mirror of
https://github.com/seaweedfs/seaweedfs.git
synced 2026-09-19 13:00:45 +02:00
* fix(filer): leave reader cache unbounded without an explicit budget NewReaderCache silently installed a 256MiB ReaderCacheBudget when the caller passed none. Only weed mount opts into a budget; every other caller (S3 gateway, WebDAV, query engine, mq logstore) inherited the cap. Under ~90 concurrent S3 GETs of medium objects, prefetch wants far more than 64 chunk buffers, so reserve() serialized chunk fetches, clients timed out and retried, and the retry re-downloaded chunks the cancelled request had already fetched. A nil budget now means unbounded, restoring the pre-4.47 behavior for callers that never asked for a memory cap; reserve/complete/release are nil-safe. The mount path is unchanged and still enforces -readerCacheSizeMB. Fixes #11380 * feat(s3): expose -s3.readerCacheSizeMB reader buffer budget Operators who want the S3 gateway read path memory-bounded can now opt in: -s3.readerCacheSizeMB on weed filer/server/mini and -readerCacheSizeMB on standalone weed s3, matching the mount flag. The default 0 keeps the unbounded pre-4.47 behavior; a positive value installs a shared ReaderCacheBudget across in-flight and retained chunk buffers for all S3 GETs. * fix(filer): validate chunk size before consulting the reader budget A nil budget returned early and skipped the negative chunkSize check, letting a corrupted size reach mem.Allocate and panic. Also drop the command-specific flag prefix from the S3 validation error since standalone weed s3 exposes the option as -readerCacheSizeMB. * filer: drop chunk buffers once fully consumed ReaderCache retained every completed chunk buffer in the downloaders map until the slot limit evicted it, so buffers lingered after all readers finished with them. Track attached readers on each SingleChunkCacher and remove the cacher when the last reader consumes the buffer to its end. In-flight download deduplication and the prefetch handoff are unchanged: a buffer always survives until fully read, partial reads keep it available, and an attached reader pins a consumed buffer until it detaches. Repeat reads now go through the chunk cache where enabled, or refetch. * filer: drop consumed buffers on last detach, rechecked under cache lock Two review findings on the drop-on-consume change: - Removal only fired when the detaching reader itself reached the chunk end. If the end-reaching reader finished first and the last remaining reader did a partial read or cancelled, the consumed buffer and its budget reservation lingered until eviction. Track a persistent consumed flag instead, so any end-reaching read marks the buffer and the last detach drops it. - remove() checked only map identity, so a reader attaching between the reader count hitting zero and removal could attach to a cacher that was then deleted underneath it. removeConsumed() re-checks identity, readers == 0, and consumed under the ReaderCache lock; a raced attach keeps the cacher and its own detach retries the removal.
727 lines
22 KiB
Go
727 lines
22 KiB
Go
package filer
|
|
|
|
import (
|
|
"context"
|
|
"fmt"
|
|
"net/http"
|
|
"net/http/httptest"
|
|
"sync"
|
|
"sync/atomic"
|
|
"testing"
|
|
"time"
|
|
|
|
util_http "github.com/seaweedfs/seaweedfs/weed/util/http"
|
|
)
|
|
|
|
// mockChunkCacheForReaderCache implements chunk cache for testing
|
|
type mockChunkCacheForReaderCache struct {
|
|
data map[string][]byte
|
|
hitCount int32
|
|
mu sync.Mutex
|
|
}
|
|
|
|
type mockCacheInvalidatorForReaderCache struct {
|
|
calls int32
|
|
mu sync.Mutex
|
|
fileId string
|
|
}
|
|
|
|
func (m *mockCacheInvalidatorForReaderCache) InvalidateCache(fileId string) {
|
|
atomic.AddInt32(&m.calls, 1)
|
|
m.mu.Lock()
|
|
defer m.mu.Unlock()
|
|
m.fileId = fileId
|
|
}
|
|
|
|
func (m *mockCacheInvalidatorForReaderCache) lastFileId() string {
|
|
m.mu.Lock()
|
|
defer m.mu.Unlock()
|
|
return m.fileId
|
|
}
|
|
|
|
func newMockChunkCacheForReaderCache() *mockChunkCacheForReaderCache {
|
|
return &mockChunkCacheForReaderCache{
|
|
data: make(map[string][]byte),
