mirror of
https://github.com/seaweedfs/seaweedfs.git
synced 2026-08-16 12:16:36 +00:00
* mount: invalidate hot directory listings by section A cached directory used to be dropped whole when it saw 64 changes in 2s: with a continuous writer the listing cycled through wipe, direct listing and full rebuild for as long as the writer kept going, and every sibling lookup fell through to the filer in between. Split each cached listing into name-range sections of 1024 entries. A burst of foreign changes invalidates just the section it lands in; entries stay served and events keep applying, and the next readdir re-lists only that range from the filer, reconciled through the version gate so it cannot roll back newer applied events. Lookups in an invalidated section read through until then. The mount's own writes no longer invalidate anything: they are ground truth for its cache. * meta_cache: drop the version floor with a deleted or moved directory The other teardown paths already clear both maps; a floor left behind here would fence the listing of a directory re-created at the same path. * mount: harden section refresh An unversioned listing (pre-upgrade filer) now only fills gaps instead of reconciling: without a snapshot to order against, an overwrite or the deletion sweep could roll back an event applied after the listing. The section table can be rebuilt or re-split between the listing and its apply, so the refresh only marks fresh or splits when the section still covers the range it read. Splicing bounds from a stale range into a rebuilt table could leave them unsorted. Bound the wait: a readdir gives a refresh five seconds before serving the maintained-but-unverified cache. Bound the size: a range grown past four sections aborts the refresh and drops the directory cache, re-tiling it with a full rebuild, with that request served direct. Cover the filer-facing path with a listing server: paging with the snapshot pinned across pages, the section cutoff, no calls for a fresh section, and the overgrown-range abort. * meta_cache: make the section table a self-contained state machine Churn counting, freshness, stale-range scanning and the refresh completion with its guard and split now live on dirSections itself, free of the lock, the store and the apply loop, so they test directly with synthetic clocks and tables. MetaCache keeps thin wrappers that hold its mutex and find the directory's table. * meta_cache: keep section internals out of the apply request The request now carries the completed build's table and one refresh as opaque values built by section code, and the boundary-derivation rule moves out of the build loop into a collector next to the rest of the section logic. * mount: fence refreshed sections with a snapshot floor A refresh versioned the entries it fetched and tombstoned the ones it swept, but a name absent from both cache and listing kept the old directory floor, so a delayed event between the two snapshots could resurrect it into a section already marked fresh. The section now carries its own floor, consulted next to the directory floor, covering every name in the range, present or absent — which also retires the refresh's per-entry version stamps and sweep tombstones. An unversioned listing sets no floor and vouches for nothing: it may still fill gaps, but the section stays stale and reads through until a filer that stamps snapshots re-validates it. A listing's reach is unknowable up front — a resumed handle can skip far ahead, and shrunken sections let one batch span many — so a readdir now re-validates every stale section from its start name to the end of the directory instead of the next two. * mount: fence tombstoned names with floors and gate the reconcile A tombstone answered for its name before the floors were consulted, so one at an old position let through events the newer listing floor should have fenced; a build never hit this because it prunes superseded tombstones, which a section refresh does not. The version gate now raises a tombstone to the floors like any other record. With no per-entry versions, only the section floor fences a reconcile's work, so a range the rebuilt or re-split table no longer has must not touch the store either: the range check moves ahead of the mutations, under the same lock the floor install holds. An unversioned refresh no longer retries: the section is remembered as unverifiable and skipped by the stale scan, or every batch of every readdir would re-list the same ranges against a filer that cannot vouch for them. * mount: clear beaten unversioned markers and skip refresh mid-build An unversioned marker outliving the snapshot write that replaced its content bypassed the section floor the same way an old tombstone did, letting a delayed pre-snapshot event roll the entry back. The refresh now clears the marker when its write wins; pinned local-only entries are not replaced at all, keeping their content and marker. A rebuild wipes and repopulates the store off the apply loop, so a refresh reconciling meanwhile could sweep children the build had already inserted and let it publish the directory incomplete. The refresh now skips a building directory, as events (buffered) and purges (skipped) already do; its staleness dies with the build's fresh table. * mount: clear the unversioned marker only after its replacement lands Clearing before the insert meant a failed write left the old local content claiming the listing floors, fencing the very events that were still entitled to correct it. * meta_cache: rename the section state machine to sectionList dirSections named both the type and the map of them. * mount: raise the default cacheDirMaxEntries to 100000 The low ceiling guarded against whole-listing rebuild churn: a big cached directory under writes kept re-streaming everything. Sectioned invalidation ended that — a burst now costs one range listing — so the remaining cost of caching a large directory is its one-time build, comparable to the single direct listing that read-through mode pays on every enumeration instead. * meta_cache: cover section border and edge cases A bound-named entry belongs to the section starting at the bound: the neighboring refresh's sweep stops before it, its own section's covers it. Churn past everything the build saw lands in the tail section, a rename spanning two sections invalidates both, and a listed entry at the section's end name is cut off with the ones beyond it.
