Configure Node GC Threshold

command/agent/agent.go

@@ -8,6 +8,7 @@ import (
 	"os"
 	"path/filepath"
 	"sync"
+	"time"
 
 	"github.com/hashicorp/nomad/client"
 	"github.com/hashicorp/nomad/nomad"
@@ -137,6 +138,14 @@ func (a *Agent) serverConfig() (*nomad.Config, error) {
 		conf.SerfConfig.MemberlistConfig.BindPort = port
 	}
 
+	if gcThreshold := a.config.Server.NodeGCThreshold; gcThreshold != "" {
+		dur, err := time.ParseDuration(gcThreshold)
+		if err != nil {
+			return nil, err
+		}
+		conf.NodeGCThreshold = dur
+	}
+
 	return conf, nil
 }
 

command/agent/agent_test.go

@@ -148,6 +148,17 @@ func TestAgent_ServerConfig(t *testing.T) {
 		t.Fatalf("expect 127.0.0.2, got: %s", addr)
 	}
 
+	conf.Server.NodeGCThreshold = "42g"
+	out, err = a.serverConfig()
+	if err == nil || !strings.Contains(err.Error(), "unknown unit") {
+		t.Fatalf("expected unknown unit error, got: %#v", err)
+	}
+	conf.Server.NodeGCThreshold = "10s"
+	out, err = a.serverConfig()
+	if threshold := out.NodeGCThreshold; threshold != time.Second*10 {
+		t.Fatalf("expect 10s, got: %s", threshold)
+	}
+
 	// Defaults to the global bind addr
 	conf.Addresses.RPC = ""
 	conf.Addresses.Serf = ""

command/agent/config.go

@@ -178,6 +178,9 @@ type ServerConfig struct {
 	// enabled for this server to handle. This will restrict the evaluations
 	// that the workers dequeue for processing.
 	EnabledSchedulers []string `hcl:"enabled_schedulers"`
+
+	// NodeGCThreshold contros how "old" a node must be to be collected by GC.
+	NodeGCThreshold string `hcl:"node_gc_threshold"`
 }
 
 // Telemetry is the telemetry configuration for the server
@@ -372,6 +375,9 @@ func (a *ServerConfig) Merge(b *ServerConfig) *ServerConfig {
 	if b.NumSchedulers != 0 {
 		result.NumSchedulers = b.NumSchedulers
 	}
+	if b.NodeGCThreshold != "" {
+		result.NodeGCThreshold = b.NodeGCThreshold
+	}
 
 	// Add the schedulers
 	result.EnabledSchedulers = append(result.EnabledSchedulers, b.EnabledSchedulers...)

command/agent/config_test.go

@@ -47,6 +47,7 @@ func TestConfig_Merge(t *testing.T) {
 			DataDir:         "/tmp/data1",
 			ProtocolVersion: 1,
 			NumSchedulers:   1,
+			NodeGCThreshold: "1h",
 		},
 		Ports: &Ports{
 			HTTP: 4646,
@@ -106,6 +107,7 @@ func TestConfig_Merge(t *testing.T) {
 			ProtocolVersion:   2,
 			NumSchedulers:     2,
 			EnabledSchedulers: []string{structs.JobTypeBatch},
+			NodeGCThreshold:   "12h",
 		},
 		Ports: &Ports{
 			HTTP: 20000,
@@ -369,6 +371,7 @@ func TestConfig_LoadConfigString(t *testing.T) {
 			ProtocolVersion:   3,
 			NumSchedulers:     2,
 			EnabledSchedulers: []string{"test"},
+			NodeGCThreshold:   "12h",
 		},
 		Telemetry: &Telemetry{
 			StatsiteAddr:    "127.0.0.1:1234",
@@ -441,6 +444,7 @@ server {
 	protocol_version = 3
 	num_schedulers = 2
 	enabled_schedulers = ["test"]
+	node_gc_threshold = "12h"
 }
 telemetry {
 	statsite_address = "127.0.0.1:1234"

website/source/docs/agent/config.html.md

@@ -173,6 +173,10 @@ configured on client nodes.
     sub-schedulers this server will handle. This can be used to restrict the
     evaluations that worker threads will dequeue for processing. This
     defaults to all available schedulers.
+  * `node_gc_threshold` This is a string with a unit suffix, such as "300ms",
+    "1.5h" or "25m". Valid time units are "ns", "us" (or "µs"), "ms", "s",
+    "m", "h". Controls how long a node must be in a terminal state before it is
+    garbage collected and purged from the system.
 
 ## Client-specific Options