17840: Deduplicate flag-parsing code.

[arvados.git] / services / crunch-dispatch-slurm / crunch-dispatch-slurm.go

diff --git a/services/crunch-dispatch-slurm/crunch-dispatch-slurm.go b/services/crunch-dispatch-slurm/crunch-dispatch-slurm.go
index 36bcef4f26b6cc3b3c4db53c88d9e6538b2ebad3..ff1077fae63ace1703e8c66ad95e54bb32d5a49a 100644 (file)
--- a/services/crunch-dispatch-slurm/crunch-dispatch-slurm.go
+++ b/services/crunch-dispatch-slurm/crunch-dispatch-slurm.go
@@ -7,7 +7,6 @@ package main
 // Dispatcher service for Crunch that submits containers to the slurm queue.
 
 import (
-       "bytes"
        "context"
        "flag"
        "fmt"
@@ -18,6 +17,7 @@ import (
        "strings"
        "time"
 
+       "git.arvados.org/arvados.git/lib/cmd"
        "git.arvados.org/arvados.git/lib/config"
        "git.arvados.org/arvados.git/lib/dispatchcloud"
        "git.arvados.org/arvados.git/sdk/go/arvados"
@@ -77,7 +77,7 @@ func (disp *Dispatcher) configure(prog string, args []string) error {
        if disp.logger == nil {
                disp.logger = logrus.StandardLogger()
        }
-       flags := flag.NewFlagSet(prog, flag.ExitOnError)
+       flags := flag.NewFlagSet(prog, flag.ContinueOnError)
        flags.Usage = func() { usage(flags) }
 
        loader := config.NewLoader(nil, disp.logger)
@@ -92,13 +92,9 @@ func (disp *Dispatcher) configure(prog string, args []string) error {
                false,
                "Print version information and exit.")
 
-       args = loader.MungeLegacyConfigArgs(logrus.StandardLogger(), args, "-legacy-crunch-dispatch-slurm-config")
-
-       // Parse args; omit the first arg which is the command name
-       err := flags.Parse(args)
-
-       if err == flag.ErrHelp {
-               return nil
+       args = loader.MungeLegacyConfigArgs(disp.logger, args, "-legacy-crunch-dispatch-slurm-config")
+       if ok, code := cmd.ParseFlags(flags, prog, args, "", os.Stderr); !ok {
+               os.Exit(code)
        }
 
        // Print version information if requested
@@ -202,7 +198,7 @@ var containerUuidPattern = regexp.MustCompile(`^[a-z0-9]{5}-dz642-[a-z0-9]{15}$`
 // Cancelled or Complete. See https://dev.arvados.org/issues/10979
 func (disp *Dispatcher) checkSqueueForOrphans() {
        for _, uuid := range disp.sqCheck.All() {
-               if !containerUuidPattern.MatchString(uuid) {
+               if !containerUuidPattern.MatchString(uuid) || !strings.HasPrefix(uuid, disp.cluster.ClusterID) {
                        continue
                }
                err := disp.TrackContainer(uuid)
@@ -255,6 +251,7 @@ func (disp *Dispatcher) submit(container arvados.Container, crunchRunCommand []s
        // append() here avoids modifying crunchRunCommand's
        // underlying array, which is shared with other goroutines.
        crArgs := append([]string(nil), crunchRunCommand...)
+       crArgs = append(crArgs, "--runtime-engine="+disp.cluster.Containers.RuntimeEngine)
        crArgs = append(crArgs, container.UUID)
        crScript := strings.NewReader(execScript(crArgs))
 
@@ -270,7 +267,7 @@ func (disp *Dispatcher) submit(container arvados.Container, crunchRunCommand []s
 // already in the queue).  Cancel the slurm job if the container's
 // priority changes to zero or its state indicates it's no longer
 // running.
-func (disp *Dispatcher) runContainer(_ *dispatch.Dispatcher, ctr arvados.Container, status <-chan arvados.Container) {
+func (disp *Dispatcher) runContainer(_ *dispatch.Dispatcher, ctr arvados.Container, status <-chan arvados.Container) error {
        ctx, cancel := context.WithCancel(context.Background())
        defer cancel()
 
@@ -278,37 +275,9 @@ func (disp *Dispatcher) runContainer(_ *dispatch.Dispatcher, ctr arvados.Contain
                log.Printf("Submitting container %s to slurm", ctr.UUID)
                cmd := []string{disp.cluster.Containers.CrunchRunCommand}
                cmd = append(cmd, disp.cluster.Containers.CrunchRunArgumentsList...)
-               if err := disp.submit(ctr, cmd); err != nil {
-                       var text string
-                       if err, ok := err.(dispatchcloud.ConstraintsNotSatisfiableError); ok {
-                               var logBuf bytes.Buffer
-                               fmt.Fprintf(&logBuf, "cannot run container %s: %s\n", ctr.UUID, err)
-                               if len(err.AvailableTypes) == 0 {
-                                       fmt.Fprint(&logBuf, "No instance types are configured.\n")
-                               } else {
-                                       fmt.Fprint(&logBuf, "Available instance types:\n")
-                                       for _, t := range err.AvailableTypes {
-                                               fmt.Fprintf(&logBuf,
-                                                       "Type %q: %d VCPUs, %d RAM, %d Scratch, %f Price\n",
-                                                       t.Name, t.VCPUs, t.RAM, t.Scratch, t.Price,
-                                               )
-                                       }
-                               }
-                               text = logBuf.String()
-                               disp.UpdateState(ctr.UUID, dispatch.Cancelled)
-                       } else {
-                               text = fmt.Sprintf("Error submitting container %s to slurm: %s", ctr.UUID, err)
-                       }
-                       log.Print(text)
-
-                       lr := arvadosclient.Dict{"log": arvadosclient.Dict{
-                               "object_uuid": ctr.UUID,
-                               "event_type":  "dispatch",
-                               "properties":  map[string]string{"text": text}}}
-                       disp.Arv.Create("logs", lr, nil)
-
-                       disp.Unlock(ctr.UUID)
-                       return
+               err := disp.submit(ctr, cmd)
+               if err != nil {
+                       return err
                }
        }
 
@@ -337,7 +306,7 @@ func (disp *Dispatcher) runContainer(_ *dispatch.Dispatcher, ctr arvados.Contain
                        case dispatch.Locked:
                                disp.Unlock(ctr.UUID)
                        }
-                       return
+                       return nil
                case updated, ok := <-status:
                        if !ok {
                                log.Printf("container %s is done: cancel slurm job", ctr.UUID)