Files
seaweedfs/weed/s3api/s3vectors/create_recovery_test.go
T
Chris Lu 8236c8d136 vector: integrate engine health, admin management, and local deployment (#12517)
* vector: expose request metrics and registration-aware health endpoints

* vector: report immutable read cache and storage fetch metrics

* docs: describe vector engine health, metrics, and maintenance

* vector: count canceled requests when their handler is dropped

* vector: account for uncached large immutable storage reads

* vector: reliably release the stalled credential reader in tests

* vector: observe large reads as their streamed body is consumed

* vector: bound heartbeats on established master registration streams

* test: exercise hardened vector maintenance on Go and Rust volumes

* vector: register engines in their configured filer group

* vector: make retried index creation preserve compatible datasets

* vector: allow registration to advertise a separate address

* test: release stalled credential readers without blocking FIFO cleanup

* test: allow hardened vector coverage in a named filer group

* vector: discover engines in the same filer group as their gateway

* vector: fall back when the advertised address is empty

* admin: expose vector readiness, maintenance, and management tools (#12520)

* s3vectors: share vector discovery with trusted management clients

* admin: show vector engine readiness and recent maintenance activity

* admin: suggest S3 Vectors IAM actions from the supported operations

* admin: manage vector indexes, policies, tags, encryption, and query data

* admin: validate SSE-S3 support before creating encrypted vector buckets

* admin: discover vector engines in their registration group

* admin: probe vector engines concurrently with independent time budgets

* admin: verify encryption capability on the remote S3 gateway

* admin: defer discovery until after the HTTP server can start

* admin: distinguish unknown vector health services from failed engines

* admin: count workers that can execute vector index maintenance

* admin: allow read-only users to query vectors and inspect tags

* vector: scope gateway and admin discovery to the configured filer group

* admin: refresh policy state after mutations and serialize policy controls

* vector: drain active RPCs before closing replaced discovery connections

* vector: respect request cancellation while waiting for discovery

* vector: prefer another registered engine after a transport failure

* admin: check all masters before reporting no registered vector engines

* iam: honor attached vector policies without broadening legacy grants

* vector: allow short-lived discovery clients to close their connections

* vector: refresh cached master addresses for each discovery lookup

* iam: enforce vector bucket tag conditions in attached policies

* iam: authorize vector bucket creation with the requested tags

* test: exercise vector IAM grants and tag denies on Go and Rust volumes

* admin: reject authenticated vector mutations without an admin role

* test: run vector IAM assertions before the shared engine is stopped

* vector: add native mini startup, shell tools, and a runnable Compose example (#12522)

* s3vectors: share vector discovery with trusted management clients

* admin: show vector engine readiness and recent maintenance activity

* admin: suggest S3 Vectors IAM actions from the supported operations

* admin: manage vector indexes, policies, tags, encryption, and query data

* admin: validate SSE-S3 support before creating encrypted vector buckets

* admin: discover vector engines in their registration group

* admin: probe vector engines concurrently with independent time budgets

* admin: verify encryption capability on the remote S3 gateway

* admin: defer discovery until after the HTTP server can start

* admin: distinguish unknown vector health services from failed engines

* admin: count workers that can execute vector index maintenance

* admin: allow read-only users to query vectors and inspect tags

* vector: scope gateway and admin discovery to the configured filer group

* admin: refresh policy state after mutations and serialize policy controls

* vector: drain active RPCs before closing replaced discovery connections

* vector: respect request cancellation while waiting for discovery

* vector: prefer another registered engine after a transport failure

* admin: check all masters before reporting no registered vector engines

* iam: honor attached vector policies without broadening legacy grants

* vector: allow short-lived discovery clients to close their connections

* vector: refresh cached master addresses for each discovery lookup

* iam: enforce vector bucket tag conditions in attached policies

* iam: authorize vector bucket creation with the requested tags

* test: exercise vector IAM grants and tag denies on Go and Rust volumes

* admin: reject authenticated vector mutations without an admin role

* test: run vector IAM assertions before the shared engine is stopped

* mini: optionally start and supervise the vector engine

* mini: bootstrap vector buckets with existing Lance catalog helpers

* shell: manage vector buckets, indexes, and vector data

* docs: provide a runnable vector and maintenance worker Compose stack

* shell: scope vector discovery to the session and close its connections

* mini: honor the bind address for vector and metrics listeners

* mini: reserve vector ports before assigning gRPC and compatibility listeners

* mini: wait for vector storage setup and master registration

* mini: create requested vector buckets without requiring the Lance HTTP listener