|
|
}
|
|
}
|
|
|
|
func (m *mockChunkCacheForReaderCache) GetChunk(fileId string, minSize uint64) []byte {
|
|
m.mu.Lock()
|
|
defer m.mu.Unlock()
|
|
if d, ok := m.data[fileId]; ok {
|
|
atomic.AddInt32(&m.hitCount, 1)
|
|
return d
|
|
}
|
|
return nil
|
|
}
|
|
|
|
func (m *mockChunkCacheForReaderCache) ReadChunkAt(data []byte, fileId string, offset uint64) (int, error) {
|
|
m.mu.Lock()
|
|
defer m.mu.Unlock()
|
|
if d, ok := m.data[fileId]; ok && int(offset) < len(d) {
|
|
atomic.AddInt32(&m.hitCount, 1)
|
|
n := copy(data, d[offset:])
|
|
return n, nil
|
|
}
|
|
return 0, nil
|
|
}
|
|
|
|
func (m *mockChunkCacheForReaderCache) SetChunk(fileId string, data []byte) {
|
|
m.mu.Lock()
|
|
defer m.mu.Unlock()
|
|
m.data[fileId] = data
|
|
}
|
|
|
|
func (m *mockChunkCacheForReaderCache) GetMaxFilePartSizeInCache() uint64 {
|
|
return 1024 * 1024 // 1MB
|
|
}
|
|
|
|
func (m *mockChunkCacheForReaderCache) IsInCache(fileId string, lockNeeded bool) bool {
|
|
m.mu.Lock()
|
|
defer m.mu.Unlock()
|
|
_, ok := m.data[fileId]
|
|
return ok
|
|
}
|
|
|
|
func TestReaderCacheRetryAfterCacheInvalidation(t *testing.T) {
|
|
cache := newMockChunkCacheForReaderCache()
|
|
invalidator := &mockCacheInvalidatorForReaderCache{}
|
|
fileId := "425141,ef8914e9bdbbe8cb6838191f"
|
|
staleUrl := "http://fast-volume-1/" + fileId
|
|
freshUrl := "http://fast-volume-9/" + fileId
|
|
testData := []byte("fresh chunk data after cache invalidation")
|
|
|
|
var lookupCount int32
|
|
lookupFn := func(ctx context.Context, requestedFileId string) ([]string, error) {
|
|
if requestedFileId != fileId {
|
|
return nil, fmt.Errorf("unexpected lookup file id %s", requestedFileId)
|
|
}
|
|
atomic.AddInt32(&lookupCount, 1)
|
|
if atomic.LoadInt32(&invalidator.calls) == 0 {
|
|
return []string{staleUrl}, nil
|
|
}
|
|
return []string{freshUrl}, nil
|
|
}
|
|
|
|
var fetchCount int32
|
|
fetchFn := func(ctx context.Context, buffer []byte, urlStrings []string, cipherKey []byte, isGzipped bool, isFullChunk bool, offset int64, requestedFileId string, _ util_http.RefreshUrlsFunc) (int, error) {
|
|
if requestedFileId != fileId {
|
|
return 0, fmt.Errorf("unexpected fetch file id %s", requestedFileId)
|
|
}
|
|
switch atomic.AddInt32(&fetchCount, 1) {
|
|
case 1:
|
|
if len(urlStrings) != 1 || urlStrings[0] != staleUrl {
|
|
return 0, fmt.Errorf("first fetch should use stale url %v", urlStrings)
|
|
}
|
|
return 0, fmt.Errorf("404 Not Found: not found")
|
|
case 2:
|
|
if len(urlStrings) != 1 || urlStrings[0] != freshUrl {
|
|
return 0, fmt.Errorf("retry fetch should use fresh url %v", urlStrings)
|
|
}
|
|
return copy(buffer, testData), nil
|
|
default:
|
|
return 0, fmt.Errorf("unexpected extra fetch with urls %v", urlStrings)
|
|
}
|
|
}
|
|
|
|
rc := NewReaderCache(10, cache, lookupFn, invalidator)
|
|
rc.fetchChunkDataFn = fetchFn
|
|
defer rc.destroy()
|
|
|
|
buffer := make([]byte, len(testData))
|
|
n, err := rc.ReadChunkAt(context.Background(), buffer, fileId, nil, false, 0, len(testData), true)
|
|
if err != nil {
|
|
t.Fatalf("expected successful retry, got %v", err)
|
|
}
|
|
if got := string(buffer[:n]); got != string(testData) {
|
|
t.Fatalf("expected %q, got %q", testData, got)
|
|
}
|
|
if got := atomic.LoadInt32(&invalidator.calls); got != 1 {
|
|
t.Fatalf("expected one cache invalidation, got %d", got)
|
|
}
|
|
if got := invalidator.lastFileId(); got != fileId {
|
|
t.Fatalf("expected invalidated file id %s, got %s", fileId, got)
|
|
}
|
|
if got := atomic.LoadInt32(&lookupCount); got != 2 {
|
|
t.Fatalf("expected lookup before fetch and after invalidation, got %d", got)
|
|
}
|
|
if got := atomic.LoadInt32(&fetchCount); got != 2 {
|
|
t.Fatalf("expected stale fetch and retry fetch, got %d", got)
|
|
}
|
|
}
|
|
|
|
// TestReaderCacheRefreshesLocationsAfterPartialFailure is the mount reading
|
|
// through a cached location list that still names a replica the master has
|
|
// dropped: the read succeeds on the other replica, and the cached entry must
|
|
// be dropped and looked up again so the next read starts without the dead one.
