synology_apm_repo.sdk.dedup.verify_bucket_check module

The per-bucket check dedup.verify_walk.ReachabilitySweep runs on every claimed bucket at FULL or QUICK level (see VerifyLevel), plus its multiprocess worker/executor glue.

check_one_bucket is the one implementation both the in-process path (ReachabilitySweep._check_one_bucket) and the multiprocess path (verify_bucket_worker) call, so it takes every dependency as a parameter. verify_worker_init runs once per worker process, so its Pool stays warm across every bucket that worker checks.

async synology_apm_repo.sdk.dedup.verify_bucket_check.check_one_bucket(pool, key, level)

Open, structurally check, and (at FULL) content-check one bucket.

Every Exception but StorageBackendError becomes a Finding, so one bucket’s failure never cancels its siblings in a concurrent dispatch; a transport failure says nothing about the data, so it propagates and aborts the verify, as cancellation does.

Returns (findings, key_missing): findings are untagged, and, when key_missing, include a Symptom.KEY_MISSING finding the caller deduplicates across a run.

synology_apm_repo.sdk.dedup.verify_bucket_check.verify_worker_init(descriptor)

ProcessPoolExecutor initializer: builds this worker’s Pool once per process.

class synology_apm_repo.sdk.dedup.verify_bucket_check.VerifyExecutor(pool_descriptor)

Bases: BoundProcessPool

Worker processes serving verify_bucket_worker for the repository pool_descriptor describes (a FULL-verify pool’s).

synology_apm_repo.sdk.dedup.verify_bucket_check.shared_verify_executor(repos)

One executor several catalogs’ FULL verifies can share, when every one describes the same pool (the common vault case); None when they differ (object-storage siblings can be separate stores) or one can’t be rebuilt in a worker, so each verify builds its own.

synology_apm_repo.sdk.dedup.verify_bucket_check.verify_bucket_worker(key)

The multiprocess path’s per-bucket work item — a thin wrapper around check_one_bucket. Always called at FULL level.