* shell: refresh master addresses during vector discovery

* mini: resolve sibling vector executables with platform extensions

* vector: preserve uncertain creates and fix mini listener allocation (#12525)

* vector: retain catalog records when index creation has an unknown outcome

* mini: keep chosen listener ports reserved throughout allocation

* mini: resolve hostname binds before starting vector listeners

* vector: fix integration test imports and recover interrupted index creation (#12533)

* shell: register vector operations from the command package

* s3vectors: resume interrupted index initialization safely

* mini: exclude disabled WebDAV from listener allocation

* s3vectors: retain uncertain creates across refused retries

* s3vectors: claim pending indexes before retrying initialization

* s3vectors: compare resolved distance metrics during recovery

* s3vectors: preserve metadata keys when retrying initialization

* s3vectors: verify ownership before initializing new catalog records
2026-11-10 04:21:18 -08:00

294 lines
13 KiB
Go

package s3vectors
import (
"context"
"encoding/json"
"net/http"
"net/http/httptest"
"slices"
"sync/atomic"
"testing"
"time"
"github.com/seaweedfs/seaweedfs/weed/pb"
"github.com/seaweedfs/seaweedfs/weed/pb/vector_pb"
"github.com/seaweedfs/seaweedfs/weed/s3api/s3tables"
"github.com/seaweedfs/seaweedfs/weed/s3api/s3tables/s3tablestest"
"google.golang.org/grpc"
"google.golang.org/grpc/codes"
"google.golang.org/grpc/status"
)
type failingCreateVectors struct {
*fakeVectors
err error
onCreate func()
}
func (f *failingCreateVectors) CreateIndex(_ context.Context, request *vector_pb.CreateIndexRequest, _ ...grpc.CallOption) (*vector_pb.CreateIndexResponse, error) {
f.created = append(f.created, request)
err := f.err
if f.onCreate != nil {
f.onCreate()
}
return nil, err
}
func TestCreateIndexPreservesAnUnknownOutcome(t *testing.T) {
for _, code := range []codes.Code{codes.Unavailable, codes.DeadlineExceeded, codes.Canceled, codes.Internal, codes.InvalidArgument} {
t.Run(code.String(), func(t *testing.T) {
fs := s3tablestest.Start(t)
engine := &failingCreateVectors{fakeVectors: &fakeVectors{}, err: status.Error(code, "response lost")}
server := NewServer(s3tables.NewManagerClient(fs.Client), allowAllAuth{}, engine)
postJSON(t, server.handleCreateVectorBucket, &CreateVectorBucketRequest{VectorBucketName: "vec-bucket"})
result := postJSON(t, server.handleCreateIndex, &CreateIndexRequest{VectorBucketName: "vec-bucket", IndexName: "idx", DataType: "float32", Dimension: 3, DistanceMetric: "cosine"})
if result.Code == http.StatusOK {
t.Fatal("failed engine reported success")
}
got := postJSON(t, server.handleGetIndex, &GetIndexRequest{VectorBucketName: "vec-bucket", IndexName: "idx"})
want := http.StatusOK
if code == codes.InvalidArgument {
want = http.StatusNotFound
}
if got.Code != want {
t.Fatalf("catalog after %s: %d %s", code, got.Code, got.Body)
}
})
}
}
func TestCreateIndexRetainsTheWriteErrorIfRediscoveryFails(t *testing.T) {
engine := &failingCreateVectors{fakeVectors: &fakeVectors{}, err: status.Error(codes.Unavailable, "response lost after commit")}
discovery := &vectorDiscovery{client: engine, nextResolve: time.Now().Add(time.Hour), discover: func([]pb.ServerAddress, grpc.DialOption) []string { return nil }}
_, err := discovery.CreateIndex(t.Context(), &vector_pb.CreateIndexRequest{})
if status.Convert(err).Message() != "response lost after commit" || len(engine.created) != 1 {
t.Fatalf("ambiguous result replaced by lookup failure: %v", err)
}
}
func TestCreateIndexResumesAnUnknownOutcome(t *testing.T) {