|
|
func TestReaderCacheRefreshesLocationsAfterPartialFailure(t *testing.T) {
|
|
payload := []byte("chunk contents")
|
|
live := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
|
|
w.Write(payload)
|
|
}))
|
|
defer live.Close()
|
|
dead := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {}))
|
|
deadURL := dead.URL
|
|
dead.Close()
|
|
|
|
fileId := "3,abc"
|
|
invalidator := &mockCacheInvalidatorForReaderCache{}
|
|
var lookupCount int32
|
|
lookupFn := func(ctx context.Context, requestedFileId string) ([]string, error) {
|
|
atomic.AddInt32(&lookupCount, 1)
|
|
if atomic.LoadInt32(&invalidator.calls) == 0 {
|
|
return []string{deadURL + "/" + fileId, live.URL + "/" + fileId}, nil
|
|
}
|
|
return []string{live.URL + "/" + fileId}, nil
|
|
}
|
|
|
|
rc := NewReaderCache(10, newMockChunkCacheForReaderCache(), lookupFn, invalidator)
|
|
defer rc.destroy()
|
|
|
|
buffer := make([]byte, len(payload))
|
|
n, err := rc.ReadChunkAt(context.Background(), buffer, fileId, nil, false, 0, len(payload), false)
|
|
if err != nil || string(buffer[:n]) != string(payload) {
|
|
t.Fatalf("got %q, %v; want %q", buffer[:n], err, payload)
|
|
}
|
|
if got := atomic.LoadInt32(&invalidator.calls); got != 1 {
|
|
t.Fatalf("expected the stale entry to be invalidated once, got %d", got)
|
|
}
|
|
if got := atomic.LoadInt32(&lookupCount); got != 2 {
|
|
t.Fatalf("expected a lookup before the read and one after invalidation, got %d", got)
|
|
}
|
|
}
|
|
|
|
func TestReaderCacheRemovesFailedDownloader(t *testing.T) {
|
|
cache := newMockChunkCacheForReaderCache()
|
|
fileId := "425141,failed"
|
|
url := "http://fast-volume-1/" + fileId
|
|
|
|
var lookupCount int32
|
|
lookupFn := func(ctx context.Context, requestedFileId string) ([]string, error) {
|
|
if requestedFileId != fileId {
|
|
return nil, fmt.Errorf("unexpected lookup file id %s", requestedFileId)
|
|
}
|
|
atomic.AddInt32(&lookupCount, 1)
|
|
return []string{url}, nil
|
|
}
|
|
|
|
var fetchCount int32
|
|
fetchFn := func(ctx context.Context, buffer []byte, urlStrings []string, cipherKey []byte, isGzipped bool, isFullChunk bool, offset int64, requestedFileId string, _ util_http.RefreshUrlsFunc) (int, error) {
|
|
atomic.AddInt32(&fetchCount, 1)
|
|
return 0, fmt.Errorf("fetch failed")
|
|
}
|
|
|
|
rc := NewReaderCache(10, cache, lookupFn, nil)
|
|
rc.fetchChunkDataFn = fetchFn
|
|
defer rc.destroy()
|
|
|
|
for i := 0; i < 2; i++ {
|
|
buffer := make([]byte, 8)
|
|
_, err := rc.ReadChunkAt(context.Background(), buffer, fileId, nil, false, 0, len(buffer), true)
|
|
if err == nil {
|
|
t.Fatalf("read %d should fail", i+1)
|
|
}
|
|
}
|
|
|
|
if got := atomic.LoadInt32(&lookupCount); got != 2 {
|
|
t.Fatalf("failed downloader should be removed so the second read re-lookups, got %d lookups", got)
|
|
}
|
|
if got := atomic.LoadInt32(&fetchCount); got != 2 {
|
|
t.Fatalf("failed downloader should be removed so the second read refetches, got %d fetches", got)
