chore: optimize GetPrebuiltWorkspaces query (#18717)

* Adds GetRunningPrebuiltWorkspacesOptimized query
* Runs both original and updated query side-by-side and logs diffs

coderd/database/dbauthz/dbauthz.go

@@ -2596,6 +2596,14 @@ func (q *querier) GetRunningPrebuiltWorkspaces(ctx context.Context) ([]database.
 	return q.db.GetRunningPrebuiltWorkspaces(ctx)
 }
 
+func (q *querier) GetRunningPrebuiltWorkspacesOptimized(ctx context.Context) ([]database.GetRunningPrebuiltWorkspacesOptimizedRow, error) {
+	// This query returns only prebuilt workspaces, but we decided to require permissions for all workspaces.
+	if err := q.authorizeContext(ctx, policy.ActionRead, rbac.ResourceWorkspace.All()); err != nil {
+		return nil, err
+	}
+	return q.db.GetRunningPrebuiltWorkspacesOptimized(ctx)
+}
+
 func (q *querier) GetRuntimeConfig(ctx context.Context, key string) (string, error) {
 	if err := q.authorizeContext(ctx, policy.ActionRead, rbac.ResourceSystem); err != nil {
 		return "", err

coderd/database/dbauthz/dbauthz_test.go

@@ -178,7 +178,8 @@ func TestDBAuthzRecursive(t *testing.T) {
 		if method.Name == "InTx" ||
 			method.Name == "Ping" ||
 			method.Name == "Wrappers" ||
-			method.Name == "PGLocks" {
+			method.Name == "PGLocks" ||
+			method.Name == "GetRunningPrebuiltWorkspacesOptimized" {
 			continue
 		}
 		// easy to know which method failed.

coderd/database/dbauthz/setup_test.go

@@ -41,6 +41,8 @@ var skipMethods = map[string]string{
 	"Wrappers":       "Not relevant",
 	"AcquireLock":    "Not relevant",
 	"TryAcquireLock": "Not relevant",
+	// This method will be removed once we know this works correctly.
+	"GetRunningPrebuiltWorkspacesOptimized": "Not relevant",
 }
 
 // TestMethodTestSuite runs MethodTestSuite.

coderd/database/dbgen/dbgen.go

@@ -359,6 +359,14 @@ func Workspace(t testing.TB, db database.Store, orig database.WorkspaceTable) da
 		NextStartAt:       orig.NextStartAt,
 	})
 	require.NoError(t, err, "insert workspace")
+	if orig.Deleted {
+		err = db.UpdateWorkspaceDeletedByID(genCtx, database.UpdateWorkspaceDeletedByIDParams{
+			ID:      workspace.ID,
+			Deleted: true,
+		})
+		require.NoError(t, err, "set workspace as deleted")
+		workspace.Deleted = true
+	}
 	return workspace
 }
 

coderd/database/dbmetrics/querymetrics.go

@@ -1335,6 +1335,13 @@ func (m queryMetricsStore) GetRunningPrebuiltWorkspaces(ctx context.Context) ([]
 	return r0, r1
 }
 
+func (m queryMetricsStore) GetRunningPrebuiltWorkspacesOptimized(ctx context.Context) ([]database.GetRunningPrebuiltWorkspacesOptimizedRow, error) {
+	start := time.Now()
+	r0, r1 := m.s.GetRunningPrebuiltWorkspacesOptimized(ctx)
+	m.queryLatencies.WithLabelValues("GetRunningPrebuiltWorkspacesOptimized").Observe(time.Since(start).Seconds())
+	return r0, r1
+}
+
 func (m queryMetricsStore) GetRuntimeConfig(ctx context.Context, key string) (string, error) {
 	start := time.Now()
 	r0, r1 := m.s.GetRuntimeConfig(ctx, key)

coderd/database/dbmock/dbmock.go

@@ -2778,6 +2778,21 @@ func (mr *MockStoreMockRecorder) GetRunningPrebuiltWorkspaces(ctx any) *gomock.C
 	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetRunningPrebuiltWorkspaces", reflect.TypeOf((*MockStore)(nil).GetRunningPrebuiltWorkspaces), ctx)
 }
 
