Files
seaweedfs/weed/filer/reader_cache_test.go
T
d695d101b5 Fix/s3 abandoned chunk download (#12616)
* filer: cancel a chunk download that no reader waits for

* filer: stop waiting for reader cache budget once cancelled

* filer: keep a cancelled download's re-lookups off its context

The refresh and retry lookups took the download's context. FilerClient
coalesces concurrent lookups of a volume, so a cancelled download leading
one failed the other readers waiting on it. Fetches still stop on cancel.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* filer: test that cancelling a download leaves its re-lookup running

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* filer: cancel unheld downloads when the reader cache is destroyed

ReaderCache.destroy waited for every downloader, so closing a reader
waited for a stalled prefetch nobody read. Downloads with no reader or
stream on them are now cancelled first; ones still in use keep running.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* filer: test that destroying the reader cache cancels an unread prefetch

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* filer: cancel a destroyed download once its last holder leaves

destroy skipped downloads a reader or stream still held, and once they
were out of the map nothing cancelled them, so destroy waited for a fetch
nobody used. removeConsumedLocked now cancels any consumable download,
whether or not it is still in the map.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* filer: test a destroyed download cancelled after its last stream leaves

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* filer: release the stream only after destroy dropped the download

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* filer: test a cancelled stream leaving a waiting reader's download running

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Chris Lu <chrislusf@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-13 12:45:34 -08:00