|
|
}
|
|
}
|
|
|
|
// TestReaderCacheContextCancellation tests that a reader can cancel its wait
|
|
// while the download continues for other readers
|
|
func TestReaderCacheContextCancellation(t *testing.T) {
|
|
cache := newMockChunkCacheForReaderCache()
|
|
|
|
// Create a ReaderCache - we can't easily test the full flow without mocking HTTP,
|
|
// but we can test the context cancellation in readChunkAt
|
|
rc := NewReaderCache(10, cache, nil, nil)
|
|
defer rc.destroy()
|
|
|
|
// Pre-populate cache to avoid HTTP calls
|
|
testData := []byte("test data for context cancellation")
|
|
cache.SetChunk("test-file-1", testData)
|
|
|
|
// Test that context cancellation works
|
|
ctx, cancel := context.WithCancel(context.Background())
|
|
|
|
buffer := make([]byte, len(testData))
|
|
n, err := rc.ReadChunkAt(ctx, buffer, "test-file-1", nil, false, 0, len(testData), true)
|
|
if err != nil {
|
|
t.Errorf("Expected no error, got: %v", err)
|
|
}
|
|
if n != len(testData) {
|
|
t.Errorf("Expected %d bytes, got %d", len(testData), n)
|
|
}
|
|
|
|
// Cancel context and verify it doesn't affect already completed reads
|
|
cancel()
|
|
|
|
// Subsequent read with cancelled context should still work from cache
|
|
buffer2 := make([]byte, len(testData))
|
|
n2, err2 := rc.ReadChunkAt(ctx, buffer2, "test-file-1", nil, false, 0, len(testData), true)
|
|
// Note: This may or may not error depending on whether it hits cache
|
|
_ = n2
|
|
_ = err2
|
|
}
|
|
|
|
// TestReaderCacheFallbackToChunkCache tests that when a cacher returns n=0, err=nil,
|
|
// we fall back to the chunkCache
|
|
func TestReaderCacheFallbackToChunkCache(t *testing.T) {
|
|
cache := newMockChunkCacheForReaderCache()
|
|
|
|
// Pre-populate the chunk cache with data
|
|
testData := []byte("fallback test data that should be found in chunk cache")
|
|
cache.SetChunk("fallback-file", testData)
|
|
|
|
rc := NewReaderCache(10, cache, nil, nil)
|
|
defer rc.destroy()
|
|
|
|
// Read should hit the chunk cache
|
|
buffer := make([]byte, len(testData))
|
|
n, err := rc.ReadChunkAt(context.Background(), buffer, "fallback-file", nil, false, 0, len(testData), true)
|
|
|
|
if err != nil {
|
|
t.Errorf("Expected no error, got: %v", err)
|
|
}
|
|
if n != len(testData) {
|
|
t.Errorf("Expected %d bytes, got %d", len(testData), n)
|
|
}
|
|
|
|
// Verify cache was hit
|
|
if cache.hitCount == 0 {
|
|
t.Error("Expected chunk cache to be hit")
|
|
}
|
|
}
|
|
|
|
// TestReaderCacheMultipleReadersWaitForSameChunk tests that multiple readers
|
|
// can wait for the same chunk download to complete
|
|
func TestReaderCacheMultipleReadersWaitForSameChunk(t *testing.T) {
|
|
cache := newMockChunkCacheForReaderCache()
|
|
|
|
// Pre-populate cache so we don't need HTTP
|
|
testData := make([]byte, 1024)
|
|
for i := range testData {
|
|
testData[i] = byte(i % 256)
|
|
}
|
|
cache.SetChunk("shared-chunk", testData)
|
|
|
|
rc := NewReaderCache(10, cache, nil, nil)