+// GetRunningPrebuiltWorkspacesOptimized mocks base method.
+func (m *MockStore) GetRunningPrebuiltWorkspacesOptimized(ctx context.Context) ([]database.GetRunningPrebuiltWorkspacesOptimizedRow, error) {
+	m.ctrl.T.Helper()
+	ret := m.ctrl.Call(m, "GetRunningPrebuiltWorkspacesOptimized", ctx)
+	ret0, _ := ret[0].([]database.GetRunningPrebuiltWorkspacesOptimizedRow)
+	ret1, _ := ret[1].(error)
+	return ret0, ret1
+}
+
+// GetRunningPrebuiltWorkspacesOptimized indicates an expected call of GetRunningPrebuiltWorkspacesOptimized.
+func (mr *MockStoreMockRecorder) GetRunningPrebuiltWorkspacesOptimized(ctx any) *gomock.Call {
+	mr.mock.ctrl.T.Helper()
+	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetRunningPrebuiltWorkspacesOptimized", reflect.TypeOf((*MockStore)(nil).GetRunningPrebuiltWorkspacesOptimized), ctx)
+}
+
 // GetRuntimeConfig mocks base method.
 func (m *MockStore) GetRuntimeConfig(ctx context.Context, key string) (string, error) {
 	m.ctrl.T.Helper()

coderd/database/querier.go

@@ -298,6 +298,7 @@ type sqlcQuerier interface {
 	GetReplicaByID(ctx context.Context, id uuid.UUID) (Replica, error)
 	GetReplicasUpdatedAfter(ctx context.Context, updatedAt time.Time) ([]Replica, error)
 	GetRunningPrebuiltWorkspaces(ctx context.Context) ([]GetRunningPrebuiltWorkspacesRow, error)
+	GetRunningPrebuiltWorkspacesOptimized(ctx context.Context) ([]GetRunningPrebuiltWorkspacesOptimizedRow, error)
 	GetRuntimeConfig(ctx context.Context, key string) (string, error)
 	GetTailnetAgents(ctx context.Context, id uuid.UUID) ([]TailnetAgent, error)
 	GetTailnetClientsForAgent(ctx context.Context, agentID uuid.UUID) ([]TailnetClient, error)

