fn: lb & pure-runner slot hash id communication (#1007)

* fn: lb & pure-runner slot hash id communication

With this change, LB can pre-calculate the slot hash
key and pass it to runners. If LB knows/calculates
the slot hash ids, then it can also make better
estimates on which runner can successfully execute
it especially when status messages from runner
include a small summary of idle slots for a given
slot hash id. (TODO)

* fn: fix mock test

api/agent/call.go

@@ -315,12 +315,17 @@ type call struct {
 	execDeadline   time.Time
 	requestState   RequestState
 	containerState ContainerState
+	slotHashId     string
 }
 
 func (c *call) LbDeadline() time.Time {
 	return c.lbDeadline
 }
 
+func (c *call) SlotHashId() string {
+	return c.slotHashId
+}
+
 func (c *call) RequestBody() io.ReadCloser {
 	if c.req.Body != nil && c.req.GetBody != nil {
 		rdr, err := c.req.GetBody()

api/agent/grpc/runner.pb.go

@@ -44,6 +44,7 @@ const _ = proto.ProtoPackageIsVersion2 // please upgrade the proto package
 // Request to allocate a slot for a call
 type TryCall struct {
 	ModelsCallJson string `protobuf:"bytes,1,opt,name=models_call_json,json=modelsCallJson" json:"models_call_json,omitempty"`
+	SlotHashId     string `protobuf:"bytes,2,opt,name=slot_hash_id,json=slotHashId" json:"slot_hash_id,omitempty"`
 }
 
 func (m *TryCall) Reset()                    { *m = TryCall{} }
@@ -58,6 +59,13 @@ func (m *TryCall) GetModelsCallJson() string {
 	return ""
 }
 
+func (m *TryCall) GetSlotHashId() string {
+	if m != nil {
+		return m.SlotHashId
+	}
+	return ""
+}
+
 // Data sent C2S and S2C - as soon as the runner sees the first of these it
 // will start running. If empty content, there must be one of these with eof.
 // The runner will send these for the body of the response, AFTER it has sent
@@ -705,37 +713,39 @@ var _RunnerProtocol_serviceDesc = grpc.ServiceDesc{
 func init() { proto.RegisterFile("runner.proto", fileDescriptor0) }
 