636 lines
21 KiB
Go
636 lines
21 KiB
Go
package meta_cache
|
|
|
|
import (
|
|
"context"
|
|
"fmt"
|
|
"io"
|
|
"sync"
|
|
"testing"
|
|
"time"
|
|
|
|
"github.com/seaweedfs/seaweedfs/weed/filer"
|
|
"github.com/seaweedfs/seaweedfs/weed/pb/filer_pb"
|
|
"github.com/seaweedfs/seaweedfs/weed/util"
|
|
"google.golang.org/grpc"
|
|
"google.golang.org/grpc/metadata"
|
|
)
|
|
|
|
type buildListStream struct {
|
|
responses []*filer_pb.ListEntriesResponse
|
|
onFirstRecv func()
|
|
once sync.Once
|
|
index int
|
|
}
|
|
|
|
func (s *buildListStream) Recv() (*filer_pb.ListEntriesResponse, error) {
|
|
s.once.Do(func() {
|
|
if s.onFirstRecv != nil {
|
|
s.onFirstRecv()
|
|
}
|
|
})
|
|
if s.index >= len(s.responses) {
|
|
return nil, io.EOF
|
|
}
|
|
resp := s.responses[s.index]
|
|
s.index++
|
|
return resp, nil
|
|
}
|
|
|
|
func (s *buildListStream) Header() (metadata.MD, error) { return metadata.MD{}, nil }
|
|
func (s *buildListStream) Trailer() metadata.MD { return metadata.MD{} }
|
|
func (s *buildListStream) CloseSend() error { return nil }
|
|
func (s *buildListStream) Context() context.Context { return context.Background() }
|
|
func (s *buildListStream) SendMsg(any) error { return nil }
|
|
func (s *buildListStream) RecvMsg(any) error { return nil }
|
|
|
|
type buildListClient struct {
|
|
filer_pb.SeaweedFilerClient
|
|
responses []*filer_pb.ListEntriesResponse
|
|
onFirstRecv func()
|
|
}
|
|
|
|
func (c *buildListClient) ListEntries(ctx context.Context, in *filer_pb.ListEntriesRequest, opts ...grpc.CallOption) (grpc.ServerStreamingClient[filer_pb.ListEntriesResponse], error) {
|
|
return &buildListStream{
|
|
responses: c.responses,
|
|
onFirstRecv: c.onFirstRecv,
|
|
}, nil
|
|
}
|
|
|
|
type buildFilerAccessor struct {
|
|
client filer_pb.SeaweedFilerClient
|
|
}
|
|
|
|
func (a *buildFilerAccessor) WithFilerClient(_ bool, fn func(filer_pb.SeaweedFilerClient) error) error {
|
|
return fn(a.client)
|
|
}
|
|
|
|
func (a *buildFilerAccessor) AdjustedUrl(*filer_pb.Location) string { return "" }
|
|
func (a *buildFilerAccessor) GetDataCenter() string { return "" }
|
|
|
|
func TestEnsureVisitedReplaysBufferedEventsAfterSnapshot(t *testing.T) {
|
|
mc, _, _, _ := newTestMetaCache(t, map[util.FullPath]bool{
|
|
"/": true,
|
|
})
|
|
defer mc.Shutdown()
|
|
|
|
var applyErr error
|
|
accessor := &buildFilerAccessor{
|
|
client: &buildListClient{
|
|
responses: []*filer_pb.ListEntriesResponse{
|
|
{
|
|
Entry: &filer_pb.Entry{
|
|
Name: "base.txt",
|
|
Attributes: &filer_pb.FuseAttributes{
|
|
Crtime: 1,
|
|
Mtime: 1,
|
|
FileMode: 0100644,
|
|
FileSize: 3,
|
|
},
|
|
},
|
|
SnapshotTsNs: 100,
|
|
},
|
|
},
|
|
onFirstRecv: func() {
|
|
applyErr = mc.ApplyMetadataResponse(context.Background(), &filer_pb.SubscribeMetadataResponse{
|
|
Directory: "/dir",
|
|
EventNotification: &filer_pb.EventNotification{
|
|
NewEntry: &filer_pb.Entry{
|
|
Name: "after.txt",
|
|
Attributes: &filer_pb.FuseAttributes{
|
|
Crtime: 2,
|
|
Mtime: 2,
|
|
FileMode: 0100644,
|
|
FileSize: 9,
|
|
},
|
|
},
|
|
},
|
|