coderd/database/querier_test.go

@@ -5020,3 +5020,102 @@ func requireUsersMatch(t testing.TB, expected []database.User, found []database.
 	t.Helper()
 	require.ElementsMatch(t, expected, database.ConvertUserRows(found), msg)
 }
+
+// TestGetRunningPrebuiltWorkspaces ensures the correct behavior of the
+// GetRunningPrebuiltWorkspaces query.
+func TestGetRunningPrebuiltWorkspaces(t *testing.T) {
+	t.Parallel()
+
+	if !dbtestutil.WillUsePostgres() {
+		t.Skip("Test requires PostgreSQL for complex queries")
+	}
+
+	ctx := testutil.Context(t, testutil.WaitLong)
+	db, _ := dbtestutil.NewDB(t)
+	now := dbtime.Now()
+
+	// Given: a prebuilt workspace with a successful start build and a stop build.
+	org := dbgen.Organization(t, db, database.Organization{})
+	user := dbgen.User(t, db, database.User{})
+	template := dbgen.Template(t, db, database.Template{
+		CreatedBy:      user.ID,
+		OrganizationID: org.ID,
+	})
+	templateVersion := dbgen.TemplateVersion(t, db, database.TemplateVersion{
+		TemplateID:     uuid.NullUUID{UUID: template.ID, Valid: true},
+		OrganizationID: org.ID,
+		CreatedBy:      user.ID,
+	})
+	preset := dbgen.Preset(t, db, database.InsertPresetParams{
+		TemplateVersionID: templateVersion.ID,
+		DesiredInstances:  sql.NullInt32{Int32: 1, Valid: true},
+	})
+
+	setupFixture := func(t *testing.T, db database.Store, name string, deleted bool, transition database.WorkspaceTransition, jobStatus database.ProvisionerJobStatus) database.WorkspaceTable {
+		t.Helper()
+		ws := dbgen.Workspace(t, db, database.WorkspaceTable{
+			OwnerID:    database.PrebuildsSystemUserID,
+			TemplateID: template.ID,
+			Name:       name,
+			Deleted:    deleted,
+		})
+		var canceledAt sql.NullTime
+		var jobError sql.NullString
+		switch jobStatus {
+		case database.ProvisionerJobStatusFailed:
+			jobError = sql.NullString{String: assert.AnError.Error(), Valid: true}
+		case database.ProvisionerJobStatusCanceled:
+			canceledAt = sql.NullTime{Time: now, Valid: true}
+		}
+		pj := dbgen.ProvisionerJob(t, db, nil, database.ProvisionerJob{
+			OrganizationID: org.ID,
+			InitiatorID:    database.PrebuildsSystemUserID,
+			Provisioner:    database.ProvisionerTypeEcho,
+			Type:           database.ProvisionerJobTypeWorkspaceBuild,
+			StartedAt:      sql.NullTime{Time: now.Add(-time.Minute), Valid: true},
+			CanceledAt:     canceledAt,
+			CompletedAt:    sql.NullTime{Time: now, Valid: true},
+			Error:          jobError,
+		})
+		wb := dbgen.WorkspaceBuild(t, db, database.WorkspaceBuild{
+			WorkspaceID:             ws.ID,
+			TemplateVersionID:       templateVersion.ID,
+			TemplateVersionPresetID: uuid.NullUUID{UUID: preset.ID, Valid: true},
+			JobID:                   pj.ID,
+			BuildNumber:             1,
+			Transition:              transition,
+			InitiatorID:             database.PrebuildsSystemUserID,
+			Reason:                  database.BuildReasonInitiator,
+		})
+		// Ensure things are set up as expectd
+		require.Equal(t, transition, wb.Transition)
+		require.Equal(t, int32(1), wb.BuildNumber)
+		require.Equal(t, jobStatus, pj.JobStatus)
+		require.Equal(t, deleted, ws.Deleted)
+
+		return ws
+	}
+
+	// Given: a number of prebuild workspaces with different states exist.
+	runningPrebuild := setupFixture(t, db, "running-prebuild", false, database.WorkspaceTransitionStart, database.ProvisionerJobStatusSucceeded)
+	_ = setupFixture(t, db, "stopped-prebuild", false, database.WorkspaceTransitionStop, database.ProvisionerJobStatusSucceeded)
+	_ = setupFixture(t, db, "failed-prebuild", false, database.WorkspaceTransitionStart, database.ProvisionerJobStatusFailed)
+	_ = setupFixture(t, db, "canceled-prebuild", false, database.WorkspaceTransitionStart, database.ProvisionerJobStatusCanceled)
+	_ = setupFixture(t, db, "deleted-prebuild", true, database.WorkspaceTransitionStart, database.ProvisionerJobStatusSucceeded)
+
+	// Given: a regular workspace also exists.
+	_ = dbfake.WorkspaceBuild(t, db, database.WorkspaceTable{
+		OwnerID:    user.ID,
+		TemplateID: template.ID,
+		Name:       "test-running-regular-workspace",
+		Deleted:    false,
+	})
+
+	// When: we query for running prebuild workspaces
+	runningPrebuilds, err := db.GetRunningPrebuiltWorkspaces(ctx)
+	require.NoError(t, err)
+
+	// Then: only the running prebuild workspace should be returned.
+	require.Len(t, runningPrebuilds, 1, "expected only one running prebuilt workspace")
+	require.Equal(t, runningPrebuild.ID, runningPrebuilds[0].ID, "expected the running prebuilt workspace to be returned")
+}

coderd/database/queries.sql.go

@@ -6843,6 +6843,7 @@ FROM workspace_prebuilds p
 		INNER JOIN workspace_latest_builds b ON b.workspace_id = p.id
 WHERE (b.transition = 'start'::workspace_transition
 	AND b.job_status = 'succeeded'::provisioner_job_status)
+ORDER BY p.id
 `
 
 type GetRunningPrebuiltWorkspacesRow struct {
@@ -6886,6 +6887,106 @@ func (q *sqlQuerier) GetRunningPrebuiltWorkspaces(ctx context.Context) ([]GetRun
 	return items, nil
 }
 
