feat: delete pending canceled prebuilds (#20499)

## Description

PR https://github.com/coder/coder/pull/20387 introduced canceling
pending prebuild jobs from inactive template versions to avoid
provisioning obsolete workspaces. However, the associated prebuilds
remained in the database with "Canceled" status, visible in the UI.

This PR now orphan-deletes these canceled prebuilt workspaces. Since the
canceled jobs were never processed by a provisioner, no Terraform
resources were created, making orphan deletion safe.

Orphan deletion always creates a provisioner job, but behaves
differently based on provisioner availability:
- If no provisioner daemon is available, the job is immediately marked
as completed and the workspace is marked as deleted without any
provisioner processing
- If a provisioner daemon is available, it processes the delete job with
empty Terraform state (no actual resources to destroy)

The job cancellation and workspace deletion occur atomically in the same
transaction. We don't split this into two separate reconciliation runs
because there's no way to distinguish between system-canceled prebuilds
and user-canceled workspaces. If we deleted canceled workspaces in a
later run, we'd delete user-canceled workspaces that users may want to
keep for troubleshooting.

Note: This only applies to system-generated prebuilds from inactive
template versions.

## Changes

* Update `UpdatePrebuildProvisionerJobWithCancel` query to return job
ID, workspace ID, template ID, and template version preset ID
* Add `DeprovisionMode` enum to support orphan deletion in the provision
flow
* Update `ActionTypeCancelPending` handler to cancel jobs and
orphan-delete associated workspaces atomically
This commit is contained in:
Susana Ferreira
2025-10-29 10:37:28 +00:00
committed by GitHub
parent 0d765f56f7
commit c3e3bb58f2
9 changed files with 226 additions and 102 deletions
+2 -2
View File
@@ -4933,10 +4933,10 @@ func (q *querier) UpdateOrganizationDeletedByID(ctx context.Context, arg databas
return deleteQ(q.log, q.auth, q.db.GetOrganizationByID, deleteF)(ctx, arg.ID)
}
func (q *querier) UpdatePrebuildProvisionerJobWithCancel(ctx context.Context, arg database.UpdatePrebuildProvisionerJobWithCancelParams) ([]uuid.UUID, error) {
func (q *querier) UpdatePrebuildProvisionerJobWithCancel(ctx context.Context, arg database.UpdatePrebuildProvisionerJobWithCancelParams) ([]database.UpdatePrebuildProvisionerJobWithCancelRow, error) {
// Prebuild operation for canceling pending prebuild jobs from non-active template versions
if err := q.authorizeContext(ctx, policy.ActionUpdate, rbac.ResourcePrebuiltWorkspace); err != nil {
return []uuid.UUID{}, err
return []database.UpdatePrebuildProvisionerJobWithCancelRow{}, err
}
return q.db.UpdatePrebuildProvisionerJobWithCancel(ctx, arg)
}
+6 -3
View File
@@ -646,10 +646,13 @@ func (s *MethodTestSuite) TestProvisionerJob() {
PresetID: uuid.NullUUID{UUID: uuid.New(), Valid: true},
Now: dbtime.Now(),
}
jobIDs := []uuid.UUID{uuid.New(), uuid.New()}
canceledJobs := []database.UpdatePrebuildProvisionerJobWithCancelRow{
{ID: uuid.New(), WorkspaceID: uuid.New(), TemplateID: uuid.New(), TemplateVersionPresetID: uuid.NullUUID{UUID: uuid.New(), Valid: true}},