TsNs: 101,
|
|
}, SubscriberMetadataResponseApplyOptions)
|
|
},
|
|
},
|
|
}
|
|
|
|
if err := EnsureVisited(mc, accessor, util.FullPath("/dir"), 0); err != nil {
|
|
t.Fatalf("ensure visited: %v", err)
|
|
}
|
|
if applyErr != nil {
|
|
t.Fatalf("apply buffered event: %v", applyErr)
|
|
}
|
|
if !mc.IsDirectoryCached(util.FullPath("/dir")) {
|
|
t.Fatal("directory /dir should be cached after build completes")
|
|
}
|
|
|
|
baseEntry, _, err := mc.FindEntry(context.Background(), util.FullPath("/dir/base.txt"))
|
|
if err != nil {
|
|
t.Fatalf("find base entry: %v", err)
|
|
}
|
|
if baseEntry.FileSize != 3 {
|
|
t.Fatalf("base entry size = %d, want 3", baseEntry.FileSize)
|
|
}
|
|
|
|
afterEntry, _, err := mc.FindEntry(context.Background(), util.FullPath("/dir/after.txt"))
|
|
if err != nil {
|
|
t.Fatalf("find replayed entry: %v", err)
|
|
}
|
|
if afterEntry.FileSize != 9 {
|
|
t.Fatalf("replayed entry size = %d, want 9", afterEntry.FileSize)
|
|
}
|
|
}
|
|
|
|
// TestDirectoryNotificationsSuppressedDuringBuild verifies that metadata events
|
|
// targeting a directory under active build do NOT fire onDirectoryUpdate for
|
|
// that directory. In production, onDirectoryUpdate can trigger
|
|
// markDirectoryReadThrough → DeleteFolderChildren, which would wipe entries
|
|
// that EnsureVisited already inserted mid-build.
|
|
func TestDirectoryNotificationsSuppressedDuringBuild(t *testing.T) {
|
|
mc, _, notifications, _ := newTestMetaCache(t, map[util.FullPath]bool{
|
|
"/": true,
|
|
})
|
|
defer mc.Shutdown()
|
|
|
|
// Start building /dir (simulates the beginning of EnsureVisited)
|
|
if err := mc.BeginDirectoryBuild(context.Background(), util.FullPath("/dir")); err != nil {
|
|
t.Fatalf("begin build: %v", err)
|
|
}
|
|
|
|
// Insert an entry as EnsureVisited would during the filer listing
|
|
if err := mc.InsertEntry(context.Background(), &filer.Entry{
|
|
FullPath: "/dir/existing.txt",
|
|
Attr: filer.Attr{
|
|
Crtime: time.Unix(1, 0),
|
|
Mtime: time.Unix(1, 0),
|
|
Mode: 0100644,
|
|
FileSize: 100,
|
|
},
|
|
}, 0); err != nil {
|
|
t.Fatalf("insert entry during build: %v", err)
|
|
}
|
|
|
|
// Simulate multiple metadata events arriving for /dir while the build
|
|
// is in progress. Each event would normally call noteDirectoryUpdate,
|
|
// which in production can trigger markDirectoryReadThrough and wipe entries.
|
|
for i := 0; i < 5; i++ {
|
|
resp := &filer_pb.SubscribeMetadataResponse{
|
|
Directory: "/dir",
|
|
EventNotification: &filer_pb.EventNotification{
|
|
NewEntry: &filer_pb.Entry{
|
|
Name: fmt.Sprintf("new-%d.txt", i),
|
|
Attributes: &filer_pb.FuseAttributes{
|
|
Crtime: int64(10 + i),
|
|
Mtime: int64(10 + i),
|
|
FileMode: 0100644,
|
|
FileSize: uint64(i + 1),
|
|
},
|
|
},
|
|
},
|
|
TsNs: int64(200 + i),
|
|
}
|
|
if err := mc.ApplyMetadataResponse(context.Background(), resp, SubscriberMetadataResponseApplyOptions); err != nil {
|
|
t.Fatalf("apply event %d: %v", i, err)
|
|
}
|
|
}
|
|
|
|
// The building directory /dir must NOT have received any notifications.