|
|
defer rc.destroy()
|
|
|
|
// Launch multiple concurrent readers for the same chunk
|
|
numReaders := 10
|
|
var wg sync.WaitGroup
|
|
errors := make(chan error, numReaders)
|
|
bytesRead := make(chan int, numReaders)
|
|
|
|
for i := 0; i < numReaders; i++ {
|
|
wg.Add(1)
|
|
go func() {
|
|
defer wg.Done()
|
|
buffer := make([]byte, len(testData))
|
|
n, err := rc.ReadChunkAt(context.Background(), buffer, "shared-chunk", nil, false, 0, len(testData), true)
|
|
if err != nil {
|
|
errors <- err
|
|
}
|
|
bytesRead <- n
|
|
}()
|
|
}
|
|
|
|
wg.Wait()
|
|
close(errors)
|
|
close(bytesRead)
|
|
|
|
// Check for errors
|
|
for err := range errors {
|
|
t.Errorf("Reader got error: %v", err)
|
|
}
|
|
|
|
// Verify all readers got the expected data
|
|
for n := range bytesRead {
|
|
if n != len(testData) {
|
|
t.Errorf("Expected %d bytes, got %d", len(testData), n)
|
|
}
|
|
}
|
|
}
|
|
|
|
// TestReaderCachePartialRead tests reading at different offsets
|
|
func TestReaderCachePartialRead(t *testing.T) {
|
|
cache := newMockChunkCacheForReaderCache()
|
|
|
|
testData := []byte("0123456789ABCDEFGHIJ")
|
|
cache.SetChunk("partial-read-file", testData)
|
|
|
|
rc := NewReaderCache(10, cache, nil, nil)
|
|
defer rc.destroy()
|
|
|
|
tests := []struct {
|
|
name string
|
|
offset int64
|
|
size int
|
|
expected []byte
|
|
}{
|
|
{"read from start", 0, 5, []byte("01234")},
|
|
{"read from middle", 5, 5, []byte("56789")},
|
|
{"read to end", 15, 5, []byte("FGHIJ")},
|
|
{"read single byte", 10, 1, []byte("A")},
|
|
}
|
|
|
|
for _, tt := range tests {
|
|
t.Run(tt.name, func(t *testing.T) {
|
|
buffer := make([]byte, tt.size)
|
|
n, err := rc.ReadChunkAt(context.Background(), buffer, "partial-read-file", nil, false, tt.offset, len(testData), true)
|
|
|
|
if err != nil {
|
|
t.Errorf("Expected no error, got: %v", err)
|
|
}
|
|
if n != tt.size {
|
|
t.Errorf("Expected %d bytes, got %d", tt.size, n)
|
|
}
|
|
if string(buffer[:n]) != string(tt.expected) {
|
|
t.Errorf("Expected %q, got %q", tt.expected, buffer[:n])
|
|
}
|
|
})
|
|
}
|
|
}
|
|
|
|
// TestReaderCacheCleanup tests that old downloaders are cleaned up
|
|
func TestReaderCacheCleanup(t *testing.T) {
|
|
cache := newMockChunkCacheForReaderCache()
|
|
|
|
// Create cache with limit of 3
|
|
rc := NewReaderCache(3, cache, nil, nil)
|
|
defer rc.destroy()
|
|
|
|
// Add data for multiple files
|
|
for i := 0; i < 5; i++ {
|
|
fileId := string(rune('A' + i))
|
|
data := []byte("data for file " + fileId)
|
|
cache.SetChunk(fileId, data)
|
|
}
|
|
|
|
// Read from multiple files - should trigger cleanup when exceeding limit
|
|
for i := 0; i < 5; i++ {
|
|
fileId := string(rune('A' + i))
|
|
buffer := make([]byte, 20)
|
|
_, err := rc.ReadChunkAt(context.Background(), buffer, fileId, nil, false, 0, 20, true)
|
|
if err != nil {
|
|
t.Errorf("Read error for file %s: %v", fileId, err)
|
|
}
|
|
}
|
|
|
|
// Cache should still work - reads should succeed