fs := s3tablestest.Start(t)
engine := &failingCreateVectors{fakeVectors: &fakeVectors{}, err: status.Error(codes.Unavailable, "request lost")}
server := NewServer(s3tables.NewManagerClient(fs.Client), allowAllAuth{}, engine)
postJSON(t, server.handleCreateVectorBucket, &CreateVectorBucketRequest{VectorBucketName: "vec-bucket"})
request := &CreateIndexRequest{VectorBucketName: "vec-bucket", IndexName: "idx", DataType: "float32", Dimension: 3, DistanceMetric: "cosine",
MetadataConfiguration: &MetadataConfiguration{NonFilterableMetadataKeys: []string{"b", "a"}}, Tags: map[string]string{"team": "search"}}
if result := postJSON(t, server.handleCreateIndex, request); result.Code == http.StatusOK {
t.Fatal("unavailable engine reported success")
}
before, err := server.indexRecord(t.Context(), "vec-bucket", "idx")
if err != nil {
t.Fatal(err)
}
for _, change := range []func(*CreateIndexRequest){
func(r *CreateIndexRequest) { r.Dimension++ },
func(r *CreateIndexRequest) { r.DistanceMetric = "euclidean" },
func(r *CreateIndexRequest) { r.MetadataConfiguration = nil },
} {
other := *request
change(&other)
if result := postJSON(t, server.handleCreateIndex, &other); result.Code != http.StatusConflict {
t.Fatalf("incompatible retry: %d %s", result.Code, result.Body)
}
}
if len(engine.created) != 1 {
t.Fatal("incompatible retry reached the engine")
}
engine.err = status.Error(codes.InvalidArgument, "retry refused")
postJSON(t, server.handleCreateIndex, request)
if _, err := server.indexRecord(t.Context(), "vec-bucket", "idx"); err != nil {
t.Fatalf("retry removed an uncertain dataset's catalog: %v", err)
}
engine.err = nil
request.MetadataConfiguration.NonFilterableMetadataKeys = []string{"a", "b"}
if result := postJSON(t, server.handleCreateIndex, request); result.Code != http.StatusOK {
t.Fatalf("resume: %d %s", result.Code, result.Body)
}
after, err := server.indexRecord(t.Context(), "vec-bucket", "idx")
if err != nil || recordNonce(before) != recordNonce(after) {
t.Fatalf("resume replaced the catalog: %v", err)
}
if result := postJSON(t, server.handleCreateIndex, request); result.Code != http.StatusConflict {
t.Fatalf("completed index was initialized again: %d %s", result.Code, result.Body)
}
result := postJSON(t, server.handleListTagsForResource, &ListTagsForResourceRequest{ResourceArn: server.indexArn("vec-bucket", "idx")})
var tags ListTagsForResourceResponse
if err := json.Unmarshal(result.Body.Bytes(), &tags); err != nil || tags.Tags["team"] != "search" {
t.Fatalf("resume lost tags: %s (%v)", result.Body, err)
}
}
func TestCreateIndexRollbackPreservesACompletedRetry(t *testing.T) {
fs := s3tablestest.Start(t)
engine := &failingCreateVectors{fakeVectors: &fakeVectors{}, err: status.Error(codes.InvalidArgument, "first attempt failed")}
server := NewServer(s3tables.NewManagerClient(fs.Client), allowAllAuth{}, engine)
postJSON(t, server.handleCreateVectorBucket, &CreateVectorBucketRequest{VectorBucketName: "vec-bucket"})
request := &CreateIndexRequest{VectorBucketName: "vec-bucket", IndexName: "idx", DataType: "float32", Dimension: 3, DistanceMetric: "cosine"}
engine.onCreate = func() {
engine.onCreate = nil
engine.err = nil
if result := postJSON(t, server.handleCreateIndex, request); result.Code != http.StatusOK {
t.Fatalf("concurrent retry: %d %s", result.Code, result.Body)
}
}
if result := postJSON(t, server.handleCreateIndex, request); result.Code == http.StatusOK {
t.Fatal("first attempt reported success")
}
if _, err := server.indexRecord(t.Context(), "vec-bucket", "idx"); err != nil {