|
|
// If it did, markDirectoryReadThrough would wipe the cache mid-build.
|
|
for _, p := range notifications.paths() {
|
|
if p == util.FullPath("/dir") {
|
|
t.Fatal("onDirectoryUpdate was called for /dir during build; this would cause markDirectoryReadThrough to wipe entries mid-build")
|
|
}
|
|
}
|
|
|
|
// The entry inserted during the build must still be present
|
|
entry, _, err := mc.FindEntry(context.Background(), util.FullPath("/dir/existing.txt"))
|
|
if err != nil {
|
|
t.Fatalf("entry wiped during build: %v", err)
|
|
}
|
|
if entry.FileSize != 100 {
|
|
t.Fatalf("entry size = %d, want 100", entry.FileSize)
|
|
}
|
|
|
|
// Complete the build — buffered events should be replayed
|
|
if err := mc.CompleteDirectoryBuild(context.Background(), util.FullPath("/dir"), 150, nil); err != nil {
|
|
t.Fatalf("complete build: %v", err)
|
|
}
|
|
|
|
// After build completes, the entry from the listing should still exist
|
|
entry, _, err = mc.FindEntry(context.Background(), util.FullPath("/dir/existing.txt"))
|
|
if err != nil {
|
|
t.Fatalf("entry lost after build completion: %v", err)
|
|
}
|
|
if entry.FileSize != 100 {
|
|
t.Fatalf("entry size after build = %d, want 100", entry.FileSize)
|
|
}
|
|
|
|
// Buffered events with TsNs > snapshotTsNs (150) should have been replayed
|
|
for i := 0; i < 5; i++ {
|
|
name := fmt.Sprintf("new-%d.txt", i)
|
|
e, _, err := mc.FindEntry(context.Background(), util.FullPath("/dir/"+name))
|
|
if err != nil {
|
|
t.Fatalf("replayed entry %s not found: %v", name, err)
|
|
}
|
|
if e.FileSize != uint64(i+1) {
|
|
t.Fatalf("replayed entry %s size = %d, want %d", name, e.FileSize, i+1)
|
|
}
|
|
}
|
|
}
|
|
|
|
// TestEmptyDirectoryBuildReplaysAllBufferedEvents verifies that when a
|
|
// directory build completes with snapshotTsNs=0 (empty directory — server
|
|
// returned no entries and no snapshot), ALL buffered events are replayed
|
|
// without any TsNs filtering. This prevents clock-skew between client and
|
|
// filer from dropping legitimate mutations.
|
|
func TestEmptyDirectoryBuildReplaysAllBufferedEvents(t *testing.T) {
|
|
mc, _, _, _ := newTestMetaCache(t, map[util.FullPath]bool{
|
|
"/": true,
|
|
})
|
|
defer mc.Shutdown()
|
|
|
|
if err := mc.BeginDirectoryBuild(context.Background(), util.FullPath("/empty")); err != nil {
|
|
t.Fatalf("begin build: %v", err)
|
|
}
|
|
|
|
// Buffer events with a range of TsNs values — some very old, some recent.
|
|
// With a client-synthesized snapshot, old events could be incorrectly filtered.
|
|
tsValues := []int64{1, 50, 500, 5000, 50000}
|
|
for i, ts := range tsValues {
|
|
resp := &filer_pb.SubscribeMetadataResponse{
|
|
Directory: "/empty",
|
|
EventNotification: &filer_pb.EventNotification{
|
|
NewEntry: &filer_pb.Entry{
|
|
Name: fmt.Sprintf("file-%d.txt", i),
|
|
Attributes: &filer_pb.FuseAttributes{
|
|
Crtime: ts,
|
|
Mtime: ts,
|
|
FileMode: 0100644,
|
|
FileSize: uint64(i + 10),
|
|
},
|
|
},
|
|
},
|
|
TsNs: ts,
|
|
}
|
|
if err := mc.ApplyMetadataResponse(context.Background(), resp, SubscriberMetadataResponseApplyOptions); err != nil {
|
|
t.Fatalf("apply event %d: %v", i, err)
|
|
}
|
|
}
|
|
|
|
// Complete with snapshotTsNs=0 — simulates empty directory listing
|
|
if err := mc.CompleteDirectoryBuild(context.Background(), util.FullPath("/empty"), 0, nil); err != nil {
|
|
t.Fatalf("complete build: %v", err)
|
|
}
|
|
|
|
// Every buffered event must have been replayed, regardless of TsNs
|
|
for i := range tsValues {
|
|
name := fmt.Sprintf("file-%d.txt", i)
|
|
e, _, err := mc.FindEntry(context.Background(), util.FullPath("/empty/"+name))
|
|
if err != nil {
|
|
t.Fatalf("replayed entry %s not found: %v", name, err)
|
|
}
|
|
if e.FileSize != uint64(i+10) {
|
|
t.Fatalf("replayed entry %s size = %d, want %d", name, e.FileSize, i+10)
|
|
}
|
|
}
|
|
|
|
if !mc.IsDirectoryCached(util.FullPath("/empty")) {
|
|
t.Fatal("/empty should be marked cached after build completes")
|
|
}
|
|
}
|
|
|
|
// TestBuildCompletionSurvivesCallerCancellation verifies that once
|
|
// CompleteDirectoryBuild is enqueued, a cancelled caller context does not
|
|
// prevent the build from completing. The apply loop uses context.Background()
|
|
// internally, so the operation finishes even if the caller gives up waiting.