|
|
for i := 0; i < 5; i++ {
|
|
fileId := string(rune('A' + i))
|
|
buffer := make([]byte, 20)
|
|
n, err := rc.ReadChunkAt(context.Background(), buffer, fileId, nil, false, 0, 20, true)
|
|
if err != nil {
|
|
t.Errorf("Second read error for file %s: %v", fileId, err)
|
|
}
|
|
if n == 0 {
|
|
t.Errorf("Expected data for file %s, got 0 bytes", fileId)
|
|
}
|
|
}
|
|
}
|
|
|
|
// TestSingleChunkCacherDoneSignal tests that done channel is always closed
|
|
func TestSingleChunkCacherDoneSignal(t *testing.T) {
|
|
cache := newMockChunkCacheForReaderCache()
|
|
rc := NewReaderCache(10, cache, nil, nil)
|
|
defer rc.destroy()
|
|
|
|
// Test that we can read even when data is in cache (done channel should work)
|
|
testData := []byte("done signal test")
|
|
cache.SetChunk("done-signal-test", testData)
|
|
|
|
// Multiple goroutines reading same chunk
|
|
var wg sync.WaitGroup
|
|
for i := 0; i < 5; i++ {
|
|
wg.Add(1)
|
|
go func() {
|
|
defer wg.Done()
|
|
buffer := make([]byte, len(testData))
|
|
ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
|
|
defer cancel()
|
|
|
|
n, err := rc.ReadChunkAt(ctx, buffer, "done-signal-test", nil, false, 0, len(testData), true)
|
|
if err != nil && err != context.DeadlineExceeded {
|
|
t.Errorf("Unexpected error: %v", err)
|
|
}
|
|
if n == 0 && err == nil {
|
|
t.Error("Got 0 bytes with no error")
|
|
}
|
|
}()
|
|
}
|
|
|
|
// Should complete without hanging
|
|
done := make(chan struct{})
|
|
go func() {
|
|
wg.Wait()
|
|
close(done)
|
|
}()
|
|
|
|
select {
|
|
case <-done:
|
|
// Success
|
|
case <-time.After(10 * time.Second):
|
|
t.Fatal("Test timed out - done channel may not be signaled correctly")
|
|
}
|
|
}
|
|
|
|
// ============================================================================
|
|
// Tests that exercise SingleChunkCacher concurrency logic
|
|
// ============================================================================
|
|
//
|
|
// These tests use blocking lookupFileIdFn to exercise the wait/cancellation
|
|
// logic in SingleChunkCacher without requiring HTTP calls.
|
|
|
|
// TestSingleChunkCacherLookupError tests handling of lookup errors
|
|
func TestSingleChunkCacherLookupError(t *testing.T) {
|
|
cache := newMockChunkCacheForReaderCache()
|
|
|
|
// Lookup function that returns an error
|
|
lookupFn := func(ctx context.Context, fileId string) ([]string, error) {
|
|
return nil, fmt.Errorf("lookup failed for %s", fileId)
|
|
}
|
|
|
|
rc := NewReaderCache(10, cache, lookupFn, nil)
|
|
defer rc.destroy()
|
|
|
|
buffer := make([]byte, 100)
|
|
_, err := rc.ReadChunkAt(context.Background(), buffer, "error-test", nil, false, 0, 100, true)
|
|
|
|
if err == nil {
|
|
t.Error("Expected an error, got nil")
|
|
}
|
|
}
|
|
|
|
// TestSingleChunkCacherContextCancellationDuringLookup tests that a reader can
|
|
// cancel its wait while the lookup is in progress. This exercises the actual
|
|
// SingleChunkCacher wait/cancel logic.