 var fileDescriptor0 = []byte{
-	// 511 bytes of a gzipped FileDescriptorProto
+	// 529 bytes of a gzipped FileDescriptorProto
-	0x1f, 0x8b, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0xff, 0x84, 0x53, 0x5d, 0x8b, 0xd3, 0x40,
+	0x1f, 0x8b, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x02, 0xff, 0x84, 0x53, 0x5d, 0x8b, 0x13, 0x31,
-	0x14, 0x6d, 0xb6, 0x9f, 0xb9, 0xc9, 0xae, 0x65, 0x10, 0x09, 0x75, 0xc1, 0x12, 0x3f, 0x08, 0x08,
+	0x14, 0xed, 0xf4, 0xbb, 0x77, 0x66, 0xd7, 0x12, 0x44, 0x86, 0xba, 0x60, 0x19, 0x3f, 0x28, 0x08,
-	0x53, 0xed, 0xea, 0xab, 0x0f, 0xd6, 0x5d, 0x82, 0xb0, 0x20, 0x53, 0xf1, 0xb5, 0x4c, 0x93, 0xdb,
+	0xa9, 0x56, 0x7d, 0xf5, 0xc1, 0xba, 0xcb, 0x28, 0x2c, 0x48, 0xaa, 0xbe, 0x0e, 0xe9, 0x24, 0x9d,
-	0xb4, 0x3a, 0xcd, 0x94, 0x99, 0xc9, 0x42, 0xc1, 0x3f, 0xe2, 0xbf, 0x95, 0x99, 0xa4, 0xd9, 0xe2,
+	0x19, 0x4d, 0x27, 0x25, 0xc9, 0x2c, 0x14, 0xfc, 0x23, 0xfe, 0x5b, 0x49, 0x66, 0x3a, 0x5b, 0x7c,
-	0x8b, 0x6f, 0x73, 0xe6, 0xdc, 0xaf, 0x73, 0xee, 0x0c, 0x84, 0xaa, 0x2a, 0x4b, 0x54, 0xf4, 0xa0,
+	0xd9, 0xb7, 0x9c, 0x7b, 0x72, 0x3f, 0xce, 0xb9, 0x09, 0x04, 0xaa, 0x2a, 0x4b, 0xae, 0xf0, 0x41,
-	0xa4, 0x91, 0x93, 0xe7, 0x85, 0x94, 0x85, 0xc0, 0x99, 0x43, 0xeb, 0x6a, 0x33, 0xc3, 0xfd, 0xc1,
+	0x49, 0x23, 0x67, 0x4f, 0x33, 0x29, 0x33, 0xc1, 0x97, 0x0e, 0x6d, 0xab, 0xdd, 0x92, 0xef, 0x0f,
-	0x1c, 0x6b, 0x32, 0xbe, 0x81, 0xe1, 0x77, 0x75, 0x5c, 0x70, 0x21, 0x48, 0x02, 0xe3, 0xbd, 0xcc,
+	0xe6, 0x58, 0x93, 0xd1, 0x0f, 0x18, 0x7d, 0x57, 0xc7, 0x35, 0x15, 0x02, 0x2d, 0x60, 0xba, 0x97,
-	0x51, 0xe8, 0x55, 0xc6, 0x85, 0x58, 0xfd, 0xd4, 0xb2, 0x8c, 0xbc, 0xa9, 0x97, 0xf8, 0xec, 0xaa,
+	0x8c, 0x0b, 0x9d, 0xa4, 0x54, 0x88, 0xe4, 0x97, 0x96, 0x65, 0xe8, 0xcd, 0xbd, 0xc5, 0x84, 0x5c,
-	0xbe, 0xb7, 0x51, 0x5f, 0xb5, 0x2c, 0xe3, 0xf7, 0xe0, 0x7f, 0xe1, 0x86, 0xdf, 0x29, 0xbe, 0x47,
+	0xd6, 0x71, 0x7b, 0xeb, 0xab, 0x96, 0x25, 0x9a, 0x43, 0xa0, 0x85, 0x34, 0x49, 0x4e, 0x75, 0x9e,
-	0x42, 0xa0, 0x97, 0x73, 0xc3, 0x5d, 0x68, 0xc8, 0xdc, 0x99, 0x8c, 0xa1, 0x8b, 0x72, 0x13, 0x5d,
+	0x14, 0x2c, 0xec, 0xba, 0x5b, 0x60, 0x63, 0x31, 0xd5, 0xf9, 0x17, 0x16, 0xbd, 0x85, 0xc9, 0x67,
-	0x4c, 0xbd, 0x64, 0xc4, 0xec, 0x31, 0xfe, 0x00, 0x90, 0x1a, 0x73, 0x48, 0x91, 0xe7, 0xa8, 0x2c,