+const getRunningPrebuiltWorkspacesOptimized = `-- name: GetRunningPrebuiltWorkspacesOptimized :many
+WITH latest_prebuilds AS (
+	-- All workspaces that match the following criteria:
+	-- 1. Owned by prebuilds user
+	-- 2. Not deleted
+	-- 3. Latest build is a 'start' transition
+	-- 4. Latest build was successful
+	SELECT
+		workspaces.id,
+		workspaces.name,
+		workspaces.template_id,
+		workspace_latest_builds.template_version_id,
+		workspace_latest_builds.job_id,
+		workspaces.created_at
+	FROM workspace_latest_builds
+	JOIN workspaces ON workspaces.id = workspace_latest_builds.workspace_id
+	WHERE workspace_latest_builds.transition = 'start'::workspace_transition
+	AND workspace_latest_builds.job_status = 'succeeded'::provisioner_job_status
+	AND workspaces.owner_id = 'c42fdf75-3097-471c-8c33-fb52454d81c0'::UUID
+	AND NOT workspaces.deleted
+),
+workspace_latest_presets AS (
+	-- For each of the above workspaces, the preset_id of the most recent
+	-- successful start transition.
+	SELECT DISTINCT ON (latest_prebuilds.id)
+		latest_prebuilds.id AS workspace_id,
+		workspace_builds.template_version_preset_id AS current_preset_id
+	FROM latest_prebuilds
+	JOIN workspace_builds ON workspace_builds.workspace_id = latest_prebuilds.id
+	WHERE workspace_builds.transition = 'start'::workspace_transition
+	AND   workspace_builds.template_version_preset_id IS NOT NULL
+	ORDER BY latest_prebuilds.id, workspace_builds.build_number DESC
+),
+ready_agents AS (
+	-- For each of the above workspaces, check if all agents are ready.
+	SELECT
+		latest_prebuilds.job_id,
+		BOOL_AND(workspace_agents.lifecycle_state = 'ready'::workspace_agent_lifecycle_state)::boolean AS ready
+	FROM latest_prebuilds
+	JOIN workspace_resources ON workspace_resources.job_id = latest_prebuilds.job_id
+	JOIN workspace_agents ON workspace_agents.resource_id = workspace_resources.id
+	WHERE workspace_agents.deleted = false
+	AND workspace_agents.parent_id IS NULL
+	GROUP BY latest_prebuilds.job_id
+)
+SELECT
+	latest_prebuilds.id,
+	latest_prebuilds.name,
+	latest_prebuilds.template_id,
+	latest_prebuilds.template_version_id,
+	workspace_latest_presets.current_preset_id,
+	COALESCE(ready_agents.ready, false)::boolean AS ready,
+	latest_prebuilds.created_at
+FROM latest_prebuilds
+LEFT JOIN ready_agents ON ready_agents.job_id = latest_prebuilds.job_id
+LEFT JOIN workspace_latest_presets ON workspace_latest_presets.workspace_id = latest_prebuilds.id
+ORDER BY latest_prebuilds.id
+`
+
+type GetRunningPrebuiltWorkspacesOptimizedRow struct {
+	ID                uuid.UUID     `db:"id" json:"id"`
+	Name              string        `db:"name" json:"name"`
+	TemplateID        uuid.UUID     `db:"template_id" json:"template_id"`
+	TemplateVersionID uuid.UUID     `db:"template_version_id" json:"template_version_id"`
+	CurrentPresetID   uuid.NullUUID `db:"current_preset_id" json:"current_preset_id"`
+	Ready             bool          `db:"ready" json:"ready"`
+	CreatedAt         time.Time     `db:"created_at" json:"created_at"`
+}
+
+func (q *sqlQuerier) GetRunningPrebuiltWorkspacesOptimized(ctx context.Context) ([]GetRunningPrebuiltWorkspacesOptimizedRow, error) {
+	rows, err := q.db.QueryContext(ctx, getRunningPrebuiltWorkspacesOptimized)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	var items []GetRunningPrebuiltWorkspacesOptimizedRow
+	for rows.Next() {
+		var i GetRunningPrebuiltWorkspacesOptimizedRow
+		if err := rows.Scan(
+			&i.ID,
+			&i.Name,
+			&i.TemplateID,
+			&i.TemplateVersionID,
+			&i.CurrentPresetID,
+			&i.Ready,
+			&i.CreatedAt,
+		); err != nil {
+			return nil, err
+		}
+		items = append(items, i)
+	}
+	if err := rows.Close(); err != nil {
+		return nil, err
+	}
+	if err := rows.Err(); err != nil {
+		return nil, err
+	}
+	return items, nil
+}
+
 const getTemplatePresetsWithPrebuilds = `-- name: GetTemplatePresetsWithPrebuilds :many
 SELECT
 		t.id                        AS template_id,

