Phase 9: two-phase garbage collection engine
GC engine (internal/gc/): Collector.Run() implements the two-phase algorithm — Phase 1 finds unreferenced blobs and deletes DB rows in a single transaction, Phase 2 deletes blob files from storage. Registry-wide mutex blocks concurrent GC runs. Collector.Reconcile() scans filesystem for orphaned files with no DB row (crash recovery). Wired into admin_gc.go: POST /v1/gc now launches the real collector in a goroutine with gc_started/gc_completed audit events. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
230
internal/gc/gc_test.go
Normal file
230
internal/gc/gc_test.go
Normal file
@@ -0,0 +1,230 @@
|
||||
package gc
|
||||
|
||||
import (
|
||||
"context"
|
||||
"errors"
|
||||
"sync"
|
||||
"testing"
|
||||
)
|
||||
|
||||
// fakeDB implements gc.DB for tests.
|
||||
type fakeDB struct {
|
||||
mu sync.Mutex
|
||||
unreferenced []UnreferencedBlob
|
||||
blobsExist map[string]bool
|
||||
}
|
||||
|
||||
func newFakeDB() *fakeDB {
|
||||
return &fakeDB{
|
||||
blobsExist: make(map[string]bool),
|
||||
}
|
||||
}
|
||||
|
||||
func (f *fakeDB) FindAndDeleteUnreferencedBlobs() ([]UnreferencedBlob, error) {
|
||||
f.mu.Lock()
|
||||
defer f.mu.Unlock()
|
||||
result := make([]UnreferencedBlob, len(f.unreferenced))
|
||||
copy(result, f.unreferenced)
|
||||
// Simulate deletion by removing from blobsExist.
|
||||
for _, b := range f.unreferenced {
|
||||
delete(f.blobsExist, b.Digest)
|
||||
}
|
||||
f.unreferenced = nil
|
||||
return result, nil
|
||||
}
|
||||
|
||||
func (f *fakeDB) BlobExistsByDigest(digest string) (bool, error) {
|
||||
f.mu.Lock()
|
||||
defer f.mu.Unlock()
|
||||
return f.blobsExist[digest], nil
|
||||
}
|
||||
|
||||
// fakeStorage implements gc.Storage for tests.
|
||||
type fakeStorage struct {
|
||||
mu sync.Mutex
|
||||
blobs map[string]int64 // digest -> size
|
||||
deleted []string
|
||||
}
|
||||
|
||||
func newFakeStorage() *fakeStorage {
|
||||
return &fakeStorage{
|
||||
blobs: make(map[string]int64),
|
||||
}
|
||||
}
|
||||
|
||||
func (f *fakeStorage) Delete(digest string) error {
|
||||
f.mu.Lock()
|
||||
defer f.mu.Unlock()
|
||||
if _, ok := f.blobs[digest]; !ok {
|
||||
return errors.New("not found")
|
||||
}
|
||||
delete(f.blobs, digest)
|
||||
f.deleted = append(f.deleted, digest)
|
||||
return nil
|
||||
}
|
||||
|
||||
func (f *fakeStorage) ListBlobDigests() ([]string, error) {
|
||||
f.mu.Lock()
|
||||
defer f.mu.Unlock()
|
||||
var digests []string
|
||||
for d := range f.blobs {
|
||||
digests = append(digests, d)
|
||||
}
|
||||
return digests, nil
|
||||
}
|
||||
|
||||
func TestGCRemovesUnreferencedBlobs(t *testing.T) {
|
||||
db := newFakeDB()
|
||||
db.unreferenced = []UnreferencedBlob{
|
||||
{Digest: "sha256:dead1", Size: 100},
|
||||
{Digest: "sha256:dead2", Size: 200},
|
||||
}
|
||||
db.blobsExist["sha256:dead1"] = true
|
||||
db.blobsExist["sha256:dead2"] = true
|
||||
db.blobsExist["sha256:alive"] = true // referenced, not in unreferenced list
|
||||
|
||||
store := newFakeStorage()
|
||||
store.blobs["sha256:dead1"] = 100
|
||||
store.blobs["sha256:dead2"] = 200
|
||||
store.blobs["sha256:alive"] = 300
|
||||
|
||||
c := New(db, store)
|
||||
result, err := c.Run(context.Background())
|
||||
if err != nil {
|
||||
t.Fatalf("Run: %v", err)
|
||||
}
|
||||
|
||||
if result.BlobsRemoved != 2 {
|
||||
t.Fatalf("BlobsRemoved: got %d, want 2", result.BlobsRemoved)
|
||||
}
|
||||
if result.BytesFreed != 300 {
|
||||
t.Fatalf("BytesFreed: got %d, want 300", result.BytesFreed)
|
||||
}
|
||||
|
||||
// Dead blobs should be deleted from storage.
|
||||
if _, ok := store.blobs["sha256:dead1"]; ok {
|
||||
t.Fatal("sha256:dead1 should have been deleted from storage")
|
||||
}
|
||||
if _, ok := store.blobs["sha256:dead2"]; ok {
|
||||
t.Fatal("sha256:dead2 should have been deleted from storage")
|
||||
}
|
||||
|
||||
// Alive blob should still exist.