|
|
func TestBuildCompletionSurvivesCallerCancellation(t *testing.T) {
|
|
mc, _, _, _ := newTestMetaCache(t, map[util.FullPath]bool{
|
|
"/": true,
|
|
})
|
|
defer mc.Shutdown()
|
|
|
|
if err := mc.BeginDirectoryBuild(context.Background(), util.FullPath("/dir")); err != nil {
|
|
t.Fatalf("begin build: %v", err)
|
|
}
|
|
|
|
// Insert an entry during the build (as EnsureVisited would)
|
|
if err := mc.InsertEntry(context.Background(), &filer.Entry{
|
|
FullPath: "/dir/kept.txt",
|
|
Attr: filer.Attr{
|
|
Crtime: time.Unix(1, 0),
|
|
Mtime: time.Unix(1, 0),
|
|
Mode: 0100644,
|
|
FileSize: 42,
|
|
},
|
|
}, 0); err != nil {
|
|
t.Fatalf("insert entry: %v", err)
|
|
}
|
|
|
|
// Buffer an event that should be replayed
|
|
if err := mc.ApplyMetadataResponse(context.Background(), &filer_pb.SubscribeMetadataResponse{
|
|
Directory: "/dir",
|
|
EventNotification: &filer_pb.EventNotification{
|
|
NewEntry: &filer_pb.Entry{
|
|
Name: "buffered.txt",
|
|
Attributes: &filer_pb.FuseAttributes{
|
|
Crtime: 5,
|
|
Mtime: 5,
|
|
FileMode: 0100644,
|
|
FileSize: 77,
|
|
},
|
|
},
|
|
},
|
|
TsNs: 200,
|
|
}, SubscriberMetadataResponseApplyOptions); err != nil {
|
|
t.Fatalf("apply event: %v", err)
|
|
}
|
|
|
|
// Complete with an already-cancelled context. The operation should still
|
|
// succeed because enqueueAndWait sets req.ctx = context.Background().
|
|
cancelledCtx, cancel := context.WithCancel(context.Background())
|
|
cancel() // cancel immediately
|
|
|
|
// CompleteDirectoryBuild may return ctx.Err() if the select picks
|
|
// ctx.Done() first, but the operation itself still completes in the
|
|
// apply loop. Poll for the observable side effect instead of using
|
|
// a fixed sleep.
|
|
_ = mc.CompleteDirectoryBuild(cancelledCtx, util.FullPath("/dir"), 100, nil)
|
|
|
|
// Poll until the build completes or a deadline elapses.
