Merge pull request #2337 from hashicorp/f-driver-errors

Drivers log during fingerprinting

client/driver/docker.go

@@ -29,6 +29,7 @@ import (
 	"github.com/hashicorp/nomad/client/driver/executor"
 	dstructs "github.com/hashicorp/nomad/client/driver/structs"
 	cstructs "github.com/hashicorp/nomad/client/structs"
+	"github.com/hashicorp/nomad/helper"
 	"github.com/hashicorp/nomad/helper/fields"
 	shelpers "github.com/hashicorp/nomad/helper/stats"
 	"github.com/hashicorp/nomad/nomad/structs"
@@ -108,6 +109,10 @@ type DockerDriver struct {
 
 	driverConfig *DockerDriverConfig
 	imageID      string
+
+	// A tri-state boolean to know if the fingerprinting has happened and
+	// whether it has been successful
+	fingerprintSuccess *bool
 }
 
 type DockerDriverAuth struct {
@@ -262,6 +267,48 @@ func NewDockerDriver(ctx *DriverContext) Driver {
 	return &DockerDriver{DriverContext: *ctx}
 }
 
+func (d *DockerDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool, error) {
+	// Initialize docker API clients
+	client, _, err := d.dockerClients()
+	if err != nil {
+		if d.fingerprintSuccess == nil || *d.fingerprintSuccess {
+			d.logger.Printf("[INFO] driver.docker: failed to initialize client: %s", err)
+		}
+		delete(node.Attributes, dockerDriverAttr)
+		d.fingerprintSuccess = helper.BoolToPtr(false)
+		return false, nil
+	}
+
+	// This is the first operation taken on the client so we'll try to
+	// establish a connection to the Docker daemon. If this fails it means
+	// Docker isn't available so we'll simply disable the docker driver.
+	env, err := client.Version()
+	if err != nil {
+		delete(node.Attributes, dockerDriverAttr)
+		if d.fingerprintSuccess == nil || *d.fingerprintSuccess {
+			d.logger.Printf("[DEBUG] driver.docker: could not connect to docker daemon at %s: %s", client.Endpoint(), err)
+		}
+		d.fingerprintSuccess = helper.BoolToPtr(false)
+		return false, nil
+	}
+
+	node.Attributes[dockerDriverAttr] = "1"
+	node.Attributes["driver.docker.version"] = env.Get("Version")
+
+	privileged := d.config.ReadBoolDefault(dockerPrivilegedConfigOption, false)
+	if privileged {
+		node.Attributes[dockerPrivilegedConfigOption] = "1"
+	}
+
+	// Advertise if this node supports Docker volumes
+	if d.config.ReadBoolDefault(dockerVolumesConfigOption, dockerVolumesConfigDefault) {
+		node.Attributes["driver."+dockerVolumesConfigOption] = "1"
+	}
+
+	d.fingerprintSuccess = helper.BoolToPtr(true)
+	return true, nil
+}
+
 // Validate is used to validate the driver configuration
 func (d *DockerDriver) Validate(config map[string]interface{}) error {
 	fd := &fields.FieldData{
@@ -618,49 +665,6 @@ func (d *DockerDriver) dockerClients() (*docker.Client, *docker.Client, error) {
 	return client, waitClient, merr.ErrorOrNil()
 }
 
