Skip to content

[4/N] feat(submitqueue): activate request status materialization #343

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
albertywu merged 1 commit into from
Jul 15, 2026
Merged

[4/N] feat(submitqueue): activate request status materialization #343

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 10 additions & 8 deletions service/submitqueue/gateway/server/main.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -215,8 +215,9 @@ func run() error {
// Build the topic registry. The gateway publishes to the start of the
// orchestrator pipeline (TopicKeyStart) and the cancel topic (TopicKeyCancel) —
// both publish-only. It additionally consumes the log topic (TopicKeyLog):
// the gateway is the sole writer of the request log, persisting entries that
// the orchestrator publishes there.
// the gateway persists normal pipeline entries that the orchestrator
// publishes there. Orchestrator DLQ reconciliation may repair terminal
// request-log projections directly.
registry, err := consumer.NewTopicRegistry([]consumer.TopicConfig{
{Key: topickey.TopicKeyStart, Name: "start", Queue: mysqlQueue},
{Key: topickey.TopicKeyCancel, Name: "cancel", Queue: mysqlQueue},
Expand Down Expand Up @@ -249,8 +250,8 @@ func run() error {

// Initialize storage from the shared app database connection. The land
// controller writes to this store directly; cancel/status use the request
// log store directly. The log consumer (registered below) is the sole
// persister of request log entries published by the orchestrator.
// log store directly. The log consumer registered below persists request
// log entries published by the orchestrator's normal pipeline.
store, err := mysqlstorage.NewStorage(appDB, scope.SubScope("storage"))
if err != nil {
return fmt.Errorf("failed to create storage: %w", err)
Expand All @@ -271,7 +272,7 @@ func run() error {
// Create controllers and wrap them for gRPC
pingController := controller.NewPingController(logger, scope)
landController := controller.NewLandController(logger.Sugar(), scope, cnt, store, queueConfigs, registry)
cancelController := controller.NewCancelController(logger.Sugar(), scope, requestLogStore, registry)
cancelController := controller.NewCancelController(logger.Sugar(), scope, store, registry)
statusController := controller.NewStatusController(logger.Sugar(), scope, requestLogStore)
gatewayServer := &GatewayServer{
pingController: pingController,
Expand All @@ -285,9 +286,10 @@ func run() error {
// Register reflection service for debugging with grpcurl
reflection.Register(grpcServer)

// Create the queue consumer and register the log controller. The gateway is
// the sole persister of the request log: the orchestrator publishes entries
// to the log topic and this consumer writes them to storage.
// Create the queue consumer and register the log controller. The orchestrator
// publishes normal pipeline entries to the log topic and this consumer writes
// them to storage. Orchestrator DLQ reconciliation repairs terminal entries
// directly so reconciliation does not depend on another asynchronous path.
logConsumer := consumer.New(logger.Sugar(), scope.SubScope("consumer"), registry,
errs.NewClassifierProcessor(
// Storage (submitqueue/extension/storage/mysql) and queue (platform/extension/messagequeue/mysql)
Expand Down
31 changes: 15 additions & 16 deletions service/submitqueue/orchestrator/server/main.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -415,10 +415,9 @@ func newTopicRegistry(q extqueue.Queue, subscriberName string) (consumer.TopicRe
})
}

