// loadCheckState is used to restore the persisted state of a check.
func (a *Agent) loadCheckState(check *structs.HealthCheck) error {
// Try to read the persisted state for this check
file := filepath.Join(a.config.DataDir, checkStateDir, stringHash(check.CheckID))
buf, err := ioutil.ReadFile(file)
if err != nil {
if os.IsNotExist(err) {
return nil
}
return fmt.Errorf("failed reading file %q: %s", file, err)
}
// Decode the state data
var p persistedCheckState
if err := json.Unmarshal(buf, &p); err != nil {
return fmt.Errorf("failed decoding check state: %s", err)
}
// Check if the state has expired
if time.Now().Unix() >= p.Expires {
a.logger.Printf("[DEBUG] agent: check state expired for %q, not restoring", check.CheckID)
return a.purgeCheckState(check.CheckID)
}
// Restore the fields from the state
check.Output = p.Output
check.Status = p.Status
return nil
}
func TestAgentAntiEntropy_Checks(t *testing.T) {
conf := nextConfig()
dir, agent := makeAgent(t, conf)
defer os.RemoveAll(dir)
defer agent.Shutdown()
testutil.WaitForLeader(t, agent.RPC, "dc1")
// Register info
args := &structs.RegisterRequest{
Datacenter: "dc1",
Node: agent.config.NodeName,
Address: "127.0.0.1",
}
// Exists both, same (noop)
var out struct{}
chk1 := &structs.HealthCheck{
Node: agent.config.NodeName,
CheckID: "mysql",
Name: "mysql",
Status: structs.HealthPassing,
}
agent.state.AddCheck(chk1, "")
args.Check = chk1
if err := agent.RPC("Catalog.Register", args, &out); err != nil {
t.Fatalf("err: %v", err)
}
// Exists both, different (update)
chk2 := &structs.HealthCheck{
Node: agent.config.NodeName,
CheckID: "redis",
Name: "redis",
Status: structs.HealthPassing,
}
agent.state.AddCheck(chk2, "")
chk2_mod := new(structs.HealthCheck)
*chk2_mod = *chk2
chk2_mod.Status = structs.HealthCritical
args.Check = chk2_mod
if err := agent.RPC("Catalog.Register", args, &out); err != nil {
t.Fatalf("err: %v", err)
}
// Exists local (create)
chk3 := &structs.HealthCheck{
Node: agent.config.NodeName,
CheckID: "web",
Name: "web",
Status: structs.HealthPassing,
}
agent.state.AddCheck(chk3, "")
// Exists remote (delete)
chk4 := &structs.HealthCheck{
Node: agent.config.NodeName,
CheckID: "lb",
Name: "lb",
Status: structs.HealthPassing,
}
args.Check = chk4
if err := agent.RPC("Catalog.Register", args, &out); err != nil {
t.Fatalf("err: %v", err)
}
// Exists local, in sync, remote missing (create)
chk5 := &structs.HealthCheck{
Node: agent.config.NodeName,
CheckID: "cache",
Name: "cache",
Status: structs.HealthPassing,
}
agent.state.AddCheck(chk5, "")
agent.state.checkStatus["cache"] = syncStatus{inSync: true}
// Trigger anti-entropy run and wait
agent.StartSync()
time.Sleep(200 * time.Millisecond)
// Verify that we are in sync
req := structs.NodeSpecificRequest{
Datacenter: "dc1",
Node: agent.config.NodeName,
}
var checks structs.IndexedHealthChecks
if err := agent.RPC("Health.NodeChecks", &req, &checks); err != nil {
t.Fatalf("err: %v", err)
}
// We should have 5 checks (serf included)
if len(checks.HealthChecks) != 5 {
t.Fatalf("bad: %v", checks)
}
// All the checks should match
for _, chk := range checks.HealthChecks {
switch chk.CheckID {
case "mysql":
if !reflect.DeepEqual(chk, chk1) {
t.Fatalf("bad: %v %v", chk, chk1)
}
case "redis":
if !reflect.DeepEqual(chk, chk2) {
t.Fatalf("bad: %v %v", chk, chk2)
}
case "web":
if !reflect.DeepEqual(chk, chk3) {
t.Fatalf("bad: %v %v", chk, chk3)
}
case "cache":
if !reflect.DeepEqual(chk, chk5) {
t.Fatalf("bad: %v %v", chk, chk5)
}
case "serfHealth":
// ignore
default:
t.Fatalf("unexpected check: %v", chk)
}
}
// Check the local state
if len(agent.state.checks) != 4 {
t.Fatalf("bad: %v", agent.state.checks)
}
if len(agent.state.checkStatus) != 4 {
t.Fatalf("bad: %v", agent.state.checkStatus)
}
for name, status := range agent.state.checkStatus {
if !status.inSync {
t.Fatalf("should be in sync: %v %v", name, status)
}
}
}