|
|
deadline := time.After(2 * time.Second)
|
|
for !mc.IsDirectoryCached(util.FullPath("/dir")) {
|
|
select {
|
|
case <-deadline:
|
|
t.Fatal("/dir should be cached — CompleteDirectoryBuild must have executed despite cancelled context")
|
|
default:
|
|
time.Sleep(5 * time.Millisecond)
|
|
}
|
|
}
|
|
|
|
// The pre-existing entry must survive
|
|
entry, _, findErr := mc.FindEntry(context.Background(), util.FullPath("/dir/kept.txt"))
|
|
if findErr != nil {
|
|
t.Fatalf("find kept entry: %v", findErr)
|
|
}
|
|
if entry.FileSize != 42 {
|
|
t.Fatalf("kept entry size = %d, want 42", entry.FileSize)
|
|
}
|
|
|
|
// The buffered event (TsNs 200 > snapshot 100) must have been replayed
|
|
buffered, _, findErr := mc.FindEntry(context.Background(), util.FullPath("/dir/buffered.txt"))
|
|
if findErr != nil {
|
|
t.Fatalf("find buffered entry: %v", findErr)
|
|
}
|
|
if buffered.FileSize != 77 {
|
|
t.Fatalf("buffered entry size = %d, want 77", buffered.FileSize)
|
|
}
|
|
}
|
|
|
|
func TestBufferedRenameUpdatesOtherDirectoryBeforeBuildCompletes(t *testing.T) {
|
|
mc, _, _, _ := newTestMetaCache(t, map[util.FullPath]bool{
|
|
"/": true,
|
|
"/src": true,
|
|
})
|
|
defer mc.Shutdown()
|
|
|
|
if err := mc.InsertEntry(context.Background(), &filer.Entry{
|
|
FullPath: "/src/from.txt",
|
|
Attr: filer.Attr{
|
|
Crtime: time.Unix(1, 0),
|
|
Mtime: time.Unix(1, 0),
|
|
Mode: 0100644,
|
|
FileSize: 7,
|
|
},
|
|
}, 0); err != nil {
|
|
t.Fatalf("insert source entry: %v", err)
|
|
}
|
|
|
|
if err := mc.BeginDirectoryBuild(context.Background(), util.FullPath("/dst")); err != nil {
|
|
t.Fatalf("begin build: %v", err)
|
|
}
|
|
|
|
renameResp := &filer_pb.SubscribeMetadataResponse{
|
|
Directory: "/src",
|
|
EventNotification: &filer_pb.EventNotification{
|
|
OldEntry: &filer_pb.Entry{
|
|
Name: "from.txt",
|
|
},
|
|
NewEntry: &filer_pb.Entry{
|
|
Name: "to.txt",
|
|
Attributes: &filer_pb.FuseAttributes{
|
|
Crtime: 2,
|
|
Mtime: 2,
|
|
FileMode: 0100644,
|
|
FileSize: 12,
|
|
},
|
|
},
|
|
NewParentPath: "/dst",
|
|
},
|
|
TsNs: 101,
|
|
}
|
|
|
|
if err := mc.ApplyMetadataResponse(context.Background(), renameResp, SubscriberMetadataResponseApplyOptions); err != nil {
|
|
t.Fatalf("apply rename: %v", err)
|
|
}
|
|
|
|
oldEntry, _, err := mc.FindEntry(context.Background(), util.FullPath("/src/from.txt"))
|
|
if err != filer_pb.ErrNotFound {
|
|
t.Fatalf("find old path error = %v, want %v", err, filer_pb.ErrNotFound)
|
|
}
|
|
if oldEntry != nil {
|
|
t.Fatalf("old path should be removed before build completes: %+v", oldEntry)
|
|
}
|
|
|
|
newEntry, _, err := mc.FindEntry(context.Background(), util.FullPath("/dst/to.txt"))
|
|
if err != filer_pb.ErrNotFound {
|
|
t.Fatalf("find buffered new path error = %v, want %v", err, filer_pb.ErrNotFound)
|
|
}
|
|
if newEntry != nil {
|
|
t.Fatalf("new path should stay hidden until build completes: %+v", newEntry)
|
|
}
|
|
|
|
if err := mc.CompleteDirectoryBuild(context.Background(), util.FullPath("/dst"), 100, nil); err != nil {
|
|
t.Fatalf("complete build: %v", err)
|
|
}
|
|
|
|
newEntry, _, err = mc.FindEntry(context.Background(), util.FullPath("/dst/to.txt"))
|
|
if err != nil {
|
|
t.Fatalf("find replayed new path: %v", err)
|
|
}
|
|
if newEntry.FileSize != 12 {
|
|
t.Fatalf("replayed new path size = %d, want 12", newEntry.FileSize)
|
|
}
|
|
}
|
|
|
|
// TestEnsureVisitedPreservesLocalOnlyEntry reproduces the residual coherence
|
|
// gap behind the FUSE ConcurrentReadWrite ENOENT flake.
|
|
//
|
|
// A FUSE create on the writeback/deferFilerCreate path inserts the entry into
|
|
// the local store directly (weedfs_file_mkrm.go createFile), off the metaCache
|
|
// apply loop, before the filer holds it. A concurrent rebuild of the parent —
|
|
// triggered when the directory falls out of cache (idle evict, hot-dir
|
|
// read-through) — wipes the store and refills it from a filer listing that does
|
|
// not yet include the un-flushed local create, then publishes the directory
|
|
// authoritatively cached (markCachedFn). The local entry vanishes although the
|
|
// client created it: lookupEntry then returns an authoritative ENOENT for it.