+	0x6a, 0xe8, 0x8d, 0xa2, 0x7b, 0x8e, 0x10, 0xf4, 0x19, 0x35, 0xd4, 0x15, 0x0b, 0x88, 0x3b, 0xa3,
-	0xff, 0x0b, 0x8f, 0x4d, 0x75, 0x7b, 0x24, 0x4f, 0xa1, 0xff, 0xc0, 0x45, 0x85, 0x2e, 0xc7, 0x67,
+	0x29, 0xf4, 0xb8, 0xdc, 0xb9, 0xcc, 0x31, 0xb1, 0xc7, 0xe8, 0x3d, 0x40, 0x6c, 0xcc, 0x21, 0xe6,
-	0x35, 0x88, 0x7f, 0x40, 0x68, 0xb3, 0x18, 0xea, 0xc3, 0x3d, 0x1a, 0x4e, 0x5e, 0x40, 0xa0, 0x0d,
+	0x94, 0x71, 0x65, 0xf9, 0xdf, 0xfc, 0xd8, 0xf4, 0xb7, 0x47, 0xf4, 0x18, 0x06, 0x77, 0x54, 0x54,
-	0x37, 0x95, 0x5e, 0x65, 0x32, 0x47, 0x97, 0xdf, 0x67, 0x50, 0x5f, 0x2d, 0x64, 0x8e, 0xe4, 0x35,
+	0xbc, 0xe9, 0x56, 0x83, 0xe8, 0x27, 0x04, 0x36, 0x8b, 0x70, 0x7d, 0xb8, 0xe5, 0x86, 0xa2, 0x67,
-	0x0c, 0xb7, 0xae, 0x85, 0x8e, 0x2e, 0xa6, 0xdd, 0x24, 0x98, 0x07, 0xf4, 0xb1, 0x2d, 0x3b, 0x71,
+	0xe0, 0x6b, 0x43, 0x4d, 0xa5, 0x93, 0x54, 0x32, 0xee, 0xf2, 0x07, 0x04, 0xea, 0xd0, 0x5a, 0x32,
-	0xf1, 0x27, 0x78, 0x62, 0xc5, 0x30, 0xd4, 0x95, 0x30, 0x4b, 0xc3, 0x95, 0x21, 0x2f, 0xa1, 0xb7,
+	0x8e, 0x5e, 0xc2, 0x28, 0x77, 0x2d, 0x74, 0xd8, 0x9d, 0xf7, 0x16, 0xfe, 0xca, 0xc7, 0xf7, 0x6d,
-	0x35, 0xe6, 0x10, 0xe5, 0x53, 0x2f, 0x09, 0xe6, 0x97, 0xf4, 0xbc, 0x6f, 0xda, 0x61, 0x8e, 0xfc,
+	0xc9, 0x89, 0x8b, 0x3e, 0xc2, 0x23, 0x2b, 0x97, 0x70, 0x5d, 0x09, 0xb3, 0x31, 0x54, 0x19, 0xf4,
-	0x3c, 0x80, 0xde, 0x1e, 0x0d, 0x8f, 0x7f, 0x43, 0x68, 0xf3, 0xef, 0x76, 0xe5, 0x4e, 0x6f, 0x31,
+	0x1c, 0xfa, 0xb9, 0x31, 0x87, 0x90, 0xcd, 0xbd, 0x85, 0xbf, 0xba, 0xc0, 0xe7, 0x7d, 0xe3, 0x0e,
-	0x27, 0x11, 0x0c, 0x75, 0x95, 0x65, 0xa8, 0xb5, 0x9b, 0x69, 0xc4, 0x4e, 0xd0, 0x32, 0x39, 0x1a,
+	0x71, 0xe4, 0xa7, 0x21, 0xf4, 0xf7, 0xdc, 0xd0, 0xe8, 0x0f, 0x04, 0x36, 0xff, 0xa6, 0x28, 0x0b,
-	0xbe, 0x13, 0xba, 0x51, 0x76, 0x82, 0xe4, 0x1a, 0x7c, 0x54, 0x4a, 0x2a, 0x3b, 0x77, 0xd4, 0x75,
+	0x9d, 0x73, 0x86, 0x42, 0x18, 0xe9, 0x2a, 0x4d, 0xb9, 0xd6, 0x6e, 0xa6, 0x31, 0x39, 0x41, 0xcb,
-	0x4a, 0x1e, 0x2f, 0xc8, 0x04, 0x46, 0x0e, 0x2c, 0x8d, 0x8a, 0x7a, 0x2e, 0xb1, 0xc5, 0xf1, 0x12,
+	0x30, 0x6e, 0x68, 0x21, 0x74, 0xa3, 0xec, 0x04, 0xd1, 0x15, 0x4c, 0xb8, 0x52, 0x52, 0xd9, 0xb9,
-	0xfc, 0x85, 0xd8, 0x61, 0x69, 0xee, 0x75, 0x41, 0xae, 0xa1, 0x6b, 0x54, 0x6d, 0x65, 0x30, 0x1f,