|
||||
if _, ok := store.blobs["sha256:alive"]; !ok {
|
||||
t.Fatal("sha256:alive should still exist in storage")
|
||||
}
|
||||
}
|
||||
|
||||
func TestGCDoesNotRemoveReferencedBlobs(t *testing.T) {
|
||||
db := newFakeDB()
|
||||
// No unreferenced blobs.
|
||||
db.blobsExist["sha256:alive"] = true
|
||||
|
||||
store := newFakeStorage()
|
||||
store.blobs["sha256:alive"] = 500
|
||||
|
||||
c := New(db, store)
|
||||
result, err := c.Run(context.Background())
|
||||
if err != nil {
|
||||
t.Fatalf("Run: %v", err)
|
||||
}
|
||||
|
||||
if result.BlobsRemoved != 0 {
|
||||
t.Fatalf("BlobsRemoved: got %d, want 0", result.BlobsRemoved)
|
||||
}
|
||||
|
||||
if _, ok := store.blobs["sha256:alive"]; !ok {
|
||||
t.Fatal("referenced blob should not be deleted")
|
||||
}
|
||||
}
|
||||
|
||||
func TestGCConcurrentRejected(t *testing.T) {
|
||||
db := newFakeDB()
|
||||
store := newFakeStorage()
|
||||
c := New(db, store)
|
||||
|
||||
// Acquire the lock manually.
|
||||
c.mu.Lock()
|
||||
|
||||
// Try to run GC — should fail.
|
||||
_, err := c.Run(context.Background())
|
||||
if !errors.Is(err, ErrGCRunning) {
|
||||
t.Fatalf("expected ErrGCRunning, got %v", err)
|
||||
}
|
||||
|
||||
c.mu.Unlock()
|
||||
|
||||
// Now it should work.
|
||||
result, err := c.Run(context.Background())
|
||||
if err != nil {
|
||||
t.Fatalf("Run after unlock: %v", err)
|
||||
}
|
||||
if result.BlobsRemoved != 0 {
|
||||
t.Fatalf("BlobsRemoved: got %d, want 0", result.BlobsRemoved)
|
||||
}
|
||||
}
|
||||
|
||||
func TestGCEmptyRegistry(t *testing.T) {
|
||||
db := newFakeDB()
|
||||
store := newFakeStorage()
|
||||
c := New(db, store)
|
||||
|
||||
result, err := c.Run(context.Background())
|
||||
if err != nil {
|
||||
t.Fatalf("Run: %v", err)
|
||||
}
|
||||
if result.BlobsRemoved != 0 {
|
||||
t.Fatalf("BlobsRemoved: got %d, want 0", result.BlobsRemoved)
|
||||
}
|
||||
if result.Duration <= 0 {
|
||||
t.Fatal("Duration should be positive")
|
||||
}
|
||||
}
|
||||
|
||||
func TestReconcileCleansOrphanedFiles(t *testing.T) {
|
||||
db := newFakeDB()
|
||||
// Only sha256:alive has a DB row.
|
||||
db.blobsExist["sha256:alive"] = true
|
||||
|
||||
store := newFakeStorage()
|
||||
store.blobs["sha256:alive"] = 100
|
||||
store.blobs["sha256:orphan1"] = 200
|
||||
store.blobs["sha256:orphan2"] = 300
|
||||
|
||||
c := New(db, store)
|
||||
result, err := c.Reconcile(context.Background())
|
||||
if err != nil {
|
||||
t.Fatalf("Reconcile: %v", err)
|
||||
}
|
||||
|
||||
if result.BlobsRemoved != 2 {
|
||||
t.Fatalf("BlobsRemoved: got %d, want 2", result.BlobsRemoved)
|
||||
}
|
||||
|
||||
// Alive blob should still exist.
|
||||
if _, ok := store.blobs["sha256:alive"]; !ok {
|
||||
t.Fatal("sha256:alive should still exist")
|
||||
}
|
||||
|
||||
// Orphans should be gone.
|
||||
if _, ok := store.blobs["sha256:orphan1"]; ok {
|
||||
t.Fatal("sha256:orphan1 should have been deleted")
|
||||
}
|
||||
if _, ok := store.blobs["sha256:orphan2"]; ok {
|
||||
t.Fatal("sha256:orphan2 should have been deleted")
|
||||
}
|
||||
}
|
||||
|
||||
func TestReconcileEmptyStorage(t *testing.T) {
|
||||
db := newFakeDB()
|
||||
store := newFakeStorage()
|
||||
c := New(db, store)
|
||||
|
||||
result, err := c.Reconcile(context.Background())
|
||||
if err != nil {
|
||||
t.Fatalf("Reconcile: %v", err)
|
||||
}
|
||||
if result.BlobsRemoved != 0 {
|
||||
t.Fatalf("BlobsRemoved: got %d, want 0", result.BlobsRemoved)
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user