t.Fatalf("rollback removed the completed retry: %v", err)
}
if result := postJSON(t, server.handleCreateIndex, request); result.Code != http.StatusConflict {
t.Fatalf("completed retry lost its state: %d %s", result.Code, result.Body)
}
}
func TestCreateIndexRetainsAnUnknownOutcomeAfterARefusedRetry(t *testing.T) {
for _, code := range []codes.Code{codes.PermissionDenied, codes.FailedPrecondition} {
t.Run(code.String(), func(t *testing.T) {
fs := s3tablestest.Start(t)
first := &failingCreateVectors{fakeVectors: &fakeVectors{}, err: status.Error(codes.Unavailable, "response lost after commit")}
second := &failingCreateVectors{fakeVectors: &fakeVectors{}, err: status.Error(code, "retry refused")}
discovery := &vectorDiscovery{client: first, nextResolve: time.Now().Add(time.Hour)}
discovery.discover = func([]pb.ServerAddress, grpc.DialOption) []string {
discovery.mu.Lock()
discovery.client = second
discovery.mu.Unlock()
return nil
}
defer discovery.Close()
server := NewServer(s3tables.NewManagerClient(fs.Client), allowAllAuth{}, discovery)
postJSON(t, server.handleCreateVectorBucket, &CreateVectorBucketRequest{VectorBucketName: "vec-bucket"})
result := postJSON(t, server.handleCreateIndex, &CreateIndexRequest{VectorBucketName: "vec-bucket", IndexName: "idx", DataType: "float32", Dimension: 3, DistanceMetric: "cosine"})
if result.Code != http.StatusServiceUnavailable || len(first.created) != 1 || len(second.created) != 1 {
t.Fatalf("unknown outcome lost: %d %s (attempts %d, %d)", result.Code, result.Body, len(first.created), len(second.created))
}
if _, err := server.indexRecord(t.Context(), "vec-bucket", "idx"); err != nil {
t.Fatalf("refused retry erased a possibly committed dataset: %v", err)
}
})
}
}
type concurrentCreateVectors struct {
vector_pb.VectorServiceClient
create func() (*vector_pb.CreateIndexResponse, error)
}
func (f *concurrentCreateVectors) CreateIndex(context.Context, *vector_pb.CreateIndexRequest, ...grpc.CallOption) (*vector_pb.CreateIndexResponse, error) {
return f.create()
}
func TestCreateIndexRollbackPreservesAnActiveRetry(t *testing.T) {
fs := s3tablestest.Start(t)
firstStarted, retryStarted, finishRetry := make(chan struct{}), make(chan struct{}), make(chan struct{})
ctx, cancel := context.WithTimeout(t.Context(), 5*time.Second)
defer cancel()
var calls atomic.Int32
engine := &concurrentCreateVectors{create: func() (*vector_pb.CreateIndexResponse, error) {
if calls.Add(1) == 1 {
close(firstStarted)
select {
case <-retryStarted:
return nil, status.Error(codes.InvalidArgument, "first attempt refused")
case <-ctx.Done():
return nil, ctx.Err()
}
}
close(retryStarted)
select {
case <-finishRetry:
return &vector_pb.CreateIndexResponse{}, nil
case <-ctx.Done():
return nil, ctx.Err()
}
}}
server := NewServer(s3tables.NewManagerClient(fs.Client), allowAllAuth{}, engine)
postJSON(t, server.handleCreateVectorBucket, &CreateVectorBucketRequest{VectorBucketName: "vec-bucket"})
request := &CreateIndexRequest{VectorBucketName: "vec-bucket", IndexName: "idx", DataType: "float32", Dimension: 3, DistanceMetric: "cosine"}
first, retry := make(chan *httptest.ResponseRecorder, 1), make(chan *httptest.ResponseRecorder, 1)
go func() { first <- postJSON(t, server.handleCreateIndex, request) }()
select {
case <-firstStarted:
case <-ctx.Done():
t.Fatal("first create did not reach the engine")
}
go func() { retry <- postJSON(t, server.handleCreateIndex, request) }()
if result := <-first; result.Code == http.StatusOK {
t.Fatal("first attempt reported success")
}
if _, err := server.indexRecord(t.Context(), "vec-bucket", "idx"); err != nil {