|
|
func TestEnsureVisitedPreservesLocalOnlyEntry(t *testing.T) {
|
|
mc, _, _, _ := newTestMetaCache(t, map[util.FullPath]bool{"/": true})
|
|
defer mc.Shutdown()
|
|
|
|
// The mount pins the un-flushed create (open dirty handle / pending flush),
|
|
// keyed off the inode the entry carries so a kernel Forget that dropped the
|
|
// path→inode mapping cannot unpin an in-flight create.
|
|
mc.SetPinnedChildFn(func(e *filer.Entry) bool { return e.Attr.Inode == 42 })
|
|
|
|
// A deferred local create lands before the rebuild; /dir is not yet cached.
|
|
// It carries its allocated inode, as createFile's placeholder does.
|
|
if err := mc.InsertEntry(context.Background(), &filer.Entry{
|
|
FullPath: "/dir/pending.txt",
|
|
Attr: filer.Attr{Crtime: time.Unix(1, 0), Mtime: time.Unix(1, 0), Mode: 0100644, FileSize: 1, Inode: 42},
|
|
}, 0); err != nil {
|
|
t.Fatalf("insert pending entry: %v", err)
|
|
}
|
|
|
|
// A concurrent rebuild lists the filer, whose snapshot pre-dates the
|
|
// un-flushed create, so it returns only the already-persisted sibling.
|
|
accessor := &buildFilerAccessor{client: &buildListClient{
|
|
responses: []*filer_pb.ListEntriesResponse{{
|
|
Entry: &filer_pb.Entry{
|
|
Name: "base.txt",
|
|
Attributes: &filer_pb.FuseAttributes{
|
|
Crtime: 1,
|
|
Mtime: 1,
|
|
FileMode: 0100644,
|
|
FileSize: 3,
|
|
},
|
|
},
|
|
SnapshotTsNs: 100,
|
|
}},
|
|
}}
|
|
|
|
if err := EnsureVisited(mc, accessor, util.FullPath("/dir"), 0); err != nil {
|
|
t.Fatalf("ensure visited: %v", err)
|
|
}
|
|
if !mc.IsDirectoryCached(util.FullPath("/dir")) {
|
|
t.Fatal("/dir should be cached after build completes")
|
|
}
|
|
|
|
// base.txt from the listing is present.
|
|
if _, _, err := mc.FindEntry(context.Background(), util.FullPath("/dir/base.txt")); err != nil {
|
|
t.Fatalf("listed entry missing after build: %v", err)
|
|
}
|
|
|
|
// The un-flushed local create must survive the rebuild. With /dir now
|
|
// authoritatively cached, losing it is the file-vanishes flake.
|
|
if _, _, err := mc.FindEntry(context.Background(), util.FullPath("/dir/pending.txt")); err != nil {
|
|
t.Fatalf("local-only entry lost across concurrent rebuild: %v", err)
|
|
}
|
|
}
|
|
|
|
// TestEnsureVisitedDropsUnpinnedStaleEntry guards the preservation's selectivity:
|
|
// a cached child the filer listing no longer returns and that is NOT pinned must
|
|
// still be wiped, so the rebuild can't resurrect a deleted/renamed entry.
|
|
func TestEnsureVisitedDropsUnpinnedStaleEntry(t *testing.T) {
|
|
mc, _, _, _ := newTestMetaCache(t, map[util.FullPath]bool{"/": true})
|
|
defer mc.Shutdown()
|
|
|
|
mc.SetPinnedChildFn(func(*filer.Entry) bool { return false })
|
|
|
|
// A stale child sits in the cache; the filer no longer has it.
|
|
insertCacheEntry(t, mc, "/dir/stale.txt")
|
|
|
|
accessor := &buildFilerAccessor{client: &buildListClient{
|
|
responses: []*filer_pb.ListEntriesResponse{{
|
|
Entry: &filer_pb.Entry{
|
|
Name: "base.txt",
|
|
Attributes: &filer_pb.FuseAttributes{
|
|
Crtime: 1,
|
|
Mtime: 1,
|
|
FileMode: 0100644,
|
|
FileSize: 3,
|
|
},
|
|
},
|
|
SnapshotTsNs: 100,
|
|
}},
|
|
}}
|
|
|
|
if err := EnsureVisited(mc, accessor, util.FullPath("/dir"), 0); err != nil {
|
|
t.Fatalf("ensure visited: %v", err)
|
|
}
|
|
if entry, _, err := mc.FindEntry(context.Background(), util.FullPath("/dir/stale.txt")); err != filer_pb.ErrNotFound || entry != nil {
|
|
t.Fatalf("unpinned stale entry survived rebuild = %+v, %v; want nil, %v", entry, err, filer_pb.ErrNotFound)
|
|
}
|
|
}
|
|
|
|
// sequencedListClient returns a different ListEntries result per call (repeating
|
|
// the last), modelling a filer that lists empty transiently then the real entries.