-func (d *DockerDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool, error) {
-	// Get the current status so that we can log any debug messages only if the
-	// state changes
-	_, currentlyEnabled := node.Attributes[dockerDriverAttr]
-
-	// Initialize docker API clients
-	client, _, err := d.dockerClients()
-	if err != nil {
-		delete(node.Attributes, dockerDriverAttr)
-		if currentlyEnabled {
-			d.logger.Printf("[INFO] driver.docker: failed to initialize client: %s", err)
-		}
-		return false, nil
-	}
-
-	privileged := d.config.ReadBoolDefault(dockerPrivilegedConfigOption, false)
-	if privileged {
-		node.Attributes[dockerPrivilegedConfigOption] = "1"
-	}
-
-	// This is the first operation taken on the client so we'll try to
-	// establish a connection to the Docker daemon. If this fails it means
-	// Docker isn't available so we'll simply disable the docker driver.
-	env, err := client.Version()
-	if err != nil {
-		if currentlyEnabled {
-			d.logger.Printf("[DEBUG] driver.docker: could not connect to docker daemon at %s: %s", client.Endpoint(), err)
-		}
-		delete(node.Attributes, dockerDriverAttr)
-		return false, nil
-	}
-
-	node.Attributes[dockerDriverAttr] = "1"
-	node.Attributes["driver.docker.version"] = env.Get("Version")
-
-	// Advertise if this node supports Docker volumes
-	if d.config.ReadBoolDefault(dockerVolumesConfigOption, dockerVolumesConfigDefault) {
-		node.Attributes["driver."+dockerVolumesConfigOption] = "1"
-	}
-
-	return true, nil
-}
-
 func (d *DockerDriver) containerBinds(driverConfig *DockerDriverConfig, taskDir *allocdir.TaskDir,
 	task *structs.Task) ([]string, error) {
 

client/driver/exec.go

@@ -29,6 +29,10 @@ const (
 // features.
 type ExecDriver struct {
 	DriverContext
+
+	// A tri-state boolean to know if the fingerprinting has happened and
+	// whether it has been successful
+	fingerprintSuccess *bool
 }
 
 type ExecDriverConfig struct {

client/driver/exec_default.go

@@ -4,9 +4,11 @@ package driver
 
 import (
 	"github.com/hashicorp/nomad/client/config"
+	"github.com/hashicorp/nomad/helper"
 	"github.com/hashicorp/nomad/nomad/structs"
 )
 
 func (d *ExecDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool, error) {
+	d.fingerprintSuccess = helper.BoolToPtr(false)
 	return false, nil
 }

client/driver/exec_linux.go

@@ -2,33 +2,33 @@ package driver
 
 import (
 	"github.com/hashicorp/nomad/client/config"
+	"github.com/hashicorp/nomad/helper"
 	"github.com/hashicorp/nomad/nomad/structs"
 	"golang.org/x/sys/unix"
 )
 
 func (d *ExecDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool, error) {
-	// Get the current status so that we can log any debug messages only if the
-	// state changes
-	_, currentlyEnabled := node.Attributes[execDriverAttr]
-
 	// Only enable if cgroups are available and we are root
-	if _, ok := node.Attributes["unique.cgroup.mountpoint"]; !ok {
-		if currentlyEnabled {
+	if !cgroupsMounted(node) {
+		if d.fingerprintSuccess == nil || *d.fingerprintSuccess {
 			d.logger.Printf("[DEBUG] driver.exec: cgroups unavailable, disabling")
 		}
+		d.fingerprintSuccess = helper.BoolToPtr(false)
 		delete(node.Attributes, execDriverAttr)
 		return false, nil
 	} else if unix.Geteuid() != 0 {
-		if currentlyEnabled {
+		if d.fingerprintSuccess == nil || *d.fingerprintSuccess {
 			d.logger.Printf("[DEBUG] driver.exec: must run as root user, disabling")
 		}
 		delete(node.Attributes, execDriverAttr)
+		d.fingerprintSuccess = helper.BoolToPtr(false)
 		return false, nil
 	}
 
-	if !currentlyEnabled {
+	if d.fingerprintSuccess == nil || *d.fingerprintSuccess {
 		d.logger.Printf("[DEBUG] driver.exec: exec driver is enabled")
 	}
 	node.Attributes[execDriverAttr] = "1"
+	d.fingerprintSuccess = helper.BoolToPtr(true)
 	return true, nil
 }

client/driver/java.go

@@ -23,6 +23,7 @@ import (
 	dstructs "github.com/hashicorp/nomad/client/driver/structs"
 	"github.com/hashicorp/nomad/client/fingerprint"
 	cstructs "github.com/hashicorp/nomad/client/structs"
+	"github.com/hashicorp/nomad/helper"
 	"github.com/hashicorp/nomad/helper/fields"
 	"github.com/hashicorp/nomad/nomad/structs"
 )
@@ -38,6 +39,10 @@ const (
 type JavaDriver struct {
 	DriverContext
 	fingerprint.StaticFingerprinter
+
+	// A tri-state boolean to know if the fingerprinting has happened and
+	// whether it has been successful
+	fingerprintSuccess *bool
 }
 