+	0xc3, 0x9e, 0x53, 0x72, 0x1f, 0x40, 0x33, 0x18, 0x3b, 0xb0, 0x31, 0x2a, 0xec, 0xbb, 0xc4, 0x16,
-	0xd1, 0x66, 0x99, 0x69, 0x87, 0xd9, 0x6b, 0x32, 0x6d, 0x96, 0x73, 0xe1, 0x68, 0xa0, 0xed, 0xda,
+	0x47, 0x1b, 0x98, 0xac, 0x45, 0xc1, 0x4b, 0x73, 0xab, 0x33, 0x74, 0x05, 0x3d, 0xa3, 0x6a, 0x2b,
-	0xac, 0x24, 0xcb, 0x58, 0x49, 0x6b, 0x99, 0x1f, 0xe3, 0x3f, 0x1e, 0xf8, 0xcc, 0x3d, 0x1b, 0x5b,
+	0xfd, 0xd5, 0x18, 0x37, 0xeb, 0x8e, 0x3b, 0xc4, 0x86, 0xd1, 0xbc, 0x59, 0x4e, 0xd7, 0xd1, 0x80,
-	0xf5, 0x23, 0x84, 0xca, 0x99, 0xb3, 0xd2, 0xd6, 0x9d, 0xa6, 0xfc, 0x98, 0xfe, 0xe3, 0x5a, 0xda,
+	0xdb, 0xb5, 0x59, 0x49, 0x96, 0xb1, 0x92, 0xb6, 0x92, 0x1d, 0xa3, 0xbf, 0x1e, 0x4c, 0x88, 0x7b,
-	0x61, 0x81, 0x3a, 0x33, 0xf1, 0xbf, 0xed, 0xc8, 0x5b, 0x18, 0x6d, 0x1a, 0xd7, 0x9c, 0x68, 0x6b,
+	0x58, 0xb6, 0xea, 0x07, 0x08, 0x94, 0x33, 0x27, 0xd1, 0xd6, 0x9d, 0xa6, 0xfc, 0x14, 0xff, 0xe7,
-	0xf5, 0xb9, 0x95, 0x69, 0x87, 0xb5, 0x01, 0xed, 0x6c, 0x6f, 0x20, 0xac, 0x47, 0x5b, 0xba, 0x4d,
+	0x5a, 0xdc, 0x21, 0xbe, 0x3a, 0x33, 0xf1, 0xc1, 0x76, 0xe8, 0x35, 0x8c, 0x77, 0x8d, 0x6b, 0x4e,
-	0x93, 0x67, 0x30, 0xe0, 0x99, 0xd9, 0x3d, 0xd4, 0xaf, 0xa5, 0xcf, 0x1a, 0x34, 0x2f, 0xe0, 0xaa,
+	0xb4, 0xb5, 0xfa, 0xdc, 0xca, 0xb8, 0x43, 0xda, 0x0b, 0xed, 0x6c, 0xaf, 0x20, 0xa8, 0x47, 0xdb,
-	0x8e, 0xfb, 0x66, 0xdf, 0x76, 0x26, 0x05, 0x79, 0x05, 0x83, 0xdb, 0xb2, 0xe0, 0x05, 0x12, 0xa0,
+	0xb8, 0x4d, 0xa3, 0x27, 0x30, 0xa4, 0xa9, 0x29, 0xee, 0xea, 0xd7, 0x32, 0x20, 0x0d, 0x5a, 0x65,
-	0xad, 0x67, 0x13, 0xa0, 0xad, 0xd2, 0xc4, 0x7b, 0xe7, 0x91, 0x19, 0x0c, 0x4e, 0x95, 0x69, 0xfd,
+	0x70, 0x59, 0xdf, 0xfb, 0x66, 0x5f, 0x7f, 0x2a, 0x05, 0x7a, 0x01, 0xc3, 0xeb, 0x32, 0xa3, 0x19,
-	0x59, 0xe8, 0xe9, 0xb3, 0xd0, 0x5b, 0xfb, 0x59, 0x26, 0x97, 0xf4, 0x7c, 0x80, 0xf5, 0xc0, 0xd1,
+	0x47, 0x80, 0x5b, 0xcf, 0x66, 0x80, 0x5b, 0xa5, 0x0b, 0xef, 0x8d, 0x87, 0x96, 0x30, 0x3c, 0x55,
-	0x37, 0x7f, 0x03, 0x00, 0x00, 0xff, 0xff, 0xdc, 0x15, 0x71, 0xf0, 0x69, 0x03, 0x00, 0x00,
+	0xc6, 0xf5, 0x77, 0xc2, 0xa7, 0xef, 0x84, 0xaf, 0xed, 0x77, 0x9a, 0x5d, 0xe0, 0xf3, 0x01, 0xb6,
+	0x43, 0x47, 0xbf, 0xfb, 0x17, 0x00, 0x00, 0xff, 0xff, 0x0a, 0xb4, 0xdf, 0x3c, 0x8b, 0x03, 0x00,
+	0x00,
 }