t.Fatalf("rollback removed an active retry: %v", err)
}
close(finishRetry)
if result := <-retry; result.Code != http.StatusOK {
t.Fatalf("retry could not finish: %d %s", result.Code, result.Body)
}
}
func TestCreateIndexResumesWithTheDefaultMetric(t *testing.T) {
fs := s3tablestest.Start(t)
engine := &failingCreateVectors{fakeVectors: &fakeVectors{}, err: status.Error(codes.Unavailable, "request lost")}
server := NewServer(s3tables.NewManagerClient(fs.Client), allowAllAuth{}, engine)
postJSON(t, server.handleCreateVectorBucket, &CreateVectorBucketRequest{VectorBucketName: "vec-bucket"})
request := &CreateIndexRequest{VectorBucketName: "vec-bucket", IndexName: "idx", DataType: "float32", Dimension: 3}
postJSON(t, server.handleCreateIndex, request)
engine.err = nil
if result := postJSON(t, server.handleCreateIndex, request); result.Code != http.StatusOK {
t.Fatalf("default-metric recovery: %d %s", result.Code, result.Body)
}
if len(engine.created) != 2 || engine.created[1].DistanceMetric != vector_pb.DistanceMetric_L2 {
t.Fatal("recovery changed the default metric")
}
}
func TestCreateIndexRetryPreservesDuplicateMetadataKeys(t *testing.T) {
fs := s3tablestest.Start(t)
engine := &failingCreateVectors{fakeVectors: &fakeVectors{}, err: status.Error(codes.Unavailable, "response lost")}
server := NewServer(s3tables.NewManagerClient(fs.Client), allowAllAuth{}, engine)
postJSON(t, server.handleCreateVectorBucket, &CreateVectorBucketRequest{VectorBucketName: "vec-bucket"})
request := &CreateIndexRequest{VectorBucketName: "vec-bucket", IndexName: "idx", DataType: "float32", Dimension: 3, DistanceMetric: "cosine",
MetadataConfiguration: &MetadataConfiguration{NonFilterableMetadataKeys: []string{"b", "a", "a"}}}
postJSON(t, server.handleCreateIndex, request)
engine.err = nil
if result := postJSON(t, server.handleCreateIndex, request); result.Code != http.StatusOK {
t.Fatalf("duplicate-key recovery: %d %s", result.Code, result.Body)
}
if len(engine.created) != 2 {
t.Fatalf("engine calls: %d", len(engine.created))
}
for _, call := range engine.created {
keys := slices.Clone(call.NonFilterableMetadataKeys)
slices.Sort(keys)
if !slices.Equal(keys, []string{"a", "a", "b"}) {
t.Fatalf("initialization changed metadata keys: %q", keys)
}
}
}
func TestCreateIndexDoesNotInitializeAConcurrentCatalogCreate(t *testing.T) {
fs := s3tablestest.Start(t)
engine := &fakeVectors{}
server := NewServer(s3tables.NewManagerClient(fs.Client), allowAllAuth{}, engine)
postJSON(t, server.handleCreateVectorBucket, &CreateVectorBucketRequest{VectorBucketName: "vec-bucket"})
postJSON(t, server.handleCreateIndex, &CreateIndexRequest{VectorBucketName: "vec-bucket", IndexName: "seed", DataType: "float32", Dimension: 7, DistanceMetric: "cosine"})
dir, name := splitEntryPath(s3tables.GetTablePath("vec-bucket", indexNamespace, "idx"))
seed := fs.Get(dir, "seed")
var lookups int
fs.OnLookup = func(d, n string) {
if d == dir && n == name {
lookups++
if lookups == 2 {
fs.Put(dir, name, seed.Extended)
}
}
}
result := postJSON(t, server.handleCreateIndex, &CreateIndexRequest{VectorBucketName: "vec-bucket", IndexName: "idx", DataType: "float32", Dimension: 3, DistanceMetric: "cosine"})
if result.Code != http.StatusConflict || len(engine.created) != 1 {
t.Fatalf("initialized a competing record: %d %s (%d engine calls)", result.Code, result.Body, len(engine.created))
}
record, err := server.indexRecord(t.Context(), "vec-bucket", "idx")
if err != nil || indexFromRecord(server, "vec-bucket", "idx", record).Dimension != 7 {
t.Fatalf("changed the competing record: %v", err)
}
}