// Publish-only: the orchestrator emits request log entries to the log
// topic but never persists them. The gateway is the sole consumer that
// writes the request log to storage, so the orchestrator registers no
// consuming subscription (and therefore no log DLQ) for this topic.
// Publish-only: the orchestrator emits request-log entries to the log topic.
// The gateway is the sole consumer and writer of request logs and public
// projections, so the orchestrator registers no consuming subscription.
configs = append(configs, consumer.TopicConfig{
Key: topickey.TopicKeyLog,
Name: "log",
Expand Down Expand Up @@ -740,19 +739,19 @@ func registerDLQControllers(c consumer.Consumer, logger *zap.SugaredLogger, scop
name string
ctl consumer.Controller
}{
{"start_dlq", dlq.NewDLQRequestController(logger, dlqScope, store, dlq.DecodeLandRequestID, dlq.TopicKey(topickey.TopicKeyStart), "orchestrator-start-dlq")},
{"cancel_dlq", dlq.NewDLQRequestController(logger, dlqScope, store, dlq.DecodeCancelRequestID, dlq.TopicKey(topickey.TopicKeyCancel), "orchestrator-cancel-dlq")},
{"validate_dlq", dlq.NewDLQRequestController(logger, dlqScope, store, dlq.DecodeRequestID, dlq.TopicKey(topickey.TopicKeyValidate), "orchestrator-validate-dlq")},
{"mergeconflictsignal_dlq", dlq.NewDLQMergeConflictSignalController(logger, dlqScope, store, dlq.TopicKey(runwaymq.TopicKeyMergeConflictCheckSignal), "orchestrator-mergeconflictsignal-dlq")},
{"batch_dlq", dlq.NewDLQRequestController(logger, dlqScope, store, dlq.DecodeRequestID, dlq.TopicKey(topickey.TopicKeyBatch), "orchestrator-batch-dlq")},
{"score_dlq", dlq.NewDLQBatchController(logger, dlqScope, store, dlq.TopicKey(topickey.TopicKeyScore), "orchestrator-score-dlq")},
{"speculate_dlq", dlq.NewDLQBatchController(logger, dlqScope, store, dlq.TopicKey(topickey.TopicKeySpeculate), "orchestrator-speculate-dlq")},
{"start_dlq", dlq.NewDLQRequestController(logger, dlqScope, store, registry, dlq.DecodeLandRequestID, dlq.TopicKey(topickey.TopicKeyStart), "orchestrator-start-dlq")},
{"cancel_dlq", dlq.NewDLQRequestController(logger, dlqScope, store, registry, dlq.DecodeCancelRequestID, dlq.TopicKey(topickey.TopicKeyCancel), "orchestrator-cancel-dlq")},
{"validate_dlq", dlq.NewDLQRequestController(logger, dlqScope, store, registry, dlq.DecodeRequestID, dlq.TopicKey(topickey.TopicKeyValidate), "orchestrator-validate-dlq")},
{"mergeconflictsignal_dlq", dlq.NewDLQMergeConflictSignalController(logger, dlqScope, store, registry, dlq.TopicKey(runwaymq.TopicKeyMergeConflictCheckSignal), "orchestrator-mergeconflictsignal-dlq")},
{"batch_dlq", dlq.NewDLQRequestController(logger, dlqScope, store, registry, dlq.DecodeRequestID, dlq.TopicKey(topickey.TopicKeyBatch), "orchestrator-batch-dlq")},
{"score_dlq", dlq.NewDLQBatchController(logger, dlqScope, store, registry, dlq.TopicKey(topickey.TopicKeyScore), "orchestrator-score-dlq")},
{"speculate_dlq", dlq.NewDLQBatchController(logger, dlqScope, store, registry, dlq.TopicKey(topickey.TopicKeySpeculate), "orchestrator-speculate-dlq")},
{"prioritize_dlq", dlq.NewDLQQueueController(logger, dlqScope, registry, dlq.TopicKey(topickey.TopicKeyPrioritize), "orchestrator-prioritize-dlq")},
{"build_dlq", dlq.NewDLQBatchController(logger, dlqScope, store, dlq.TopicKey(topickey.TopicKeyBuild), "orchestrator-build-dlq")},
{"buildsignal_dlq", dlq.NewDLQBuildSignalController(logger, dlqScope, store, dlq.TopicKey(topickey.TopicKeyBuildSignal), "orchestrator-buildsignal-dlq")},
{"merge_dlq", dlq.NewDLQBatchController(logger, dlqScope, store, dlq.TopicKey(topickey.TopicKeyMerge), "orchestrator-merge-dlq")},
{"mergesignal_dlq", dlq.NewDLQMergeSignalController(logger, dlqScope, store, dlq.TopicKey(runwaymq.TopicKeyMergeSignal), "orchestrator-mergesignal-dlq")},
{"conclude_dlq", dlq.NewDLQBatchController(logger, dlqScope, store, dlq.TopicKey(topickey.TopicKeyConclude), "orchestrator-conclude-dlq")},
{"build_dlq", dlq.NewDLQBatchController(logger, dlqScope, store, registry, dlq.TopicKey(topickey.TopicKeyBuild), "orchestrator-build-dlq")},
{"buildsignal_dlq", dlq.NewDLQBuildSignalController(logger, dlqScope, store, registry, dlq.TopicKey(topickey.TopicKeyBuildSignal), "orchestrator-buildsignal-dlq")},
{"merge_dlq", dlq.NewDLQBatchController(logger, dlqScope, store, registry, dlq.TopicKey(topickey.TopicKeyMerge), "orchestrator-merge-dlq")},
{"mergesignal_dlq", dlq.NewDLQMergeSignalController(logger, dlqScope, store, registry, dlq.TopicKey(runwaymq.TopicKeyMergeSignal), "orchestrator-mergesignal-dlq")},
{"conclude_dlq", dlq.NewDLQBatchController(logger, dlqScope, store, registry, dlq.TopicKey(topickey.TopicKeyConclude), "orchestrator-conclude-dlq")},
}
var count int
for _, reg := range dlqRegs {
Expand Down
36 changes: 17 additions & 19 deletions submitqueue/gateway/controller/cancel.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,7 @@ import (
entityqueue "github.com/uber/submitqueue/platform/base/messagequeue"
"github.com/uber/submitqueue/platform/consumer"
"github.com/uber/submitqueue/platform/errs"
requestcore "github.com/uber/submitqueue/submitqueue/core/request"
"github.com/uber/submitqueue/submitqueue/core/topickey"
"github.com/uber/submitqueue/submitqueue/entity"
"github.com/uber/submitqueue/submitqueue/extension/storage"
Expand All @@ -35,21 +36,23 @@ import (
// and returns a response. The orchestrator-side cancel controller performs the actual
// state transitions and emits the terminal RequestStatusCancelled log entry.
type CancelController struct {
logger *zap.SugaredLogger
metricsScope tally.Scope
requestLogStore storage.RequestLogStore
registry consumer.TopicRegistry
logger *zap.SugaredLogger
metricsScope tally.Scope
requestSummaryStore storage.RequestSummaryStore
materializer *requestcore.Materializer
registry consumer.TopicRegistry
}

