| 1 | package daemon |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "os" |
| 6 | "strings" |
| 7 | "testing" |
| 8 | "time" |
| 9 | |
| 10 | "github.com/FelineStateMachine/tinyrelay/internal/event" |
| 11 | "github.com/FelineStateMachine/tinyrelay/internal/replication" |
| 12 | ) |
| 13 | |
| 14 | func TestDaemonBackupJobResumesAfterRestart(t *testing.T) { |
| 15 | ctx := context.Background() |
| 16 | dir := t.TempDir() |
| 17 | secret := strings.Repeat("0" , 63) + "1" |
| 18 | owner, err := event.PublicKey(secret) |
| 19 | if err != nil { |
| 20 | t.Fatal(err) |
| 21 | } |
| 22 | config := Config{DataDir: dir, DefaultTenant: "main" } |
| 23 | app, err := New(ctx, config) |
| 24 | if err != nil { |
| 25 | t.Fatal(err) |
| 26 | } |
| 27 | meta, err := app.Create(ctx, CreateOptions{Name: "main" , Owner: owner, Template: "default" }) |
| 28 | if err != nil { |
| 29 | t.Fatal(err) |
| 30 | } |
| 31 | if err := app.Start(ctx, "http://relay.test" ); err != nil { |
| 32 | t.Fatal(err) |
| 33 | } |
| 34 | tenant := app.tenants[meta.ID] |
| 35 | tenant.workCancel() |
| 36 | tenant.workWG.Wait() |
| 37 | if err := tenant.replication.AddJob(ctx, replication.JobSpec{ID: "restart-backup" , Kind: replication.JobBackup, Relays: []string{"restart.backup.json" }}); err != nil { |
| 38 | t.Fatal(err) |
| 39 | } |
| 40 | if err := app.Close(ctx); err != nil { |
| 41 | t.Fatal(err) |
| 42 | } |
| 43 | |
| 44 | app, err = New(ctx, config) |
| 45 | if err != nil { |
| 46 | t.Fatal(err) |
| 47 | } |
| 48 | defer app.Close(ctx) |
| 49 | if err := app.Start(ctx, "http://relay.test" ); err != nil { |
| 50 | t.Fatal(err) |
| 51 | } |
| 52 | reloaded := app.tenants[meta.ID] |
| 53 | if !waitFor(t, 8*time.Second, func() bool { |
| 54 | statuses, statusErr := reloaded.replication.ListJobStatus(ctx) |
| 55 | return statusErr == nil && len(statuses) == 1 && statuses[0].Phase == "complete" |
| 56 | }) { |
| 57 | statuses, _ := reloaded.replication.ListJobStatus(ctx) |
| 58 | t.Fatalf("backup job did not resume: %#v" , statuses) |
| 59 | } |
| 60 | if _, err := os.Stat(meta.Paths.Root + "/restart.backup.json" ); err != nil { |
| 61 | t.Fatalf("backup artifact missing after restart: %v" , err) |
| 62 | } |
| 63 | } |
| 64 | |
| 65 | func TestDaemonFailedJobStatusSurvivesRestart(t *testing.T) { |
| 66 | ctx := context.Background() |
| 67 | dir := t.TempDir() |
| 68 | secret := strings.Repeat("0" , 63) + "1" |
| 69 | owner, err := event.PublicKey(secret) |
| 70 | if err != nil { |
| 71 | t.Fatal(err) |
| 72 | } |
| 73 | config := Config{DataDir: dir, DefaultTenant: "main" , AllowPrivateRelays: true} |
| 74 | app, err := New(ctx, config) |
| 75 | if err != nil { |
| 76 | t.Fatal(err) |
| 77 | } |
| 78 | meta, err := app.Create(ctx, CreateOptions{Name: "main" , Owner: owner, Template: "default" }) |
| 79 | if err != nil { |
| 80 | t.Fatal(err) |
| 81 | } |
| 82 | if err := app.Start(ctx, "http://relay.test" ); err != nil { |
| 83 | t.Fatal(err) |
| 84 | } |
| 85 | tenant := app.tenants[meta.ID] |
| 86 | tenant.workCancel() |
| 87 | tenant.workWG.Wait() |
| 88 | if err := tenant.replication.AddJob(ctx, replication.JobSpec{ID: "restart-failure" , Kind: replication.JobPull, Relays: []string{"ws://127.0.0.1:1" }}); err != nil { |
| 89 | t.Fatal(err) |
| 90 | } |
| 91 | if err := app.Close(ctx); err != nil { |
| 92 | t.Fatal(err) |
| 93 | } |
| 94 | |
| 95 | app, err = New(ctx, config) |
| 96 | if err != nil { |
| 97 | t.Fatal(err) |
| 98 | } |
| 99 | defer app.Close(ctx) |
| 100 | if err := app.Start(ctx, "http://relay.test" ); err != nil { |
| 101 | t.Fatal(err) |
| 102 | } |
| 103 | reloaded := app.tenants[meta.ID] |
| 104 | if !waitFor(t, 8*time.Second, func() bool { |
| 105 | statuses, statusErr := reloaded.replication.ListJobStatus(ctx) |
| 106 | return statusErr == nil && len(statuses) == 1 && statuses[0].Phase == "failed" && statuses[0].Error != "" |
| 107 | }) { |
| 108 | statuses, _ := reloaded.replication.ListJobStatus(ctx) |
| 109 | t.Fatalf("failed job status was not recorded after restart: %#v" , statuses) |
| 110 | } |
| 111 | pending, err := reloaded.replication.Queue().Pending(ctx, "job" ) |
| 112 | if err != nil { |
| 113 | t.Fatal(err) |
| 114 | } |
| 115 | if len(pending) != 1 || pending[0].Attempts == 0 { |
| 116 | t.Fatalf("failed job lost retry state: %#v" , pending) |
| 117 | } |
| 118 | } |