X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/68b9c7d30c7f47f8a9f9cff8a327fa9a3812d4da..844ff7cc1dc1c93a29b7ad8eca2987b987cf89e6:/lib/crunchrun/singularity.go

diff --git a/lib/crunchrun/singularity.go b/lib/crunchrun/singularity.go
index 741f542454..1af0d420e4 100644
--- a/lib/crunchrun/singularity.go
+++ b/lib/crunchrun/singularity.go
@@ -10,6 +10,7 @@ import (
 	"os"
 	"os/exec"
 	"sort"
+	"strings"
 	"syscall"
 	"time"
 
@@ -36,6 +37,14 @@ func newSingularityExecutor(logf func(string, ...interface{})) (*singularityExec
 	}, nil
 }
 
+func (e *singularityExecutor) Runtime() string {
+	buf, err := exec.Command("singularity", "--version").CombinedOutput()
+	if err != nil {
+		return "singularity (unknown version)"
+	}
+	return strings.TrimSuffix(string(buf), "\n")
+}
+
 func (e *singularityExecutor) getOrCreateProject(ownerUuid string, name string, containerClient *arvados.Client) (*arvados.Group, error) {
 	var gp arvados.GroupList
 	err := containerClient.RequestAndDecode(&gp,
@@ -101,7 +110,7 @@ func (e *singularityExecutor) checkImageCache(dockerImageID string, container ar
 	if len(cl.Items) == 1 {
 		imageCollection = cl.Items[0]
 	} else {
-		collectionName := collectionName + " " + time.Now().UTC().Format(time.RFC3339)
+		collectionName := "converting " + collectionName
 		exp := time.Now().Add(24 * 7 * 2 * time.Hour)
 		err = containerClient.RequestAndDecode(&imageCollection,
 			arvados.EndpointCollectionCreate.Method,
@@ -112,6 +121,7 @@ func (e *singularityExecutor) checkImageCache(dockerImageID string, container ar
 					"name":       collectionName,
 					"trash_at":   exp.UTC().Format(time.RFC3339),
 				},
+				"ensure_unique_name": true,
 			})
 		if err != nil {
 			return nil, fmt.Errorf("error creating '%v' collection: %s", collectionName, err)
@@ -141,6 +151,12 @@ func (e *singularityExecutor) LoadImage(dockerImageID string, imageTarballPath s
 	}
 
 	if _, err := os.Stat(imageFilename); os.IsNotExist(err) {
+		// Make sure the docker image is readable, and error
+		// out if not.
+		if _, err := os.Stat(imageTarballPath); err != nil {
+			return err
+		}
+
 		e.logf("building singularity image")
 		// "singularity build" does not accept a
 		// docker-archive://... filename containing a ":" character,
@@ -151,7 +167,22 @@ func (e *singularityExecutor) LoadImage(dockerImageID string, imageTarballPath s
 			return err
 		}
 
+		// Set up a cache and tmp dir for singularity build
+		err = os.Mkdir(e.tmpdir+"/cache", 0700)
+		if err != nil {
+			return err
+		}
+		defer os.RemoveAll(e.tmpdir + "/cache")
+		err = os.Mkdir(e.tmpdir+"/tmp", 0700)
+		if err != nil {
+			return err
+		}
+		defer os.RemoveAll(e.tmpdir + "/tmp")
+
 		build := exec.Command("singularity", "build", imageFilename, "docker-archive://"+e.tmpdir+"/image.tar")
+		build.Env = os.Environ()
+		build.Env = append(build.Env, "SINGULARITY_CACHEDIR="+e.tmpdir+"/cache")
+		build.Env = append(build.Env, "SINGULARITY_TMPDIR="+e.tmpdir+"/tmp")
 		e.logf("%v", build.Args)
 		out, err := build.CombinedOutput()
 		// INFO:    Starting build...
@@ -217,11 +248,16 @@ func (e *singularityExecutor) Create(spec containerSpec) error {
 	return nil
 }
 
-func (e *singularityExecutor) Start() error {
-	args := []string{"singularity", "exec", "--containall", "--no-home", "--cleanenv", "--pwd", e.spec.WorkingDir}
+func (e *singularityExecutor) execCmd(path string) *exec.Cmd {
+	args := []string{path, "exec", "--containall", "--cleanenv", "--pwd", e.spec.WorkingDir}
 	if !e.spec.EnableNetwork {
 		args = append(args, "--net", "--network=none")
 	}
+
+	if e.spec.CUDADeviceCount != 0 {
+		args = append(args, "--nv")
+	}
+
 	readonlyflag := map[bool]string{
 		false: "rw",
 		true:  "ro",
@@ -233,7 +269,12 @@ func (e *singularityExecutor) Start() error {
 	sort.Strings(binds)
 	for _, path := range binds {
 		mount := e.spec.BindMounts[path]
-		args = append(args, "--bind", mount.HostPath+":"+path+":"+readonlyflag[mount.ReadOnly])
+		if path == e.spec.Env["HOME"] {
+			// Singularity treates $HOME as special case
+			args = append(args, "--home", mount.HostPath+":"+path)
+		} else {
+			args = append(args, "--bind", mount.HostPath+":"+path+":"+readonlyflag[mount.ReadOnly])
+		}
 	}
 
 	// This is for singularity 3.5.2. There are some behaviors
@@ -243,21 +284,34 @@ func (e *singularityExecutor) Start() error {
 	env := make([]string, 0, len(e.spec.Env))
 	for k, v := range e.spec.Env {
 		if k == "HOME" {
-			// $HOME is a special case
-			args = append(args, "--home="+v)
-		} else {
-			env = append(env, "SINGULARITYENV_"+k+"="+v)
+			// Singularity treates $HOME as special case, this is handled
+			// with --home above
+			continue
 		}
+		env = append(env, "SINGULARITYENV_"+k+"="+v)
+	}
+
+	// Singularity always makes all nvidia devices visible to the
+	// container.  If a resource manager such as slurm or LSF told
+	// us to select specific devices we need to propagate that.
+	if cudaVisibleDevices := os.Getenv("CUDA_VISIBLE_DEVICES"); cudaVisibleDevices != "" {
+		// If a resource manager such as slurm or LSF told
+		// us to select specific devices we need to propagate that.
+		env = append(env, "SINGULARITYENV_CUDA_VISIBLE_DEVICES="+cudaVisibleDevices)
 	}
+	// Singularity's default behavior is to evaluate each
+	// SINGULARITYENV_* env var with a shell as a double-quoted
+	// string and pass the result to the contained
+	// process. Singularity 3.10+ has an option to pass env vars
+	// through literally without evaluating, which is what we
+	// want. See https://github.com/sylabs/singularity/pull/704
+	// and https://dev.arvados.org/issues/19081
+	env = append(env, "SINGULARITY_NO_EVAL=1")
 
 	args = append(args, e.imageFilename)
 	args = append(args, e.spec.Command...)
 
-	path, err := exec.LookPath(args[0])
-	if err != nil {
-		return err
-	}
-	child := &exec.Cmd{
+	return &exec.Cmd{
 		Path:   path,
 		Args:   args,
 		Env:    env,
@@ -265,6 +319,14 @@ func (e *singularityExecutor) Start() error {
 		Stdout: e.spec.Stdout,
 		Stderr: e.spec.Stderr,
 	}
+}
+
+func (e *singularityExecutor) Start() error {
+	path, err := exec.LookPath("singularity")
+	if err != nil {
+		return err
+	}
+	child := e.execCmd(path)
 	err = child.Start()
 	if err != nil {
 		return err