coderd/database/queries/prebuilds.sql

@@ -48,6 +48,64 @@ WHERE tvp.desired_instances IS NOT NULL -- Consider only presets that have a pre
   -- AND NOT t.deleted -- We don't exclude deleted templates because there's no constraint in the DB preventing a soft deletion on a template while workspaces are running.
 	AND (t.id = sqlc.narg('template_id')::uuid OR sqlc.narg('template_id') IS NULL);
 
+-- name: GetRunningPrebuiltWorkspacesOptimized :many
+WITH latest_prebuilds AS (
+	-- All workspaces that match the following criteria:
+	-- 1. Owned by prebuilds user
+	-- 2. Not deleted
+	-- 3. Latest build is a 'start' transition
+	-- 4. Latest build was successful
+	SELECT
+		workspaces.id,
+		workspaces.name,
+		workspaces.template_id,
+		workspace_latest_builds.template_version_id,
+		workspace_latest_builds.job_id,
+		workspaces.created_at
+	FROM workspace_latest_builds
+	JOIN workspaces ON workspaces.id = workspace_latest_builds.workspace_id
+	WHERE workspace_latest_builds.transition = 'start'::workspace_transition
+	AND workspace_latest_builds.job_status = 'succeeded'::provisioner_job_status
+	AND workspaces.owner_id = 'c42fdf75-3097-471c-8c33-fb52454d81c0'::UUID
+	AND NOT workspaces.deleted
+),
+workspace_latest_presets AS (
+	-- For each of the above workspaces, the preset_id of the most recent
+	-- successful start transition.
+	SELECT DISTINCT ON (latest_prebuilds.id)
+		latest_prebuilds.id AS workspace_id,
+		workspace_builds.template_version_preset_id AS current_preset_id
+	FROM latest_prebuilds
+	JOIN workspace_builds ON workspace_builds.workspace_id = latest_prebuilds.id
+	WHERE workspace_builds.transition = 'start'::workspace_transition
+	AND   workspace_builds.template_version_preset_id IS NOT NULL
+	ORDER BY latest_prebuilds.id, workspace_builds.build_number DESC
+),
+ready_agents AS (
+	-- For each of the above workspaces, check if all agents are ready.
+	SELECT
+		latest_prebuilds.job_id,
+		BOOL_AND(workspace_agents.lifecycle_state = 'ready'::workspace_agent_lifecycle_state)::boolean AS ready
+	FROM latest_prebuilds
+	JOIN workspace_resources ON workspace_resources.job_id = latest_prebuilds.job_id
+	JOIN workspace_agents ON workspace_agents.resource_id = workspace_resources.id
+	WHERE workspace_agents.deleted = false
+	AND workspace_agents.parent_id IS NULL
+	GROUP BY latest_prebuilds.job_id
+)
+SELECT
+	latest_prebuilds.id,
+	latest_prebuilds.name,
+	latest_prebuilds.template_id,
+	latest_prebuilds.template_version_id,
+	workspace_latest_presets.current_preset_id,
+	COALESCE(ready_agents.ready, false)::boolean AS ready,
+	latest_prebuilds.created_at
+FROM latest_prebuilds
+LEFT JOIN ready_agents ON ready_agents.job_id = latest_prebuilds.job_id
+LEFT JOIN workspace_latest_presets ON workspace_latest_presets.workspace_id = latest_prebuilds.id
+ORDER BY latest_prebuilds.id;
+
 -- name: GetRunningPrebuiltWorkspaces :many
 SELECT
 		p.id,
@@ -60,7 +118,8 @@ SELECT
 FROM workspace_prebuilds p
 		INNER JOIN workspace_latest_builds b ON b.workspace_id = p.id
 WHERE (b.transition = 'start'::workspace_transition
-	AND b.job_status = 'succeeded'::provisioner_job_status);
+	AND b.job_status = 'succeeded'::provisioner_job_status)
+ORDER BY p.id;
 
 -- name: CountInProgressPrebuilds :many
 -- CountInProgressPrebuilds returns the number of in-progress prebuilds, grouped by preset ID and transition.