{ID: uuid.New(), WorkspaceID: uuid.New(), TemplateID: uuid.New(), TemplateVersionPresetID: uuid.NullUUID{UUID: uuid.New(), Valid: true}},
}
dbm.EXPECT().UpdatePrebuildProvisionerJobWithCancel(gomock.Any(), arg).Return(jobIDs, nil).AnyTimes()
check.Args(arg).Asserts(rbac.ResourcePrebuiltWorkspace, policy.ActionUpdate).Returns(jobIDs)
dbm.EXPECT().UpdatePrebuildProvisionerJobWithCancel(gomock.Any(), arg).Return(canceledJobs, nil).AnyTimes()
check.Args(arg).Asserts(rbac.ResourcePrebuiltWorkspace, policy.ActionUpdate).Returns(canceledJobs)
}))
s.Run("GetProvisionerJobsByIDs", s.Mocked(func(dbm *dbmock.MockStore, faker *gofakeit.Faker, check *expects) {
org := testutil.Fake(s.T(), faker, database.Organization{})
+1 -1
View File
@@ -3042,7 +3042,7 @@ func (m queryMetricsStore) UpdateOrganizationDeletedByID(ctx context.Context, ar
return r0
}
func (m queryMetricsStore) UpdatePrebuildProvisionerJobWithCancel(ctx context.Context, arg database.UpdatePrebuildProvisionerJobWithCancelParams) ([]uuid.UUID, error) {
func (m queryMetricsStore) UpdatePrebuildProvisionerJobWithCancel(ctx context.Context, arg database.UpdatePrebuildProvisionerJobWithCancelParams) ([]database.UpdatePrebuildProvisionerJobWithCancelRow, error) {
start := time.Now()
r0, r1 := m.s.UpdatePrebuildProvisionerJobWithCancel(ctx, arg)
m.queryLatencies.WithLabelValues("UpdatePrebuildProvisionerJobWithCancel").Observe(time.Since(start).Seconds())
+2 -2
View File
@@ -6540,10 +6540,10 @@ func (mr *MockStoreMockRecorder) UpdateOrganizationDeletedByID(ctx, arg any) *go
}
// UpdatePrebuildProvisionerJobWithCancel mocks base method.
func (m *MockStore) UpdatePrebuildProvisionerJobWithCancel(ctx context.Context, arg database.UpdatePrebuildProvisionerJobWithCancelParams) ([]uuid.UUID, error) {
func (m *MockStore) UpdatePrebuildProvisionerJobWithCancel(ctx context.Context, arg database.UpdatePrebuildProvisionerJobWithCancelParams) ([]database.UpdatePrebuildProvisionerJobWithCancelRow, error) {
m.ctrl.T.Helper()
ret := m.ctrl.Call(m, "UpdatePrebuildProvisionerJobWithCancel", ctx, arg)
ret0, _ := ret[0].([]uuid.UUID)
ret0, _ := ret[0].([]database.UpdatePrebuildProvisionerJobWithCancelRow)
ret1, _ := ret[1].(error)
return ret0, ret1
}
+1 -1
View File
@@ -667,7 +667,7 @@ type sqlcQuerier interface {
// Cancels all pending provisioner jobs for prebuilt workspaces on a specific preset from an
// inactive template version.
// This is an optimization to clean up stale pending jobs.
UpdatePrebuildProvisionerJobWithCancel(ctx context.Context, arg UpdatePrebuildProvisionerJobWithCancelParams) ([]uuid.UUID, error)
UpdatePrebuildProvisionerJobWithCancel(ctx context.Context, arg UpdatePrebuildProvisionerJobWithCancelParams) ([]UpdatePrebuildProvisionerJobWithCancelRow, error)
UpdatePresetPrebuildStatus(ctx context.Context, arg UpdatePresetPrebuildStatusParams) error
UpdateProvisionerDaemonLastSeenAt(ctx context.Context, arg UpdateProvisionerDaemonLastSeenAtParams) error
UpdateProvisionerJobByID(ctx context.Context, arg UpdateProvisionerJobByIDParams) error
+26 -12
View File
@@ -8687,12 +8687,8 @@ func (q *sqlQuerier) GetTemplatePresetsWithPrebuilds(ctx context.Context, templa
}
const updatePrebuildProvisionerJobWithCancel = `-- name: UpdatePrebuildProvisionerJobWithCancel :many
UPDATE provisioner_jobs
SET
canceled_at = $1::timestamptz,
completed_at = $1::timestamptz