|
|
func TestSingleChunkCacherContextCancellationDuringLookup(t *testing.T) {
|
|
cache := newMockChunkCacheForReaderCache()
|
|
lookupStarted := make(chan struct{})
|
|
lookupCanFinish := make(chan struct{})
|
|
|
|
// Lookup function that blocks to simulate slow operation
|
|
lookupFn := func(ctx context.Context, fileId string) ([]string, error) {
|
|
close(lookupStarted)
|
|
<-lookupCanFinish // Block until test allows completion
|
|
return nil, fmt.Errorf("lookup completed but reader should have cancelled")
|
|
}
|
|
|
|
rc := NewReaderCache(10, cache, lookupFn, nil)
|
|
defer rc.destroy()
|
|
defer close(lookupCanFinish) // Ensure cleanup
|
|
|
|
ctx, cancel := context.WithCancel(context.Background())
|
|
readResult := make(chan error, 1)
|
|
|
|
go func() {
|
|
buffer := make([]byte, 100)
|
|
_, err := rc.ReadChunkAt(ctx, buffer, "cancel-during-lookup", nil, false, 0, 100, true)
|
|
readResult <- err
|
|
}()
|
|
|
|
// Wait for lookup to start, then cancel the reader's context
|
|
select {
|
|
case <-lookupStarted:
|
|
cancel() // Cancel the reader while lookup is blocked
|
|
case <-time.After(5 * time.Second):
|
|
t.Fatal("Lookup never started")
|
|
}
|
|
|
|
// Read should return with context.Canceled
|
|
select {
|
|
case err := <-readResult:
|
|
if err != context.Canceled {
|
|
t.Errorf("Expected context.Canceled, got: %v", err)
|
|
}
|
|
case <-time.After(5 * time.Second):
|
|
t.Fatal("Read did not complete after context cancellation")
|
|
}
|
|
}
|
|
|
|
// TestSingleChunkCacherMultipleReadersWaitForDownload tests that multiple readers
|
|
// can wait for the same SingleChunkCacher download to complete. When lookup fails,
|
|
// all readers should receive the same error.
|
|
func TestSingleChunkCacherMultipleReadersWaitForDownload(t *testing.T) {
|
|
cache := newMockChunkCacheForReaderCache()
|
|
lookupStarted := make(chan struct{})
|
|
lookupCanFinish := make(chan struct{})
|
|
var lookupStartedOnce sync.Once
|
|
|
|
// Lookup function that blocks to simulate slow operation
|
|
lookupFn := func(ctx context.Context, fileId string) ([]string, error) {
|
|
lookupStartedOnce.Do(func() { close(lookupStarted) })
|
|
<-lookupCanFinish
|
|
return nil, fmt.Errorf("simulated lookup error")
|
|
}
|
|
|
|
rc := NewReaderCache(10, cache, lookupFn, nil)
|
|
defer rc.destroy()
|
|
|
|
numReaders := 5
|
|
var wg sync.WaitGroup
|
|
errors := make(chan error, numReaders)
|
|
|
|
// Start multiple readers for the same chunk
|
|
for i := 0; i < numReaders; i++ {
|
|
wg.Add(1)
|
|
go func() {
|
|
defer wg.Done()
|
|
buffer := make([]byte, 100)
|
|
_, err := rc.ReadChunkAt(context.Background(), buffer, "shared-chunk", nil, false, 0, 100, true)
|
|
errors <- err
|
|
}()
|
|
}
|
|
|
|
// Wait for lookup to start, then allow completion
|
|
select {
|
|
case <-lookupStarted:
|
|
close(lookupCanFinish)
|
|
case <-time.After(5 * time.Second):
|
|
close(lookupCanFinish)
|
|
t.Fatal("Lookup never started")
|
|
}
|
|
|
|
wg.Wait()
|
|
close(errors)
|
|
|
|
// All readers should receive an error
|
|
errorCount := 0
|
|
for err := range errors {
|
|
if err != nil {
|
|
errorCount++
|
|
}
|
|
}
|
|
if errorCount != numReaders {
|
|
t.Errorf("Expected %d errors, got %d", numReaders, errorCount)
|
|
}
|
|
}
|
|
|
|
// TestReaderCacheDownloaderDedup tests that concurrent ReadChunkAt calls for
|
|
// the same fileId result in only one network fetch (lookup call), because
|
|
// the downloaders map deduplicates in-flight downloads.
