Daan MeijerandClaude Opus 5.5
821de668d0
Waiting files counted as work, so the merge loop re-read them for its whole 30 s budget every minute: Postgres used ~a core even without workers. Now ~4% at rest, and storing a boneless range costs Postgres ~0.7–1.9 s of CPU (it was 33 s with the loop running). Chunk payloads are stored EXTERNAL (zstd already; no pglz attempt). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
37 lines
1.4 KiB
TypeScript
37 lines
1.4 KiB
TypeScript
import { BaseSchema } from '@adonisjs/lucid/schema'
|
|
import env from '#start/env'
|
|
|
|
/**
|
|
* The chunk store (REBUILD.md, part 3; decision a): one write-once row per
|
|
* release per batch, its files encoded by lib/chunks. Partitioned by month
|
|
* of writing, in CHUNK_TABLESPACE (the HDD) when set; ChunkPartitions keeps
|
|
* the next months' partitions ready. A re-run of a range replaces its chunks
|
|
* (by range_id).
|
|
*/
|
|
export default class extends BaseSchema {
|
|
async up() {
|
|
const tablespace = env.get('CHUNK_TABLESPACE')
|
|
const where = tablespace ? ` TABLESPACE ${tablespace}` : ''
|
|
this.schema.raw(`
|
|
CREATE TABLE release_chunks (
|
|
id bigserial,
|
|
release_id bigint NOT NULL,
|
|
range_id bigint NOT NULL,
|
|
batch_from bigint NOT NULL,
|
|
files integer NOT NULL,
|
|
segments integer NOT NULL,
|
|
payload bytea NOT NULL,
|
|
created_at timestamptz NOT NULL DEFAULT now(),
|
|
PRIMARY KEY (id, created_at)
|
|
) PARTITION BY RANGE (created_at)${where}`)
|
|
// The payload is zstd already: no pglz attempt on it.
|
|
this.schema.raw(`ALTER TABLE release_chunks ALTER COLUMN payload SET STORAGE EXTERNAL`)
|
|
this.schema.raw(`CREATE INDEX release_chunks_release_idx ON release_chunks (release_id)`)
|
|
this.schema.raw(`CREATE INDEX release_chunks_range_idx ON release_chunks (range_id)`)
|
|
}
|
|
|
|
async down() {
|
|
this.schema.raw(`DROP TABLE IF EXISTS release_chunks CASCADE`)
|
|
}
|
|
}
|