X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/c36ec856598f214e340e3335ddd347d131335bf8..ac312e0acae4fd114114081c9f4791d05e640831:/services/crunch-dispatch-slurm/crunch-dispatch-slurm.go?ds=inline

diff --git a/services/crunch-dispatch-slurm/crunch-dispatch-slurm.go b/services/crunch-dispatch-slurm/crunch-dispatch-slurm.go
index c31d799752..5a9ef91c3d 100644
--- a/services/crunch-dispatch-slurm/crunch-dispatch-slurm.go
+++ b/services/crunch-dispatch-slurm/crunch-dispatch-slurm.go
@@ -7,6 +7,8 @@ package dispatchslurm
 
 import (
 	"context"
+	"crypto/hmac"
+	"crypto/sha256"
 	"fmt"
 	"log"
 	"math"
@@ -17,6 +19,8 @@ import (
 	"time"
 
 	"git.arvados.org/arvados.git/lib/cmd"
+	"git.arvados.org/arvados.git/lib/controller/dblock"
+	"git.arvados.org/arvados.git/lib/ctrlctx"
 	"git.arvados.org/arvados.git/lib/dispatchcloud"
 	"git.arvados.org/arvados.git/lib/service"
 	"git.arvados.org/arvados.git/sdk/go/arvados"
@@ -53,10 +57,11 @@ const initialNiceValue int64 = 10000
 
 type Dispatcher struct {
 	*dispatch.Dispatcher
-	logger  logrus.FieldLogger
-	cluster *arvados.Cluster
-	sqCheck *SqueueChecker
-	slurm   Slurm
+	logger      logrus.FieldLogger
+	cluster     *arvados.Cluster
+	sqCheck     *SqueueChecker
+	slurm       Slurm
+	dbConnector ctrlctx.DBConnector
 
 	done chan struct{}
 	err  error
@@ -88,6 +93,7 @@ func (disp *Dispatcher) configure() error {
 	disp.Client.APIHost = disp.cluster.Services.Controller.ExternalURL.Host
 	disp.Client.AuthToken = disp.cluster.SystemRootToken
 	disp.Client.Insecure = disp.cluster.TLS.Insecure
+	disp.dbConnector = ctrlctx.DBConnector{PostgreSQL: disp.cluster.PostgreSQL}
 
 	if disp.Client.APIHost != "" || disp.Client.AuthToken != "" {
 		// Copy real configs into env vars so [a]
@@ -99,7 +105,6 @@ func (disp *Dispatcher) configure() error {
 		if disp.Client.Insecure {
 			os.Setenv("ARVADOS_API_HOST_INSECURE", "1")
 		}
-		os.Setenv("ARVADOS_EXTERNAL_CLIENT", "")
 		for k, v := range disp.cluster.Containers.SLURM.SbatchEnvironmentVariables {
 			os.Setenv(k, v)
 		}
@@ -136,6 +141,8 @@ func (disp *Dispatcher) setup() {
 }
 
 func (disp *Dispatcher) run() error {
+	dblock.Dispatch.Lock(context.Background(), disp.dbConnector.GetDB)
+	defer dblock.Dispatch.Unlock()
 	defer disp.sqCheck.Stop()
 
 	if disp.cluster != nil && len(disp.cluster.InstanceTypes) > 0 {
@@ -190,14 +197,16 @@ func (disp *Dispatcher) sbatchArgs(container arvados.Container) ([]string, error
 	if disp.cluster == nil {
 		// no instance types configured
 		args = append(args, disp.slurmConstraintArgs(container)...)
-	} else if it, err := dispatchcloud.ChooseInstanceType(disp.cluster, &container); err == dispatchcloud.ErrInstanceTypesNotConfigured {
+	} else if types, err := dispatchcloud.ChooseInstanceType(disp.cluster, &container); err == dispatchcloud.ErrInstanceTypesNotConfigured {
 		// ditto
 		args = append(args, disp.slurmConstraintArgs(container)...)
 	} else if err != nil {
 		return nil, err
 	} else {
-		// use instancetype constraint instead of slurm mem/cpu/tmp specs
-		args = append(args, "--constraint=instancetype="+it.Name)
+		// use instancetype constraint instead of slurm
+		// mem/cpu/tmp specs (note types[0] is the lowest-cost
+		// suitable instance type)
+		args = append(args, "--constraint=instancetype="+types[0].Name)
 	}
 
 	if len(container.SchedulingParameters.Partitions) > 0 {
@@ -213,7 +222,12 @@ func (disp *Dispatcher) submit(container arvados.Container, crunchRunCommand []s
 	crArgs := append([]string(nil), crunchRunCommand...)
 	crArgs = append(crArgs, "--runtime-engine="+disp.cluster.Containers.RuntimeEngine)
 	crArgs = append(crArgs, container.UUID)
-	crScript := strings.NewReader(execScript(crArgs))
+
+	h := hmac.New(sha256.New, []byte(disp.cluster.SystemRootToken))
+	fmt.Fprint(h, container.UUID)
+	authsecret := fmt.Sprintf("%x", h.Sum(nil))
+
+	crScript := strings.NewReader(execScript(crArgs, map[string]string{"GatewayAuthSecret": authsecret}))
 
 	sbArgs, err := disp.sbatchArgs(container)
 	if err != nil {