|
|
func TestReaderCacheDownloaderDedup(t *testing.T) {
|
|
cache := newMockChunkCacheForReaderCache()
|
|
|
|
var lookupCount int32
|
|
var fetchCount int32
|
|
fetchGate := make(chan struct{})
|
|
testData := []byte("deduplicated data")
|
|
|
|
lookupFn := func(ctx context.Context, fileId string) ([]string, error) {
|
|
atomic.AddInt32(&lookupCount, 1)
|
|
return []string{"http://volume/" + fileId}, nil
|
|
}
|
|
|
|
fetchFn := func(ctx context.Context, buffer []byte, urlStrings []string, cipherKey []byte, isGzipped bool, isFullChunk bool, offset int64, fileId string, _ util_http.RefreshUrlsFunc) (int, error) {
|
|
atomic.AddInt32(&fetchCount, 1)
|
|
<-fetchGate
|
|
return copy(buffer, testData), nil
|
|
}
|
|
|
|
rc := NewReaderCache(10, cache, lookupFn, nil)
|
|
rc.fetchChunkDataFn = fetchFn
|
|
defer rc.destroy()
|
|
|
|
const numReaders = 10
|
|
var wg sync.WaitGroup
|
|
wg.Add(numReaders)
|
|
|
|
for i := 0; i < numReaders; i++ {
|
|
go func() {
|
|
defer wg.Done()
|
|
buffer := make([]byte, 50)
|
|
rc.ReadChunkAt(context.Background(), buffer, "dedup-file", nil, false, 0, 100, false)
|
|
}()
|
|
}
|
|
|
|
// Allow downloads to proceed.
|
|
close(fetchGate)
|
|
wg.Wait()
|
|
|
|
if count := atomic.LoadInt32(&lookupCount); count != 1 {
|
|
t.Errorf("expected exactly 1 lookup call, got %d", count)
|
|
}
|
|
if count := atomic.LoadInt32(&fetchCount); count != 1 {
|
|
t.Errorf("expected exactly 1 fetch call, got %d", count)
|
|
}
|
|
}
|
|
|
|
// TestSingleChunkCacherOneReaderCancelsOthersContinue tests that when one reader
|
|
// cancels, other readers waiting on the same chunk continue to wait.
|
|
func TestSingleChunkCacherOneReaderCancelsOthersContinue(t *testing.T) {
|
|
cache := newMockChunkCacheForReaderCache()
|
|
lookupStarted := make(chan struct{})
|
|
lookupCanFinish := make(chan struct{})
|
|
var lookupStartedOnce sync.Once
|
|
|
|
lookupFn := func(ctx context.Context, fileId string) ([]string, error) {
|
|
lookupStartedOnce.Do(func() { close(lookupStarted) })
|
|
<-lookupCanFinish
|
|
return nil, fmt.Errorf("simulated error after delay")
|
|
}
|
|
|
|
rc := NewReaderCache(10, cache, lookupFn, nil)
|
|
defer rc.destroy()
|
|
|
|
cancelledReaderDone := make(chan error, 1)
|
|
otherReaderDone := make(chan error, 1)
|
|
|
|
ctx, cancel := context.WithCancel(context.Background())
|
|
|
|
// Start reader that will be cancelled
|
|
go func() {
|
|
buffer := make([]byte, 100)
|
|
_, err := rc.ReadChunkAt(ctx, buffer, "shared-chunk-2", nil, false, 0, 100, true)
|
|
cancelledReaderDone <- err
|
|
}()
|
|
|
|
// Start reader that will NOT be cancelled
|
|
go func() {
|
|
buffer := make([]byte, 100)
|
|
_, err := rc.ReadChunkAt(context.Background(), buffer, "shared-chunk-2", nil, false, 0, 100, true)
|
|
otherReaderDone <- err
|
|
}()
|
|
|
|
// Wait for lookup to start
|
|
select {
|
|
case <-lookupStarted:
|
|
case <-time.After(5 * time.Second):
|
|
t.Fatal("Lookup never started")
|
|
}
|
|
|
|
// Cancel the first reader
|
|
cancel()
|
|
|
|
// First reader should complete with context.Canceled quickly
|
|
select {
|
|
case err := <-cancelledReaderDone:
|
|
if err != context.Canceled {
|
|
t.Errorf("Cancelled reader: expected context.Canceled, got: %v", err)
|
|
}
|
|
case <-time.After(2 * time.Second):
|
|
t.Error("Cancelled reader did not complete quickly")
|
|
}
|
|
|
|
// Allow the download to complete
|
|
close(lookupCanFinish)
|
|
|
|
// Other reader should eventually complete (with error since lookup returns error)
|
|
select {
|
|
case err := <-otherReaderDone:
|
|
if err == nil || err == context.Canceled {
|
|
t.Errorf("Other reader: expected non-nil non-cancelled error, got: %v", err)
|
|
}
|
|
// Expected: "simulated error after delay"
|
|
case <-time.After(5 * time.Second):
|
|
t.Error("Other reader did not complete")
|
|
}
|
|
}
|