You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
benchmarks: measure the TypeScript/Postgres baseline on a host with Docker #1132
Job 030-benchmark-rig (#1131) builds the harness and measures the LMDB backend, and the same
harness can drive the Rust Postgres backend where a Postgres exists (CI's Rust Checks declares postgres:16-alpine). What it cannot measure is the TypeScript relay: src/ on rewrite/main is 3.11.0 with 47 files diverged from origin/main, and there is no Docker daemon,
no psql and no node_modules in the factory member that runs the job.
So the baseline that the cutover actually needs a number for — "did the rewrite regress against the
implementation being replaced" — is unmeasured until someone runs it on a host with Docker and a
checkout of origin/main (or the capture commit 242a3001).
What this needs
A host with a Docker daemon and Node, or CI with a Job that checks out origin/main, installs,
runs pnpm db:migrate against a Postgres service, and runs the TS relay under the workload
030 defines.
The workload file from 030 as the input, so the two columns describe the same events and reads.
The numbers appended to docs/rewrite/benchmarks.md in the table 030 leaves for them, with the
machine, the date and the commit named.
Until then benchmarks.md says not measured for that column and why. Filed from 030's recon
(.factory/jobs/030-benchmark-rig/recon.md, "Sizing: one subsystem or two").
Activity
feat(storage): the benchmark rig, and the first LMDB recording (#1141)