|
|
type sequencedListClient struct {
|
|
filer_pb.SeaweedFilerClient
|
|
mu sync.Mutex
|
|
perCall [][]*filer_pb.ListEntriesResponse
|
|
calls int
|
|
}
|
|
|
|
func (c *sequencedListClient) ListEntries(ctx context.Context, in *filer_pb.ListEntriesRequest, opts ...grpc.CallOption) (grpc.ServerStreamingClient[filer_pb.ListEntriesResponse], error) {
|
|
c.mu.Lock()
|
|
idx := c.calls
|
|
if idx >= len(c.perCall) {
|
|
idx = len(c.perCall) - 1
|
|
}
|
|
c.calls++
|
|
resp := c.perCall[idx]
|
|
c.mu.Unlock()
|
|
return &buildListStream{responses: resp}, nil
|
|
}
|
|
|
|
// TestEnsureVisitedConfirmsTransientEmptyListing: a rebuild whose first filer
|
|
// listing comes back empty must re-read and cache the real entries, not strand
|
|
// the directory cached over an empty store (the ConcurrentReadWrite ENOENT flake).
|
|
func TestEnsureVisitedConfirmsTransientEmptyListing(t *testing.T) {
|
|
mc, _, _, _ := newTestMetaCache(t, map[util.FullPath]bool{"/": true})
|
|
defer mc.Shutdown()
|
|
|
|
entry := &filer_pb.Entry{
|
|
Name: "keep.txt",
|
|
Attributes: &filer_pb.FuseAttributes{Crtime: 1, Mtime: 1, FileMode: 0100644, FileSize: 7},
|
|
}
|
|
accessor := &buildFilerAccessor{client: &sequencedListClient{
|
|
perCall: [][]*filer_pb.ListEntriesResponse{
|
|
{}, // first read: transient empty
|
|
{{Entry: entry, SnapshotTsNs: 100}}, // confirm read: the real entry
|
|
},
|
|
}}
|
|
|
|
if err := EnsureVisited(mc, accessor, util.FullPath("/dir"), 0); err != nil {
|
|
t.Fatalf("ensure visited: %v", err)
|
|
}
|
|
if !mc.IsDirectoryCached(util.FullPath("/dir")) {
|
|
t.Fatal("/dir should be cached after build completes")
|
|
}
|
|
if _, _, err := mc.FindEntry(context.Background(), util.FullPath("/dir/keep.txt")); err != nil {
|
|
t.Fatalf("/dir/keep.txt stranded after transient empty listing: %v", err)
|
|
}
|
|
}
|
|
|
|
// TestEnsureVisitedCachesGenuinelyEmptyDirectory: a really-empty directory lists
|
|
// empty on every confirm and must still end up cached.
|
|
func TestEnsureVisitedCachesGenuinelyEmptyDirectory(t *testing.T) {
|
|
mc, _, _, _ := newTestMetaCache(t, map[util.FullPath]bool{"/": true})
|
|
defer mc.Shutdown()
|
|
|
|
client := &sequencedListClient{
|
|
perCall: [][]*filer_pb.ListEntriesResponse{{}}, // always empty
|
|
}
|
|
accessor := &buildFilerAccessor{client: client}
|
|
|
|
if err := EnsureVisited(mc, accessor, util.FullPath("/empty"), 0); err != nil {
|
|
t.Fatalf("ensure visited: %v", err)
|
|
}
|
|
if !mc.IsDirectoryCached(util.FullPath("/empty")) {
|
|
t.Fatal("/empty should be cached even though it has no entries")
|
|
}
|
|
// The empty result must have been confirmed, not trusted on the first read.
|
|
client.mu.Lock()
|
|
calls := client.calls
|
|
client.mu.Unlock()
|
|
if calls != emptyRebuildConfirmations+1 {
|
|
t.Fatalf("list calls = %d, want %d (initial + confirmations)", calls, emptyRebuildConfirmations+1)
|
|
}
|
|
}
|