1179 lines
36 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()
reachServer(t, live)
deadURL := deadServerURL(t)
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)
}
}
// reachServer has srv answer one request. The read path orders replicas by
// whether their host recently failed, and srv's port may be one an earlier test
// left marked as failed, so the answer clears that record.
func reachServer(t *testing.T, srv *httptest.Server) {
t.Helper()
if _, _, err := util_http.Get(srv.URL); err != nil {
t.Fatalf("reach %s: %v", srv.URL, err)
}
}
// deadServerURL is a replica that answered once and then went away.
func deadServerURL(t *testing.T) string {
t.Helper()
dead := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {}))
defer dead.Close()
reachServer(t, dead)
return dead.URL
}
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")
}
}
// TestReaderCacheBookkeepingOffCacheLock guards the contention fix: pin
// lifecycle calls that cannot remove a cacher must not block on the
// ReaderCache lock.
func TestReaderCacheBookkeepingOffCacheLock(t *testing.T) {
cache := newMockChunkCacheForReaderCache()
rc := NewReaderCache(10, cache, nil, nil)
defer rc.destroy()
cacher := &SingleChunkCacher{parent: rc, chunkFileId: "pinned-chunk"}
atomic.StoreInt32(&cacher.readers, 1) // a read in flight: not consumable
rc.Lock()
started := make(chan struct{})
done := make(chan struct{})
go func() {
close(started)
defer close(done)
var stream chunkStream
stream.pin(cacher)
rc.releaseStream(&stream)
}()
<-started
select {
case <-done:
case <-time.After(time.Second):
rc.Unlock()
t.Fatal("stream pin lifecycle blocked on the cache lock")
}
rc.Unlock()
}
// Releasing the last stream of an unfinished download cancels it without waiting for it.
func TestAbandonedChunkDownloadStopsWithItsLastStream(t *testing.T) {
lookupFn := func(ctx context.Context, fileId string) ([]string, error) {
return []string{"http://volume/" + fileId}, nil
}
rc := NewReaderCache(10, newMockChunkCacheForReaderCache(), lookupFn, nil)
defer rc.destroy()
fetchStarted := make(chan struct{})
fetchEnded := make(chan error, 1)
rc.fetchChunkDataFn = func(ctx context.Context, buffer []byte, urlStrings []string, cipherKey []byte, isGzipped bool, isFullChunk bool, offset int64, fileId string, refreshUrls util_http.RefreshUrlsFunc) (int, error) {
close(fetchStarted)
select {
case <-ctx.Done():
fetchEnded <- ctx.Err()
return 0, ctx.Err()
case <-time.After(10 * time.Second):
fetchEnded <- nil
return 0, fmt.Errorf("volume server never answered")
}
}
// a gzipped chunk is read through the shared whole-chunk download
var stream chunkStream
ctx, cancel := context.WithCancel(context.Background())
readDone := make(chan error, 1)
go func() {
_, err := rc.readChunkAt(ctx, &stream, make([]byte, 100), "3077,02c3475d67aaf564", nil, true, 0, 4<<20, false)
readDone <- err
}()
<-fetchStarted
cancel()
if err := <-readDone; err != context.Canceled {
t.Fatalf("cancelled read: got %v, want context.Canceled", err)
}
released := make(chan struct{})
go func() {
rc.releaseStream(&stream)
close(released)
}()
select {
case <-released:
case <-time.After(2 * time.Second):
t.Fatal("releasing the stream waited for a download nobody wants")
}
select {
case err := <-fetchEnded:
if err != context.Canceled {
t.Fatalf("abandoned download: got %v, want context.Canceled", err)
}
case <-time.After(2 * time.Second):
t.Fatal("the abandoned download kept running")
}
}
// One stream leaving a shared download must not cancel it for another stream.
func TestAbandonedChunkDownloadKeepsRunningForOtherStream(t *testing.T) {
content := []byte("gzipped chunk 01")
fileId := "3077,02c3475d67aaf564"
lookupFn := func(ctx context.Context, fileId string) ([]string, error) {
return []string{"http://volume/" + fileId}, nil
}
rc := NewReaderCache(10, newMockChunkCacheForReaderCache(), lookupFn, nil)
defer rc.destroy()
fetchStarted := make(chan struct{})
finishFetch := make(chan struct{})
rc.fetchChunkDataFn = func(ctx context.Context, buffer []byte, urlStrings []string, cipherKey []byte, isGzipped bool, isFullChunk bool, offset int64, fileId string, refreshUrls util_http.RefreshUrlsFunc) (int, error) {
close(fetchStarted)
select {
case <-ctx.Done():
return 0, ctx.Err()
case <-finishFetch:
return copy(buffer, content), nil
}
}
read := func(ctx context.Context, stream *chunkStream, done chan<- error, buffer []byte) {
_, err := rc.readChunkAt(ctx, stream, buffer, fileId, nil, true, 0, len(content), false)
done <- err
}
var leaving, staying chunkStream
leavingCtx, cancel := context.WithCancel(context.Background())
leavingDone, stayingDone := make(chan error, 1), make(chan error, 1)
stayingBuffer := make([]byte, len(content))
go read(leavingCtx, &leaving, leavingDone, make([]byte, len(content)))
