12630: Use CUDADeviceCount instead of EnableCUDA

[arvados.git] / lib / crunchrun / crunchrun.go

diff --git a/lib/crunchrun/crunchrun.go b/lib/crunchrun/crunchrun.go
index ce060151dc133f232ea9bac237bc59de49612e21..589a046a34c0edb94ed97c48b7776cd63c1091cd 100644 (file)
--- a/lib/crunchrun/crunchrun.go
+++ b/lib/crunchrun/crunchrun.go
@@ -414,11 +414,14 @@ func (runner *ContainerRunner) SetupMounts() (map[string]bindmount, error) {
        arvMountCmd := []string{
                "arv-mount",
                "--foreground",
-               "--allow-other",
                "--read-write",
                "--storage-classes", strings.Join(runner.Container.OutputStorageClasses, ","),
                fmt.Sprintf("--crunchstat-interval=%v", runner.statInterval.Seconds())}
 
+       if runner.executor.Runtime() == "docker" {
+               arvMountCmd = append(arvMountCmd, "--allow-other")
+       }
+
        if runner.Container.RuntimeConstraints.KeepCacheRAM > 0 {
                arvMountCmd = append(arvMountCmd, "--file-cache", fmt.Sprintf("%d", runner.Container.RuntimeConstraints.KeepCacheRAM))
        }
@@ -614,10 +617,15 @@ func (runner *ContainerRunner) SetupMounts() (map[string]bindmount, error) {
        }
 
        if pdhOnly {
-               arvMountCmd = append(arvMountCmd, "--mount-by-pdh", "by_id")
+               // If we are only mounting collections by pdh, make
+               // sure we don't subscribe to websocket events to
+               // avoid putting undesired load on the API server
+               arvMountCmd = append(arvMountCmd, "--mount-by-pdh", "by_id", "--disable-event-listening")
        } else {
                arvMountCmd = append(arvMountCmd, "--mount-by-id", "by_id")
        }
+       // the by_uuid mount point is used by singularity when writing
+       // out docker images converted to SIF
        arvMountCmd = append(arvMountCmd, "--mount-by-id", "by_uuid")
        arvMountCmd = append(arvMountCmd, runner.ArvMountPoint)
 
@@ -979,19 +987,20 @@ func (runner *ContainerRunner) CreateContainer(imageID string, bindmounts map[st
        runner.executorStdout = stdout
        runner.executorStderr = stderr
        return runner.executor.Create(containerSpec{
-               Image:         imageID,
-               VCPUs:         runner.Container.RuntimeConstraints.VCPUs,
-               RAM:           ram,
-               WorkingDir:    workdir,
-               Env:           env,
-               BindMounts:    bindmounts,
-               Command:       runner.Container.Command,
-               EnableNetwork: enableNetwork,
-               NetworkMode:   runner.networkMode,
-               CgroupParent:  runner.setCgroupParent,
-               Stdin:         stdin,
-               Stdout:        stdout,
-               Stderr:        stderr,
+               Image:           imageID,
+               VCPUs:           runner.Container.RuntimeConstraints.VCPUs,
+               RAM:             ram,
+               WorkingDir:      workdir,
+               Env:             env,
+               BindMounts:      bindmounts,
+               Command:         runner.Container.Command,
+               EnableNetwork:   enableNetwork,
+               CUDADeviceCount: runner.Container.RuntimeConstraints.CUDADeviceCount,
+               NetworkMode:     runner.networkMode,
+               CgroupParent:    runner.setCgroupParent,
+               Stdin:           stdin,
+               Stdout:          stdout,
+               Stderr:          stderr,
        })
 }
 
@@ -1691,11 +1700,11 @@ func (command) RunCommand(prog string, args []string, stdin io.Reader, stdout, s
                ignoreDetachFlag = true
        }
 
-       if err := flags.Parse(args); err == flag.ErrHelp {
-               return 0
-       } else if err != nil {
-               log.Print(err)
-               return 1
+       if ok, code := cmd.ParseFlags(flags, prog, args, "container-uuid", stderr); !ok {
+               return code
+       } else if !*list && flags.NArg() != 1 {
+               fmt.Fprintf(stderr, "missing required argument: container-uuid (try -help)\n")
+               return 2
        }
 
        containerUUID := flags.Arg(0)
@@ -1775,14 +1784,36 @@ func (command) RunCommand(prog string, args []string, stdin io.Reader, stdout, s
                return 1
        }
 
-       if keepstore != nil {
-               w, err := cr.NewLogWriter("keepstore")
+       if keepstore == nil {
+               // Log explanation (if any) for why we're not running
+               // a local keepstore.
+               var buf bytes.Buffer
+               keepstoreLogbuf.SetWriter(&buf)
+               if buf.Len() > 0 {
+                       cr.CrunchLog.Printf("%s", strings.TrimSpace(buf.String()))
+               }
+       } else if logWhat := conf.Cluster.Containers.LocalKeepLogsToContainerLog; logWhat == "none" {
+               cr.CrunchLog.Printf("using local keepstore process (pid %d) at %s", keepstore.Process.Pid, os.Getenv("ARVADOS_KEEP_SERVICES"))
+               keepstoreLogbuf.SetWriter(io.Discard)
+       } else {
+               cr.CrunchLog.Printf("using local keepstore process (pid %d) at %s, writing logs to keepstore.txt in log collection", keepstore.Process.Pid, os.Getenv("ARVADOS_KEEP_SERVICES"))
+               logwriter, err := cr.NewLogWriter("keepstore")
                if err != nil {
                        log.Print(err)
                        return 1
                }
-               cr.keepstoreLogger = NewThrottledLogger(w)
-               err = keepstoreLogbuf.SetWriter(cr.keepstoreLogger)
+               cr.keepstoreLogger = NewThrottledLogger(logwriter)
+
+               var writer io.WriteCloser = cr.keepstoreLogger
+               if logWhat == "errors" {
+                       writer = &filterKeepstoreErrorsOnly{WriteCloser: writer}
+               } else if logWhat != "all" {
+                       // should have been caught earlier by
+                       // dispatcher's config loader
+                       log.Printf("invalid value for Containers.LocalKeepLogsToContainerLog: %q", logWhat)
+                       return 1
+               }
+               err = keepstoreLogbuf.SetWriter(writer)
                if err != nil {
                        log.Print(err)
                        return 1
@@ -1881,6 +1912,16 @@ func startLocalKeepstore(configData ConfigData, logbuf io.Writer) (*exec.Cmd, er
        if configData.Cluster == nil || configData.KeepBuffers < 1 {
                return nil, nil
        }
+       for uuid, vol := range configData.Cluster.Volumes {
+               if len(vol.AccessViaHosts) > 0 {
+                       fmt.Fprintf(logbuf, "not starting a local keepstore process because a volume (%s) uses AccessViaHosts\n", uuid)
+                       return nil, nil
+               }
+               if !vol.ReadOnly && vol.Replication < configData.Cluster.Collections.DefaultReplication {
+                       fmt.Fprintf(logbuf, "not starting a local keepstore process because a writable volume (%s) has replication less than Collections.DefaultReplication (%d < %d)\n", uuid, vol.Replication, configData.Cluster.Collections.DefaultReplication)
+                       return nil, nil
+               }
+       }
 
        // Rather than have an alternate way to tell keepstore how
        // many buffers to use when starting it this way, we just