Nest restart fields in CheckRestart

api/tasks.go

@@ -79,36 +79,43 @@ func (r *RestartPolicy) Merge(rp *RestartPolicy) {
 	}
 }
 
+// CheckRestart describes if and when a task should be restarted based on
+// failing health checks.
+type CheckRestart struct {
+	Limit     int           `mapstructure:"limit"`
+	Grace     time.Duration `mapstructure:"grace_period"`
+	OnWarning bool          `mapstructure:"on_warning"`
+}
+
 // The ServiceCheck data model represents the consul health check that
 // Nomad registers for a Task
 type ServiceCheck struct {
-	Id             string
+	Id            string
-	Name           string
+	Name          string
-	Type           string
+	Type          string
-	Command        string
+	Command       string
-	Args           []string
+	Args          []string
-	Path           string
+	Path          string
-	Protocol       string
+	Protocol      string
-	PortLabel      string `mapstructure:"port"`
+	PortLabel     string `mapstructure:"port"`
-	Interval       time.Duration
+	Interval      time.Duration
-	Timeout        time.Duration
+	Timeout       time.Duration
-	InitialStatus  string `mapstructure:"initial_status"`
+	InitialStatus string `mapstructure:"initial_status"`
-	TLSSkipVerify  bool   `mapstructure:"tls_skip_verify"`
+	TLSSkipVerify bool   `mapstructure:"tls_skip_verify"`
-	Header         map[string][]string
+	Header        map[string][]string
-	Method         string
+	Method        string
-	RestartAfter   int
+	CheckRestart  *CheckRestart `mapstructure:"check_restart"`
-	RestartGrace   time.Duration
-	RestartWarning bool
 }
 
 // The Service model represents a Consul service definition
 type Service struct {
-	Id          string
+	Id           string
-	Name        string
+	Name         string
-	Tags        []string
+	Tags         []string
-	PortLabel   string `mapstructure:"port"`
+	PortLabel    string `mapstructure:"port"`
-	AddressMode string `mapstructure:"address_mode"`
+	AddressMode  string `mapstructure:"address_mode"`
-	Checks      []ServiceCheck
+	Checks       []ServiceCheck
+	CheckRestart *CheckRestart `mapstructure:"check_restart"`
 }
 
 func (s *Service) Canonicalize(t *Task, tg *TaskGroup, job *Job) {

command/agent/job_endpoint.go

@@ -685,27 +685,38 @@ func ApiTaskToStructsTask(apiTask *api.Task, structsTask *structs.Task) {
 				Tags:        service.Tags,
 				AddressMode: service.AddressMode,
 			}
+			if service.CheckRestart != nil {
+				structsTask.Services[i].CheckRestart = &structs.CheckRestart{
+					Limit:     service.CheckRestart.Limit,
+					Grace:     service.CheckRestart.Grace,
+					OnWarning: service.CheckRestart.OnWarning,
+				}
+			}
 
 			if l := len(service.Checks); l != 0 {
 				structsTask.Services[i].Checks = make([]*structs.ServiceCheck, l)
 				for j, check := range service.Checks {
 					structsTask.Services[i].Checks[j] = &structs.ServiceCheck{
-						Name:           check.Name,
+						Name:          check.Name,
-						Type:           check.Type,
+						Type:          check.Type,
-						Command:        check.Command,
+						Command:       check.Command,
-						Args:           check.Args,
+						Args:          check.Args,
-						Path:           check.Path,
+						Path:          check.Path,
-						Protocol:       check.Protocol,
+						Protocol:      check.Protocol,
-						PortLabel:      check.PortLabel,
+						PortLabel:     check.PortLabel,
-						Interval:       check.Interval,
+						Interval:      check.Interval,
-						Timeout:        check.Timeout,
+						Timeout:       check.Timeout,
-						InitialStatus:  check.InitialStatus,
+						InitialStatus: check.InitialStatus,
-						TLSSkipVerify:  check.TLSSkipVerify,
+						TLSSkipVerify: check.TLSSkipVerify,
-						Header:         check.Header,
+						Header:        check.Header,
-						Method:         check.Method,
+						Method:        check.Method,
-						RestartAfter:   check.RestartAfter,
+					}
-						RestartGrace:   check.RestartGrace,
+					if check.CheckRestart != nil {
-						RestartWarning: check.RestartWarning,
+						structsTask.Services[i].Checks[j].CheckRestart = &structs.CheckRestart{
+							Limit:     check.CheckRestart.Limit,
+							Grace:     check.CheckRestart.Grace,
+							OnWarning: check.CheckRestart.OnWarning,
+						}
 					}
 				}
 			}

nomad/structs/structs.go

@@ -2757,6 +2757,70 @@ func (tg *TaskGroup) GoString() string {
 	return fmt.Sprintf("*%#v", *tg)
 }
 