<-fetchStarted
go read(context.Background(), &staying, stayingDone, stayingBuffer)
// wait until the second stream is attached to the same download
deadline := time.Now().Add(2 * time.Second)
for {
rc.RLock()
cacher := rc.downloaders[fileId]
rc.RUnlock()
if cacher != nil && atomic.LoadInt32(&cacher.pins) == 2 {
break
}
if time.Now().After(deadline) {
t.Fatal("second stream never attached to the download")
}
time.Sleep(time.Millisecond)
}
cancel()
if err := <-leavingDone; err != context.Canceled {
t.Fatalf("cancelled read: got %v, want context.Canceled", err)
}
rc.releaseStream(&leaving)
close(finishFetch)
select {
case err := <-stayingDone:
if err != nil {
t.Fatalf("staying stream: got %v, want the chunk", err)
}
if string(stayingBuffer) != string(content) {
t.Fatalf("staying stream read %q, want %q", stayingBuffer, content)
}
case <-time.After(2 * time.Second):
t.Fatal("staying stream never got its data")
}
rc.releaseStream(&staying)
}
func TestAbandonedChunkDownloadKeepsRunningForWaitingReader(t *testing.T) {
content := []byte("gzipped chunk 01")
fileId := "3077,02c3475d67aaf564"
lookupFn := func(ctx context.Context, fileId string) ([]string, error) {
return []string{"http://volume/" + fileId}, nil
}
rc := NewReaderCache(10, newMockChunkCacheForReaderCache(), lookupFn, nil)
defer rc.destroy()
fetchStarted := make(chan struct{})
finishFetch := make(chan struct{})
rc.fetchChunkDataFn = func(ctx context.Context, buffer []byte, urlStrings []string, cipherKey []byte, isGzipped bool, isFullChunk bool, offset int64, fileId string, refreshUrls util_http.RefreshUrlsFunc) (int, error) {
close(fetchStarted)
select {
case <-ctx.Done():
return 0, ctx.Err()
case <-finishFetch:
return copy(buffer, content), nil
}
}
var leaving chunkStream
leavingCtx, cancel := context.WithCancel(context.Background())
leavingDone, waitingDone := make(chan error, 1), make(chan error, 1)
go func() {
_, err := rc.readChunkAt(leavingCtx, &leaving, make([]byte, len(content)), fileId, nil, true, 0, len(content), false)
leavingDone <- err
}()
<-fetchStarted
waitingBuffer := make([]byte, len(content))
go func() {
_, err := rc.ReadChunkAt(context.Background(), waitingBuffer, fileId, nil, true, 0, len(content), false)
waitingDone <- err
}()
deadline := time.Now().Add(2 * time.Second)
for {
rc.RLock()
cacher := rc.downloaders[fileId]
rc.RUnlock()
if cacher != nil && atomic.LoadInt32(&cacher.readers) == 2 {
break
}
if time.Now().After(deadline) {
t.Fatal("waiting reader never attached to the download")
}
time.Sleep(time.Millisecond)
}
cancel()
if err := <-leavingDone; err != context.Canceled {
t.Fatalf("cancelled read: got %v, want context.Canceled", err)
}
rc.releaseStream(&leaving)
close(finishFetch)
select {
case err := <-waitingDone:
if err != nil {
t.Fatalf("waiting reader: got %v, want the chunk", err)
}
if string(waitingBuffer) != string(content) {
t.Fatalf("waiting reader read %q, want %q", waitingBuffer, content)
}
case <-time.After(2 * time.Second):
t.Fatal("waiting reader never got its data")
}
}
// A download cancelled mid-copy must not cache the partly filled buffer.
func TestAbandonedChunkDownloadIsNotCached(t *testing.T) {
fileId := "3077,02c3475d67aaf564"
lookupFn := func(ctx context.Context, fileId string) ([]string, error) {
return []string{"http://volume/" + fileId}, nil
}
cache := newMockChunkCacheForReaderCache()
rc := NewReaderCache(10, cache, lookupFn, nil)
defer rc.destroy()
fetchStarted := make(chan struct{})
rc.fetchChunkDataFn = func(ctx context.Context, buffer []byte, urlStrings []string, cipherKey []byte, isGzipped bool, isFullChunk bool, offset int64, fileId string, refreshUrls util_http.RefreshUrlsFunc) (int, error) {
close(fetchStarted)
<-ctx.Done()
// like readEncryptedUrl: the body was read, the copy was skipped, no error
return 0, nil
}
var stream chunkStream
ctx, cancel := context.WithCancel(context.Background())
readDone := make(chan error, 1)
go func() {
_, err := rc.readChunkAt(ctx, &stream, make([]byte, 100), fileId, nil, true, 0, 100, true)
readDone <- err
}()
<-fetchStarted
cancel()
if err := <-readDone; err != context.Canceled {
t.Fatalf("cancelled read: got %v, want context.Canceled", err)
}
rc.releaseStream(&stream)
cache.mu.Lock()
defer cache.mu.Unlock()
if _, cached := cache.data[fileId]; cached {
t.Fatal("the partly filled buffer of a cancelled download was cached")
}
}
// Other readers can join a re-lookup through FilerClient, so cancelling the download must not cancel it.
func TestAbandonedChunkDownloadKeepsItsRelookupAlive(t *testing.T) {
fileId := "3077,02c3475d67aaf564"
var lookups int32
relookupStarted := make(chan struct{})
downloadCancelled := make(chan struct{})
relookupErr := make(chan error, 1)
lookupFn := func(ctx context.Context, fileId string) ([]string, error) {
