Eviction deleted an object before checking its record still pointed at
it. If a refetch had moved the record, that object was queued and a
request that read the record earlier could still be opening it. Eviction
now clears first, skips the delete if the record moved, and queues the
object when the delete fails after a clear.
Reclaim queues a failed delete again, which moves it behind the rest, so
objects the backend keeps refusing cannot fill every batch.
Also drop the clearArtifactCache passthrough and ClearCachedArtifact's
unused context, note that only tests call ArtifactPath, and document that
max_size does not count objects waiting to be reclaimed.
Every fetch of an artifact wrote one path, so fetches with different URLs
or digests, which do not share a coalescing key, could overwrite each
other's bytes, serve the wrong ones, or delete them on a digest mismatch
while the other was about to open them.
storeArtifact now writes {ecosystem}/{name}/{version}/{fetch id}/{filename},
with a random id per fetch. Most ecosystems learn the digest only after
the fetch, and storage has no rename, so the id is the one rule that fits
all of them. Existing records keep their paths and stay readable. On a
file:// bucket, Delete now removes an emptied fetch directory, which
fileblob leaves behind, and no other directory, since another fetch may
be creating its own inside it.
A path a record stops pointing at is queued in pending_deletes rather
than deleted, since a request that read the record may still open it. A
loop deletes queued paths after max(1h, direct_serve_ttl), whether or not
max_size is set. UpsertArtifact and clears apply only if the record still
holds the path the caller read, so racing commits each queue the path
they replaced and a clear never orphans a newer commit. Eviction still
deletes inline, and counts space as freed only when the record still
pointed at what it deleted.
A failed delete leaves the record eligible, so the same batch came back
on the next iteration while the recorded size never dropped. Against a
backend refusing deletes the pass retried forever, burning CPU and
flooding the log. A canceled context made it faster: every delete fails
at once, so the loop spun as quickly as the database could answer.
The pass now ends when a batch clears no record, and returns as soon as
the context is canceled. Progress counts records rather than bytes,
because a cached record can have no size recorded. A pass can therefore
end with the cache still above its limit, leaving the records that
failed for the next pass to retry.
* Add GCS storage backend with Workload Identity support
Register gocloud.dev/blob/gcsblob so gs:// URLs are accepted as a storage
backend. Authentication uses Application Default Credentials, which makes
GKE Workload Identity work out of the box; signed URLs (direct_serve)
fall back to the IAM Credentials signBlob API when no private key is
available.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* Replace gcsblob with lightweight GCS backend
* Extract GCS client into standalone module
---------
Co-authored-by: Anthony A. <github@anthony-arnaud.fr>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
* Bump go tool golangci-lint to v2.13.1
The .golangci.yml goconst.ignore-tests setting was added in v2.12.0
(golangci/golangci-lint#6480). On the previously pinned v2.10.1,
config verify fails with "additional properties 'ignore-tests' not
allowed" and the setting is silently ignored at run time, so goconst
counts test-file literals toward min-occurrences.
* Apply gofmt and CutSuffix simplification
- gofmt -w internal/server/health_test.go
- Replace HasSuffix+TrimSuffix with CutSuffix in ParseSize
* Remove dead code and migrate tests off legacy Filesystem storage
Migrate the three test call sites of storage.NewFilesystem to
storage.OpenBucket("file://...") and drop the deprecated
StorageConfig.Path field from test configs, then delete code that
deadcode reports as unreachable from cmd/proxy:
- internal/storage/filesystem.go and its tests
- storage.HashingReader
- enrichment.Service.BulkCheckVulnerabilities and NormalizeLicense
- server.ActiveRequestsMiddleware (no-op body; the real tracking
is the inline r.Use at server.go:226)
- mirror.RegistrySource (unimplemented stub)
metrics.UpdateCircuitBreakerState and RecordCircuitBreakerTrip are
kept because #275 wires them.
Update the CONTRIBUTING.md storage section to reflect blob.go.
Closes#99. The max_size storage config was parsed and validated but
never enforced. This adds a background eviction loop that periodically
checks total cache size and evicts least recently used artifacts when
the limit is exceeded.