// NewCancelController creates a new instance of the gateway cancel controller.
// The controller writes a RequestStatusCancelling log entry through requestLogStore and
// The controller writes a RequestStatusCancelling log entry through the shared materializer and
// publishes cancel requests to the topic registered under topickey.TopicKeyCancel.
func NewCancelController(logger *zap.SugaredLogger, scope tally.Scope, requestLogStore storage.RequestLogStore, registry consumer.TopicRegistry) *CancelController {
func NewCancelController(logger *zap.SugaredLogger, scope tally.Scope, store storage.Storage, registry consumer.TopicRegistry) *CancelController {
return &CancelController{
logger: logger,
metricsScope: scope,
requestLogStore: requestLogStore,
registry: registry,
logger: logger,
metricsScope: scope,
requestSummaryStore: store.GetRequestSummaryStore(),
materializer: requestcore.NewMaterializer(store),
registry: registry,
}
}

Expand Down Expand Up @@ -78,18 +81,13 @@ func (c *CancelController) Cancel(ctx context.Context, req entity.CancelRequest)
"reason", req.Reason,
)

// Verify the sqid exists before recording intent or publishing. Cancel is opt-in
// by sqid; an unknown sqid is a user error and must never leave a cancelling log
// row or a queue message behind for a request that never existed. The Land
// controller writes its "accepted" log entry synchronously to the same store, so
// a NotFound here reliably means "this sqid was never accepted by the gateway"
// rather than "in flight" — there is no false-negative race window.
if _, err := c.requestLogStore.List(ctx, req.ID); err != nil {
// Verify the sqid exists before recording intent or publishing.
if _, err := c.requestSummaryStore.Get(ctx, req.ID); err != nil {
if storage.IsNotFound(err) {
c.metricsScope.Counter("cancel_request_not_found").Inc(1)
return errs.NewUserError(&RequestNotFoundError{Sqid: req.ID})
}
return fmt.Errorf("CancelController failed to look up request log for sqid=%s: %w", req.ID, err)
return fmt.Errorf("CancelController failed to look up request summary for sqid=%s: %w", req.ID, err)
}

// Record the user's intent in the request log before publishing. Writing direct to the
Expand All @@ -100,7 +98,7 @@ func (c *CancelController) Cancel(ctx context.Context, req entity.CancelRequest)
metadata["reason"] = req.Reason
}
logEntry := entity.NewRequestLog(req.ID, entity.RequestStatusCancelling, 0, "", metadata)
if err := c.requestLogStore.Insert(ctx, logEntry); err != nil {
if err := c.materializer.PersistLog(ctx, logEntry); err != nil {
return fmt.Errorf("CancelController failed to insert cancelling log for sqid=%s: %w", req.ID, err)
}

Expand Down
Loading
Loading