if atomic.AddInt32(&lookups, 1) > 1 {
close(relookupStarted)
<-downloadCancelled
relookupErr <- ctx.Err()
}
return []string{"http://volume/" + fileId}, nil
}
rc := NewReaderCache(10, newMockChunkCacheForReaderCache(), lookupFn, &mockCacheInvalidatorForReaderCache{})
defer rc.destroy()
rc.fetchChunkDataFn = func(ctx context.Context, buffer []byte, urlStrings []string, cipherKey []byte, isGzipped bool, isFullChunk bool, offset int64, fileId string, refreshUrls util_http.RefreshUrlsFunc) (int, error) {
return 0, fmt.Errorf("connection reset by peer")
}
var stream chunkStream
ctx, cancel := context.WithCancel(context.Background())
readDone := make(chan error, 1)
go func() {
_, err := rc.readChunkAt(ctx, &stream, make([]byte, 100), fileId, nil, true, 0, 100, false)
readDone <- err
}()
<-relookupStarted
cancel()
if err := <-readDone; err != context.Canceled {
t.Fatalf("cancelled read: got %v, want context.Canceled", err)
}
released := make(chan struct{})
go func() {
rc.releaseStream(&stream)
close(released)
}()
// the cacher leaves the map under the same lock that cancels it
deadline := time.Now().Add(2 * time.Second)
for {
rc.RLock()
_, found := rc.downloaders[fileId]
rc.RUnlock()
if !found {
break
}
if time.Now().After(deadline) {
t.Fatal("releasing the last stream never dropped the download")
}
time.Sleep(time.Millisecond)
}
close(downloadCancelled)
if err := <-relookupErr; err != nil {
t.Fatalf("re-lookup context: got %v, want it independent of the download", err)
}
select {
case <-released:
case <-time.After(2 * time.Second):
t.Fatal("releasing the stream never returned")
}
}
// Closing the reader cache cancels a stalled prefetch nobody reads instead of waiting for it.
func TestReaderCacheDestroyCancelsUnreadPrefetch(t *testing.T) {
lookupFn := func(ctx context.Context, fileId string) ([]string, error) {
return []string{"http://volume/" + fileId}, nil
}
rc := NewReaderCache(10, newMockChunkCacheForReaderCache(), lookupFn, nil)
fetchStarted := make(chan struct{})
rc.fetchChunkDataFn = func(ctx context.Context, buffer []byte, urlStrings []string, cipherKey []byte, isGzipped bool, isFullChunk bool, offset int64, fileId string, refreshUrls util_http.RefreshUrlsFunc) (int, error) {
close(fetchStarted)
select {
case <-ctx.Done():
return 0, ctx.Err()
case <-time.After(10 * time.Second):
return 0, fmt.Errorf("volume server never answered")
}
}
rc.MaybeCache(&Interval[*ChunkView]{Value: &ChunkView{FileId: "3077,02c3475d67aaf564", ViewSize: 1024, ChunkSize: 1024}}, 1)
<-fetchStarted
destroyed := make(chan struct{})
go func() {
rc.destroy()
close(destroyed)
}()
select {
case <-destroyed:
case <-time.After(2 * time.Second):
t.Fatal("destroy waited for a prefetch nobody reads")
}
}
// A download destroy dropped while a stream still held it is cancelled once that stream leaves.
func TestReaderCacheDestroyCancelsDownloadAfterItsLastStream(t *testing.T) {
lookupFn := func(ctx context.Context, fileId string) ([]string, error) {
return []string{"http://volume/" + fileId}, nil
}
rc := NewReaderCache(10, newMockChunkCacheForReaderCache(), lookupFn, nil)
fetchStarted := make(chan struct{})
fetchEnded := make(chan error, 1)
rc.fetchChunkDataFn = func(ctx context.Context, buffer []byte, urlStrings []string, cipherKey []byte, isGzipped bool, isFullChunk bool, offset int64, fileId string, refreshUrls util_http.RefreshUrlsFunc) (int, error) {
close(fetchStarted)
select {
case <-ctx.Done():
fetchEnded <- ctx.Err()
return 0, ctx.Err()
case <-time.After(10 * time.Second):
fetchEnded <- nil
return 0, fmt.Errorf("volume server never answered")
}
}
var stream chunkStream
ctx, cancel := context.WithCancel(context.Background())
readDone := make(chan error, 1)
go func() {
_, err := rc.readChunkAt(ctx, &stream, make([]byte, 100), "3077,02c3475d67aaf564", nil, true, 0, 4<<20, false)
readDone <- err
}()
<-fetchStarted
cancel()
if err := <-readDone; err != context.Canceled {
t.Fatalf("cancelled read: got %v, want context.Canceled", err)
}
destroyed := make(chan struct{})
go func() {
rc.destroy()
close(destroyed)
}()
deadline := time.Now().Add(2 * time.Second)
for {
rc.RLock()
mapped := len(rc.downloaders)
rc.RUnlock()
if mapped == 0 {
break
}
if time.Now().After(deadline) {
t.Fatal("destroy never dropped the download from the map")
}
time.Sleep(time.Millisecond)
}
select {
case err := <-fetchEnded:
t.Fatalf("destroy cancelled a download a stream still holds: %v", err)
case <-time.After(50 * time.Millisecond):
}
rc.releaseStream(&stream)
select {
case <-destroyed:
case <-time.After(2 * time.Second):
t.Fatal("destroy kept waiting for a download its last stream left")
}
if err := <-fetchEnded; err != context.Canceled {
t.Fatalf("orphaned download: got %v, want context.Canceled", err)
}
}