 type JavaDriverConfig struct {
@@ -105,16 +110,13 @@ func (d *JavaDriver) Abilities() DriverAbilities {
 }
 
 func (d *JavaDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool, error) {
-	// Get the current status so that we can log any debug messages only if the
-	// state changes
-	_, currentlyEnabled := node.Attributes[javaDriverAttr]
-
 	// Only enable if we are root and cgroups are mounted when running on linux systems.
-	if runtime.GOOS == "linux" && (syscall.Geteuid() != 0 || !d.cgroupsMounted(node)) {
-		if currentlyEnabled {
+	if runtime.GOOS == "linux" && (syscall.Geteuid() != 0 || !cgroupsMounted(node)) {
+		if d.fingerprintSuccess == nil || *d.fingerprintSuccess {
 			d.logger.Printf("[DEBUG] driver.java: root priviledges and mounted cgroups required on linux, disabling")
 		}
 		delete(node.Attributes, "driver.java")
+		d.fingerprintSuccess = helper.BoolToPtr(false)
 		return false, nil
 	}
 
@@ -128,6 +130,7 @@ func (d *JavaDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool,
 	if err != nil {
 		// assume Java wasn't found
 		delete(node.Attributes, javaDriverAttr)
+		d.fingerprintSuccess = helper.BoolToPtr(false)
 		return false, nil
 	}
 
@@ -143,10 +146,11 @@ func (d *JavaDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool,
 	}
 
 	if infoString == "" {
-		if currentlyEnabled {
+		if d.fingerprintSuccess == nil || *d.fingerprintSuccess {
 			d.logger.Println("[WARN] driver.java: error parsing Java version information, aborting")
 		}
 		delete(node.Attributes, javaDriverAttr)
+		d.fingerprintSuccess = helper.BoolToPtr(false)
 		return false, nil
 	}
 
@@ -163,6 +167,7 @@ func (d *JavaDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool,
 	node.Attributes["driver.java.version"] = versionString
 	node.Attributes["driver.java.runtime"] = info[1]
 	node.Attributes["driver.java.vm"] = info[2]
+	d.fingerprintSuccess = helper.BoolToPtr(true)
 
 	return true, nil
 }
@@ -295,13 +300,6 @@ func (d *JavaDriver) Start(ctx *ExecContext, task *structs.Task) (DriverHandle,
 
 func (d *JavaDriver) Cleanup(*ExecContext, *CreatedResources) error { return nil }
 
-// cgroupsMounted returns true if the cgroups are mounted on a system otherwise
-// returns false
-func (d *JavaDriver) cgroupsMounted(node *structs.Node) bool {
-	_, ok := node.Attributes["unique.cgroup.mountpoint"]
-	return ok
-}
-
 type javaId struct {
 	Version         string
 	KillTimeout     time.Duration

client/driver/qemu.go

@@ -105,10 +105,6 @@ func (d *QemuDriver) FSIsolation() cstructs.FSIsolation {
 }
 
 func (d *QemuDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool, error) {
-	// Get the current status so that we can log any debug messages only if the
-	// state changes
-	_, currentlyEnabled := node.Attributes[qemuDriverAttr]
-
 	bin := "qemu-system-x86_64"
 	if runtime.GOOS == "windows" {
 		// On windows, the "qemu-system-x86_64" command does not respond to the
@@ -128,9 +124,6 @@ func (d *QemuDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool,
 		return false, fmt.Errorf("Unable to parse Qemu version string: %#v", matches)
 	}
 
