51smedian of 121 sampled runs, up to 6 jobs at once · measured
65m 33swith Boat, up to 16 jobs at once · estimated

apache/kafka

Today the median run takes 177m 51s; resharded on Boat the plan estimates 65m 33s, 63% less wall clock. Estimated, not measured.

Today measured177m 51smedian of 121 runs · bars from run #44213233047 · up to 6 jobs at once
  1. Configure Workflow3 s
  2. Compile and Check (Merge Ref)15m 30s
  3. Load Test Catalog18 s
  4. JUnit tests Java 25 (flaky)6m 58s
  5. JUnit tests Java 17139m 59s
  6. JUnit tests Java 25131m 30s
  7. JUnit tests Java 17 (flaky)11m 32s
  8. API Checker Java 2538 s
  9. JUnit tests Java 17 (new)9m 52s
  10. JUnit tests Java 25 (new)10m 11s
  11. CI checks completed4 s
  12. Collate Test Catalog18 s
Recommended with Boat estimated65m 33swall clock · up to 16 jobs at once
  1. Configure Workflow6 s
  2. Compile and Check (Merge Ref)48 s
  3. Load Test Catalog17 s
  4. JUnit tests Java 25 (flaky)13m 54s
  5. JUnit tests Java 17 ×648m 32s
  6. JUnit tests Java 25 ×645m 37s
  7. JUnit tests Java 17 (flaky)23m 11s
  8. API Checker Java 251m 05s
  9. JUnit tests Java 17 (new)19m 54s
  10. JUnit tests Java 25 (new)20m 27s
  11. CI checks completed8 s
  12. Collate Test Catalog31 s

Times: medians of 121 sampled runs; bars: one real pull request run (#44213233047) from public GitHub Actions data; recommended plan estimated. Analyzed 2026-11-20.

More detail

What to change

Recommended: 19 machines, $0.566 per run estimated The recommended plan keeps the 8 jobs of the workflow, starts each from a warm snapshot and splits the test work of the slowest job over more machines: 19 machines against 14 today.

  • Fork from a warm snapshot instead of repeating setupestimated

    Each job on the critical path spends its first steps on checkout, tool setup and installs. A boat.dev sandbox forks from a snapshot that already holds that state, so the job starts at the first test step.

    Saves 22 s per run
  • Balance shards by measured test timeestimated

    A sharded job finishes when its slowest shard does. Splitting tests by recorded duration brings the slowest shard down to the mean shard time.

    Saves 1m 21s per run
  • Run the critical-path tests on 12 machinesestimated

    Sandboxes are billed per second, so the same test work spread over more machines costs about the same and finishes sooner. Only per-shard setup stays constant.

    Saves 111m 56s per run
  • Skip jobs the change cannot affectestimated

    Path filters or a changed-files step let documentation-only and single-package changes skip jobs that cannot fail because of them.

    68.3 runner hours a day
  • Cache dependencies between runsestimated

    The workflow reinstalls dependencies on every job. Restoring them from a cache removes part of the setup time of every job.

    7.1 runner hours a day
  • Retry failed tests in parallel inside the same runestimated

    A rerun repeats the whole workflow after someone notices the failure. Retrying only the failed tests on a fresh fork while the run is still going removes that second cycle.

    Saves 56m 01s per run
  • Test merges speculatively before they landestimated

    A red default branch costs at least one more full CI cycle to fix or revert. Testing each merge result on a fork before it lands keeps those failures off the default branch.

    Saves 8m 28s per run

How this CI evolved

Daily wall time of successful runs, p50 and p95, 2026-11-17 to 2026-11-20.

2026-11-172026-11-20

p50 p95

What's next: the same climb, automatic

apache/kafka can climb the same hill. Boat's agent will run that loop for you: try a plan, measure it, keep what is faster, try again. This is not live yet; when measured rounds exist, they will appear on this page.

See your own CI