+// CheckRestart describes if and when a task should be restarted based on
+// failing health checks.
+type CheckRestart struct {
+	Limit     int           // Restart task after this many unhealthy intervals
+	Grace     time.Duration // Grace time to give tasks after starting to get healthy
+	OnWarning bool          // If true treat checks in `warning` as unhealthy
+}
+
+func (c *CheckRestart) Copy() *CheckRestart {
+	if c == nil {
+		return nil
+	}
+
+	nc := new(CheckRestart)
+	*nc = *c
+	return nc
+}
+
+// Merge non-zero values from other CheckRestart into a copy of this
+// CheckRestart. Returns nil iff both are nil.
+func (c *CheckRestart) Merge(o *CheckRestart) *CheckRestart {
+	if c == nil {
+		// Just return other
+		return o
+	}
+
+	nc := c.Copy()
+
+	if o == nil {
+		// Nothing to merge
+		return nc.Copy()
+	}
+
+	if nc.Limit == 0 {
+		nc.Limit = o.Limit
+	}
+
+	if nc.Grace == 0 {
+		nc.Grace = o.Grace
+	}
+
+	if !nc.OnWarning {
+		nc.OnWarning = o.OnWarning
+	}
+
+	return nc
+}
+
+func (c *CheckRestart) Validate() error {
+	if c == nil {
+		return nil
+	}
+
+	if c.Limit < 0 {
+		return fmt.Errorf("limit must be greater than or equal to 0 but found %d", c.Limit)
+	}
+
+	if c.Grace < 0 {
+		return fmt.Errorf("grace period must be greater than or equal to 0 but found %d", c.Grace)
+	}
+
+	return nil
+}
+
 const (
 	ServiceCheckHTTP   = "http"
 	ServiceCheckTCP    = "tcp"
@@ -2775,22 +2839,20 @@ const (
 // The ServiceCheck data model represents the consul health check that
 // Nomad registers for a Task
 type ServiceCheck struct {
-	Name           string              // Name of the check, defaults to id
+	Name          string              // Name of the check, defaults to id
-	Type           string              // Type of the check - tcp, http, docker and script
+	Type          string              // Type of the check - tcp, http, docker and script
-	Command        string              // Command is the command to run for script checks
+	Command       string              // Command is the command to run for script checks
-	Args           []string            // Args is a list of argumes for script checks
+	Args          []string            // Args is a list of argumes for script checks
-	Path           string              // path of the health check url for http type check
+	Path          string              // path of the health check url for http type check
-	Protocol       string              // Protocol to use if check is http, defaults to http
+	Protocol      string              // Protocol to use if check is http, defaults to http
-	PortLabel      string              // The port to use for tcp/http checks
+	PortLabel     string              // The port to use for tcp/http checks
-	Interval       time.Duration       // Interval of the check
+	Interval      time.Duration       // Interval of the check
-	Timeout        time.Duration       // Timeout of the response from the check before consul fails the check
+	Timeout       time.Duration       // Timeout of the response from the check before consul fails the check
-	InitialStatus  string              // Initial status of the check
+	InitialStatus string              // Initial status of the check
-	TLSSkipVerify  bool                // Skip TLS verification when Protocol=https
+	TLSSkipVerify bool                // Skip TLS verification when Protocol=https
-	Method         string              // HTTP Method to use (GET by default)
+	Method        string              // HTTP Method to use (GET by default)
-	Header         map[string][]string // HTTP Headers for Consul to set when making HTTP checks
+	Header        map[string][]string // HTTP Headers for Consul to set when making HTTP checks
-	RestartAfter   int                 // Restart task after this many unhealthy intervals
+	CheckRestart  *CheckRestart       // If and when a task should be restarted based on checks
-	RestartGrace   time.Duration       // Grace time to give tasks after starting to get healthy
-	RestartWarning bool                // If true treat checks in `warning` as unhealthy
 }
 
 func (sc *ServiceCheck) Copy() *ServiceCheck {
@@ -2801,6 +2863,7 @@ func (sc *ServiceCheck) Copy() *ServiceCheck {
 	*nsc = *sc
 	nsc.Args = helper.CopySliceString(sc.Args)
 	nsc.Header = helper.CopyMapStringSliceString(sc.Header)
+	nsc.CheckRestart = sc.CheckRestart.Copy()
 	return nsc
 }
 
@@ -2866,7 +2929,7 @@ func (sc *ServiceCheck) validate() error {
 
 	}
 
-	return nil
+	return sc.CheckRestart.Validate()
 }
 
 // RequiresPort returns whether the service check requires the task has a port.
@@ -2939,6 +3002,9 @@ type Service struct {
 
 	Tags   []string        // List of tags for the service
 	Checks []*ServiceCheck // List of checks associated with the service
+
+	// CheckRestart will be propagated to Checks if set.
+	CheckRestart *CheckRestart
 }
 
 func (s *Service) Copy() *Service {
@@ -2948,6 +3014,7 @@ func (s *Service) Copy() *Service {
 	ns := new(Service)
 	*ns = *s
 	ns.Tags = helper.CopySliceString(ns.Tags)
+	ns.CheckRestart = s.CheckRestart.Copy()
 
 	if s.Checks != nil {
 		checks := make([]*ServiceCheck, len(ns.Checks))
@@ -2983,6 +3050,14 @@ func (s *Service) Canonicalize(job string, taskGroup string, task string) {
 	for _, check := range s.Checks {
 		check.Canonicalize(s.Name)
 	}
+
+	// If CheckRestart is set propagate it to checks
+	if s.CheckRestart != nil {
+		for _, check := range s.Checks {
+			// Merge Service CheckRestart into Check's so Check's takes precedence
+			check.CheckRestart = check.CheckRestart.Merge(s.CheckRestart)
+		}
+	}
 }
 
 // Validate checks if the Check definition is valid
@@ -3016,6 +3091,11 @@ func (s *Service) Validate() error {
 			mErr.Errors = append(mErr.Errors, fmt.Errorf("check %s invalid: %v", c.Name, err))
 		}
 	}
+
+	if s.CheckRestart != nil && len(s.Checks) == 0 {
+		mErr.Errors = append(mErr.Errors, fmt.Errorf("check_restart specified but no checks"))
+	}
+
 	return mErr.ErrorOrNil()
 }