-	if !currentlyEnabled {
-		d.logger.Printf("[DEBUG] driver.qemu: enabling driver")
-	}
 	node.Attributes[qemuDriverAttr] = "1"
 	node.Attributes["driver.qemu.version"] = matches[1]
 	return true, nil

client/driver/raw_exec.go

@@ -87,17 +87,11 @@ func (d *RawExecDriver) FSIsolation() cstructs.FSIsolation {
 }
 
 func (d *RawExecDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool, error) {
-	// Get the current status so that we can log any debug messages only if the
-	// state changes
-	_, currentlyEnabled := node.Attributes[rawExecDriverAttr]
-
 	// Check that the user has explicitly enabled this executor.
 	enabled := cfg.ReadBoolDefault(rawExecConfigOption, false)
 
 	if enabled || cfg.DevMode {
-		if currentlyEnabled {
-			d.logger.Printf("[WARN] driver.raw_exec: raw exec is enabled. Only enable if needed")
-		}
+		d.logger.Printf("[WARN] driver.raw_exec: raw exec is enabled. Only enable if needed")
 		node.Attributes[rawExecDriverAttr] = "1"
 		return true, nil
 	}

client/driver/rkt.go

@@ -23,6 +23,7 @@ import (
 	"github.com/hashicorp/nomad/client/driver/executor"
 	dstructs "github.com/hashicorp/nomad/client/driver/structs"
 	cstructs "github.com/hashicorp/nomad/client/structs"
+	"github.com/hashicorp/nomad/helper"
 	"github.com/hashicorp/nomad/helper/fields"
 	"github.com/hashicorp/nomad/nomad/structs"
 	"github.com/mitchellh/mapstructure"
@@ -57,6 +58,10 @@ const (
 // planned in the future
 type RktDriver struct {
 	DriverContext
+
+	// A tri-state boolean to know if the fingerprinting has happened and
+	// whether it has been successful
+	fingerprintSuccess *bool
 }
 
 type RktDriverConfig struct {
@@ -157,22 +162,20 @@ func (d *RktDriver) Abilities() DriverAbilities {
 }
 
 func (d *RktDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool, error) {
-	// Get the current status so that we can log any debug messages only if the
-	// state changes
-	_, currentlyEnabled := node.Attributes[rktDriverAttr]
-
 	// Only enable if we are root when running on non-windows systems.
 	if runtime.GOOS != "windows" && syscall.Geteuid() != 0 {
-		if currentlyEnabled {
+		if d.fingerprintSuccess == nil || *d.fingerprintSuccess {
 			d.logger.Printf("[DEBUG] driver.rkt: must run as root user, disabling")
 		}
 		delete(node.Attributes, rktDriverAttr)
+		d.fingerprintSuccess = helper.BoolToPtr(false)
 		return false, nil
 	}
 
 	outBytes, err := exec.Command(rktCmd, "version").Output()
 	if err != nil {
 		delete(node.Attributes, rktDriverAttr)
+		d.fingerprintSuccess = helper.BoolToPtr(false)
 		return false, nil
 	}
 	out := strings.TrimSpace(string(outBytes))
@@ -181,6 +184,7 @@ func (d *RktDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool, e
 	appcMatches := reAppcVersion.FindStringSubmatch(out)
 	if len(rktMatches) != 2 || len(appcMatches) != 2 {
 		delete(node.Attributes, rktDriverAttr)
+		d.fingerprintSuccess = helper.BoolToPtr(false)
 		return false, fmt.Errorf("Unable to parse Rkt version string: %#v", rktMatches)
 	}
 
@@ -200,6 +204,7 @@ func (d *RktDriver) Fingerprint(cfg *config.Config, node *structs.Node) (bool, e
 	if d.config.ReadBoolDefault(rktVolumesConfigOption, rktVolumesConfigDefault) {
 		node.Attributes["driver."+rktVolumesConfigOption] = "1"
 	}
+	d.fingerprintSuccess = helper.BoolToPtr(true)
 	return true, nil
 }
 

client/driver/utils.go

@@ -19,6 +19,13 @@ import (
 	"github.com/hashicorp/nomad/nomad/structs"
 )
 
+// cgroupsMounted returns true if the cgroups are mounted on a system otherwise
+// returns false
+func cgroupsMounted(node *structs.Node) bool {
+	_, ok := node.Attributes["unique.cgroup.mountpoint"]
+	return ok
+}
+
 // createExecutor launches an executor plugin and returns an instance of the
 // Executor interface
 func createExecutor(w io.Writer, clientConfig *config.Config,