Merge branch '17749-s3-prefixes'

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

diff --git a/lib/crunchrun/crunchrun.go b/lib/crunchrun/crunchrun.go
index e15303a3155afe81d72e8ce61e881ce76d5282d7..42f143f1cb8fe6bb97bfff2511e9f318f37359af 100644 (file)
--- a/lib/crunchrun/crunchrun.go
+++ b/lib/crunchrun/crunchrun.go
@@ -66,7 +66,7 @@ type IKeepClient interface {
 // NewLogWriter is a factory function to create a new log writer.
 type NewLogWriter func(name string) (io.WriteCloser, error)
 
-type RunArvMount func(args []string, tok string) (*exec.Cmd, error)
+type RunArvMount func(cmdline []string, tok string) (*exec.Cmd, error)
 
 type MkTempDir func(string, string) (string, error)
 
@@ -273,8 +273,8 @@ func (runner *ContainerRunner) LoadImage() (string, error) {
        return imageID, nil
 }
 
-func (runner *ContainerRunner) ArvMountCmd(arvMountCmd []string, token string) (c *exec.Cmd, err error) {
-       c = exec.Command("arv-mount", arvMountCmd...)
+func (runner *ContainerRunner) ArvMountCmd(cmdline []string, token string) (c *exec.Cmd, err error) {
+       c = exec.Command(cmdline[0], cmdline[1:]...)
 
        // Copy our environment, but override ARVADOS_API_TOKEN with
        // the container auth token.
@@ -291,8 +291,16 @@ func (runner *ContainerRunner) ArvMountCmd(arvMountCmd []string, token string) (
                return nil, err
        }
        runner.arvMountLog = NewThrottledLogger(w)
+       scanner := logScanner{
+               Patterns: []string{
+                       "Keep write error",
+                       "Block not found error",
+                       "Unhandled exception during FUSE operation",
+               },
+               ReportFunc: runner.reportArvMountWarning,
+       }
        c.Stdout = runner.arvMountLog
-       c.Stderr = io.MultiWriter(runner.arvMountLog, os.Stderr)
+       c.Stderr = io.MultiWriter(runner.arvMountLog, os.Stderr, &scanner)
 
        runner.CrunchLog.Printf("Running %v", c.Args)
 
@@ -392,6 +400,7 @@ func (runner *ContainerRunner) SetupMounts() (map[string]bindmount, error) {
        pdhOnly := true
        tmpcount := 0
        arvMountCmd := []string{
+               "arv-mount",
                "--foreground",
                "--allow-other",
                "--read-write",
@@ -1100,6 +1109,21 @@ func (runner *ContainerRunner) updateLogs() {
        }
 }
 
+func (runner *ContainerRunner) reportArvMountWarning(pattern, text string) {
+       var updated arvados.Container
+       err := runner.DispatcherArvClient.Update("containers", runner.Container.UUID, arvadosclient.Dict{
+               "container": arvadosclient.Dict{
+                       "runtime_status": arvadosclient.Dict{
+                               "warning":       "arv-mount: " + pattern,
+                               "warningDetail": text,
+                       },
+               },
+       }, &updated)
+       if err != nil {
+               runner.CrunchLog.Printf("error updating container runtime_status: %s", err)
+       }
+}
+
 // CaptureOutput saves data from the container's output directory if
 // needed, and updates the container output accordingly.
 func (runner *ContainerRunner) CaptureOutput(bindmounts map[string]bindmount) error {
@@ -1170,6 +1194,7 @@ func (runner *ContainerRunner) CleanupDirs() {
 
                if umnterr != nil {
                        runner.CrunchLog.Printf("Error unmounting: %v", umnterr)
+                       runner.ArvMount.Process.Kill()
                } else {
                        // If arv-mount --unmount gets stuck for any reason, we
                        // don't want to wait for it forever.  Do Wait() in a goroutine
@@ -1376,7 +1401,7 @@ func (runner *ContainerRunner) NewArvLogWriter(name string) (io.WriteCloser, err
 // Run the full container lifecycle.
 func (runner *ContainerRunner) Run() (err error) {
        runner.CrunchLog.Printf("crunch-run %s started", cmd.Version.String())
-       runner.CrunchLog.Printf("Executing container '%s'", runner.Container.UUID)
+       runner.CrunchLog.Printf("Executing container '%s' using %s runtime", runner.Container.UUID, runner.executor.Runtime())
 
        hostname, hosterr := os.Hostname()
        if hosterr != nil {