feat: purge old provisioner daemons (#10949)

coderd/database/dbpurge/dbpurge.go

@@ -9,12 +9,13 @@ import (
 	"golang.org/x/sync/errgroup"
 
 	"cdr.dev/slog"
+
 	"github.com/coder/coder/v2/coderd/database"
 	"github.com/coder/coder/v2/coderd/database/dbauthz"
 )
 
 const (
-	delay = 24 * time.Hour
+	delay = 10 * time.Minute
 )
 
 // New creates a new periodically purging database instance.
@@ -23,37 +24,47 @@ const (
 // This is for cleaning up old, unused resources from the database that take up space.
 func New(ctx context.Context, logger slog.Logger, db database.Store) io.Closer {
 	closed := make(chan struct{})
+
 	ctx, cancelFunc := context.WithCancel(ctx)
 	//nolint:gocritic // The system purges old db records without user input.
 	ctx = dbauthz.AsSystemRestricted(ctx)
+
+	// Use time.Nanosecond to force an initial tick. It will be reset to the
+	// correct duration after executing once.
+	ticker := time.NewTicker(time.Nanosecond)
+	doTick := func() {
+		defer ticker.Reset(delay)
+
+		var eg errgroup.Group
+		eg.Go(func() error {
+			return db.DeleteOldWorkspaceAgentLogs(ctx)
+		})
+		eg.Go(func() error {
+			return db.DeleteOldWorkspaceAgentStats(ctx)
+		})
+		eg.Go(func() error {
+			return db.DeleteOldProvisionerDaemons(ctx)
+		})
+		err := eg.Wait()
+		if err != nil {
+			if errors.Is(err, context.Canceled) {
+				return
+			}
+			logger.Error(ctx, "failed to purge old database entries", slog.Error(err))
+		}
+	}
+
 	go func() {
 		defer close(closed)
-
-		ticker := time.NewTicker(delay)
 		defer ticker.Stop()
 		for {
 			select {
 			case <-ctx.Done():
 				return
 			case <-ticker.C:
+				ticker.Stop()
+				doTick()
 			}
-
-			var eg errgroup.Group
-			eg.Go(func() error {
-				return db.DeleteOldWorkspaceAgentLogs(ctx)
-			})
-			eg.Go(func() error {
-				return db.DeleteOldWorkspaceAgentStats(ctx)
-			})
-			err := eg.Wait()
-			if err != nil {
-				if errors.Is(err, context.Canceled) {
-					return
-				}
-				logger.Error(ctx, "failed to purge old database entries", slog.Error(err))
-			}
-
-			ticker.Reset(delay)
 		}
 	}()
 	return &instance{

coderd/database/dbpurge/dbpurge_test.go

@@ -2,15 +2,23 @@ package dbpurge_test
 
 import (
 	"context"
+	"database/sql"
 	"testing"
+	"time"
 
-	"go.uber.org/goleak"
-
+	"github.com/google/uuid"
 	"github.com/stretchr/testify/require"
+	"go.uber.org/goleak"
+	"golang.org/x/exp/slices"
 
 	"cdr.dev/slog/sloggers/slogtest"
+
+	"github.com/coder/coder/v2/coderd/database"
 	"github.com/coder/coder/v2/coderd/database/dbmem"
 	"github.com/coder/coder/v2/coderd/database/dbpurge"
+	"github.com/coder/coder/v2/coderd/database/dbtestutil"
+	"github.com/coder/coder/v2/coderd/database/dbtime"
+	"github.com/coder/coder/v2/testutil"
 )
 
 func TestMain(m *testing.M) {
@@ -24,3 +32,72 @@ func TestPurge(t *testing.T) {
 	err := purger.Close()
 	require.NoError(t, err)
 }
+
+func TestDeleteOldProvisionerDaemons(t *testing.T) {
+	t.Parallel()
+
+	db, _ := dbtestutil.NewDB(t)
+	logger := slogtest.Make(t, &slogtest.Options{IgnoreErrors: true})
+
+	ctx, cancel := context.WithTimeout(context.Background(), testutil.WaitShort)
+	defer cancel()
+
+	now := dbtime.Now()
+
+	// given
+	_, err := db.InsertProvisionerDaemon(ctx, database.InsertProvisionerDaemonParams{
+		// Provisioner daemon created 14 days ago, and checked in just before 7 days deadline.
+		ID:           uuid.New(),
+		Name:         "external-0",
+		Provisioners: []database.ProvisionerType{"echo"},
+		CreatedAt:    now.Add(-14 * 24 * time.Hour),
+		UpdatedAt:    sql.NullTime{Valid: true, Time: now.Add(-7 * 24 * time.Hour).Add(time.Minute)},
+	})
+	require.NoError(t, err)
+	_, err = db.InsertProvisionerDaemon(ctx, database.InsertProvisionerDaemonParams{
+		// Provisioner daemon created 8 days ago, and checked in last time an hour after creation.
+		ID:           uuid.New(),
+		Name:         "external-1",
+		Provisioners: []database.ProvisionerType{"echo"},
+		CreatedAt:    now.Add(-8 * 24 * time.Hour),
+		UpdatedAt:    sql.NullTime{Valid: true, Time: now.Add(-8 * 24 * time.Hour).Add(time.Hour)},
+	})
+	require.NoError(t, err)
+	_, err = db.InsertProvisionerDaemon(ctx, database.InsertProvisionerDaemonParams{
+		// Provisioner daemon created 9 days ago, and never checked in.
+		ID:           uuid.New(),
+		Name:         "external-2",
+		Provisioners: []database.ProvisionerType{"echo"},
+		CreatedAt:    now.Add(-9 * 24 * time.Hour),
+	})
+	require.NoError(t, err)
+	_, err = db.InsertProvisionerDaemon(ctx, database.InsertProvisionerDaemonParams{
+		// Provisioner daemon created 6 days ago, and never checked in.
+		ID:           uuid.New(),
+		Name:         "external-3",
+		Provisioners: []database.ProvisionerType{"echo"},
+		CreatedAt:    now.Add(-6 * 24 * time.Hour),
+		UpdatedAt:    sql.NullTime{Valid: true, Time: now.Add(-6 * 24 * time.Hour)},
+	})
+	require.NoError(t, err)
+
+	// when
+	closer := dbpurge.New(ctx, logger, db)
+	defer closer.Close()
+
+	// then
+	require.Eventually(t, func() bool {
+		daemons, err := db.GetProvisionerDaemons(ctx)
+		if err != nil {
+			return false
+		}
+		return contains(daemons, "external-0") &&
+			contains(daemons, "external-3")
+	}, testutil.WaitShort, testutil.IntervalFast)
+}
+
+func contains(daemons []database.ProvisionerDaemon, name string) bool {
+	return slices.ContainsFunc(daemons, func(d database.ProvisionerDaemon) bool {
+		return d.Name == name
+	})
+}