api/agent/grpc/runner.proto

@@ -5,6 +5,7 @@ import "google/protobuf/empty.proto";
 // Request to allocate a slot for a call
 message TryCall {
     string models_call_json = 1;
+    string slot_hash_id = 2;
 }
 
 // Data sent C2S and S2C - as soon as the runner sees the first of these it

api/agent/lb_agent.go

@@ -128,6 +128,7 @@ func (a *lbAgent) GetCall(opts ...CallOpt) (Call, error) {
 	c.req = c.req.WithContext(ctx)
 
 	c.lbDeadline = time.Now().Add(time.Duration(c.Call.Timeout) * time.Second)
+	c.slotHashId = getSlotQueueKey(&c)
 
 	return &c, nil
 }

api/agent/lb_agent_test.go

@@ -122,12 +122,17 @@ type mockRunnerCall struct {
 	rw         http.ResponseWriter
 	stdErr     io.ReadWriteCloser
 	model      *models.Call
+	slotHashId string
 }
 
 func (c *mockRunnerCall) LbDeadline() time.Time {
 	return c.lbDeadline
 }
 
+func (c *mockRunnerCall) SlotHashId() string {
+	return c.slotHashId
+}
+
 func (c *mockRunnerCall) RequestBody() io.ReadCloser {
 	return c.r.Body
 }

api/agent/pure_runner.go

@@ -536,6 +536,9 @@ func (pr *pureRunner) handleTryCall(tc *runner.TryCall, state *callHandle) error
 	}
 
 	state.c = agent_call.(*call)
+	if tc.SlotHashId != "" {
+		state.c.slotHashId = tc.SlotHashId
+	}
 	state.allocatedTime = strfmt.DateTime(time.Now())
 	pr.spawnSubmit(state)
 

api/agent/runner_client.go

@@ -124,7 +124,10 @@ func (r *gRPCRunner) TryExec(ctx context.Context, call pool.RunnerCall) (bool, e
 
 	// After this point, we assume "COMMITTED" unless pure runner
 	// send explicit NACK
-	err = runnerConnection.Send(&pb.ClientMsg{Body: &pb.ClientMsg_Try{Try: &pb.TryCall{ModelsCallJson: string(modelJSON)}}})
+	err = runnerConnection.Send(&pb.ClientMsg{Body: &pb.ClientMsg_Try{Try: &pb.TryCall{
+		ModelsCallJson: string(modelJSON),
+		SlotHashId:     call.SlotHashId(),
+	}}})
 	if err != nil {
 		logrus.WithError(err).Error("Failed to send message to runner node")
 		return true, err

api/agent/slots.go

@@ -245,7 +245,12 @@ func (a *slotQueue) exitContainerState(conType ContainerStateType) {
 // a container will be locked up forever waiting for slot to free.
 func (a *slotQueueMgr) getSlotQueue(call *call) (*slotQueue, bool) {
 
-	key := getSlotQueueKey(call)
+	var key string
+	if call.slotHashId != "" {
+		key = call.slotHashId
+	} else {
+		key = getSlotQueueKey(call)
+	}
 
 	a.hMu.Lock()
 	slots, ok := a.hot[key]

api/runnerpool/runner_pool.go

@@ -41,6 +41,7 @@ type Runner interface {
 // RunnerCall provides access to the necessary details of request in order for it to be
 // processed by a RunnerPool
 type RunnerCall interface {
+	SlotHashId() string
 	LbDeadline() time.Time
 	RequestBody() io.ReadCloser
 	ResponseWriter() http.ResponseWriter