WHERE id IN (
SELECT pj.id
WITH jobs_to_cancel AS (
SELECT pj.id, w.id AS workspace_id, w.template_id, wpb.template_version_preset_id
FROM provisioner_jobs pj
INNER JOIN workspace_prebuild_builds wpb ON wpb.job_id = pj.id
INNER JOIN workspaces w ON w.id = wpb.workspace_id
@@ -8711,7 +8707,13 @@ WHERE id IN (
AND pj.canceled_at IS NULL
AND pj.completed_at IS NULL
)
RETURNING id
UPDATE provisioner_jobs
SET
canceled_at = $1::timestamptz,
completed_at = $1::timestamptz
FROM jobs_to_cancel
WHERE provisioner_jobs.id = jobs_to_cancel.id
RETURNING jobs_to_cancel.id, jobs_to_cancel.workspace_id, jobs_to_cancel.template_id, jobs_to_cancel.template_version_preset_id
`
type UpdatePrebuildProvisionerJobWithCancelParams struct {
@@ -8719,22 +8721,34 @@ type UpdatePrebuildProvisionerJobWithCancelParams struct {
PresetID uuid.NullUUID `db:"preset_id" json:"preset_id"`
}
type UpdatePrebuildProvisionerJobWithCancelRow struct {
ID uuid.UUID `db:"id" json:"id"`
WorkspaceID uuid.UUID `db:"workspace_id" json:"workspace_id"`
TemplateID uuid.UUID `db:"template_id" json:"template_id"`
TemplateVersionPresetID uuid.NullUUID `db:"template_version_preset_id" json:"template_version_preset_id"`
}
// Cancels all pending provisioner jobs for prebuilt workspaces on a specific preset from an
// inactive template version.
// This is an optimization to clean up stale pending jobs.
func (q *sqlQuerier) UpdatePrebuildProvisionerJobWithCancel(ctx context.Context, arg UpdatePrebuildProvisionerJobWithCancelParams) ([]uuid.UUID, error) {
func (q *sqlQuerier) UpdatePrebuildProvisionerJobWithCancel(ctx context.Context, arg UpdatePrebuildProvisionerJobWithCancelParams) ([]UpdatePrebuildProvisionerJobWithCancelRow, error) {
rows, err := q.db.QueryContext(ctx, updatePrebuildProvisionerJobWithCancel, arg.Now, arg.PresetID)
if err != nil {
return nil, err
}
defer rows.Close()
var items []uuid.UUID
var items []UpdatePrebuildProvisionerJobWithCancelRow
for rows.Next() {
var id uuid.UUID
if err := rows.Scan(&id); err != nil {
var i UpdatePrebuildProvisionerJobWithCancelRow
if err := rows.Scan(
&i.ID,
&i.WorkspaceID,
&i.TemplateID,
&i.TemplateVersionPresetID,
); err != nil {
return nil, err
}
items = append(items, id)
items = append(items, i)
}
if err := rows.Close(); err != nil {
return nil, err
+9 -7
View File
@@ -300,12 +300,8 @@ GROUP BY wpb.template_version_preset_id;
-- Cancels all pending provisioner jobs for prebuilt workspaces on a specific preset from an
-- inactive template version.
-- This is an optimization to clean up stale pending jobs.
UPDATE provisioner_jobs
SET
canceled_at = @now::timestamptz,
completed_at = @now::timestamptz
WHERE id IN (
SELECT pj.id
WITH jobs_to_cancel AS (
SELECT pj.id, w.id AS workspace_id, w.template_id, wpb.template_version_preset_id
FROM provisioner_jobs pj
INNER JOIN workspace_prebuild_builds wpb ON wpb.job_id = pj.id
INNER JOIN workspaces w ON w.id = wpb.workspace_id
@@ -324,4 +320,10 @@ WHERE id IN (
AND pj.canceled_at IS NULL
AND pj.completed_at IS NULL
)
RETURNING id;
UPDATE provisioner_jobs
SET
canceled_at = @now::timestamptz,
completed_at = @now::timestamptz
FROM jobs_to_cancel
WHERE provisioner_jobs.id = jobs_to_cancel.id
RETURNING jobs_to_cancel.id, jobs_to_cancel.workspace_id, jobs_to_cancel.template_id, jobs_to_cancel.template_version_preset_id;