Skip to content

fix: move pubsub publishing out of database transactions to avoid conn exhaustion #17648

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 7 commits into from
May 5, 2025
Merged

fix: move pubsub publishing out of database transactions to avoid conn exhaustion #17648

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
ba2f90a
fix: move pubsub publishing out of database transactions to avoid con…
dannykopping May 1, 2025
74f5dd7
chore: commentary
dannykopping May 2, 2025
f6909ff
chore: polish
dannykopping May 2, 2025
1c84fac
chore: augment reconciler test to validate that broken publish does n…
dannykopping May 2, 2025
5117a5e
Merge branch 'main' of github.com:/coder/coder into dk/notify-outside-tx
dannykopping May 2, 2025
42cc3f5
chore: make lint
dannykopping May 2, 2025
9ddad0b
fix: only break pubsub after dbgen setup
dannykopping May 5, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
chore: augment reconciler test to validate that broken publish does n… 
…ot affect behaviour

Signed-off-by: Danny Kopping <dannykopping@gmail.com>
  • Loading branch information
@dannykopping
dannykopping committed May 2, 2025
commit 1c84fac5616770b4ef5cfee1f4018aba749a56ff
4 changes: 2 additions & 2 deletions coderd/database/dbgen/dbgen.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -643,8 +643,8 @@ func ProvisionerJob(t testing.TB, db database.Store, ps pubsub.Pubsub, orig data
})
require.NoError(t, err, "insert job")
if ps != nil {
err = provisionerjobs.PostJob(ps, job)
require.NoError(t, err, "post job to pubsub")
// Advisory, but not essential since acquirer has a background poller to pick up missed jobs.
_ = provisionerjobs.PostJob(ps, job)
}
if !orig.StartedAt.Time.IsZero() {
job, err = db.AcquireProvisionerJob(genCtx, database.AcquireProvisionerJobParams{
Expand Down
195 changes: 107 additions & 88 deletions enterprise/coderd/prebuilds/reconcile_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,7 @@
"time"

"github.com/prometheus/client_golang/prometheus"
"golang.org/x/xerrors"

"github.com/coder/coder/v2/coderd/database/dbtime"
"github.com/coder/coder/v2/coderd/util/slice"
Expand Down Expand Up @@ -303,107 +304,125 @@
for _, prebuildLatestTransition := range tc.prebuildLatestTransitions {
for _, prebuildJobStatus := range tc.prebuildJobStatuses {
for _, templateDeleted := range tc.templateDeleted {
t.Run(fmt.Sprintf("%s - %s - %s", tc.name, prebuildLatestTransition, prebuildJobStatus), func(t *testing.T) {
t.Parallel()
t.Cleanup(func() {
if t.Failed() {
t.Logf("failed to run test: %s", tc.name)
t.Logf("templateVersionActive: %t", templateVersionActive)
t.Logf("prebuildLatestTransition: %s", prebuildLatestTransition)
t.Logf("prebuildJobStatus: %s", prebuildJobStatus)
for _, useBrokenPubsub := range []bool{true, false} {
t.Run(fmt.Sprintf("%s - %s - %s - pubsub_broken=%v", tc.name, prebuildLatestTransition, prebuildJobStatus, useBrokenPubsub), func(t *testing.T) {
t.Parallel()
t.Cleanup(func() {
if t.Failed() {
t.Logf("failed to run test: %s", tc.name)
t.Logf("templateVersionActive: %t", templateVersionActive)
t.Logf("prebuildLatestTransition: %s", prebuildLatestTransition)
t.Logf("prebuildJobStatus: %s", prebuildJobStatus)
}
})
clock := quartz.NewMock(t)
ctx := testutil.Context(t, testutil.WaitShort)
cfg := codersdk.PrebuildsConfig{}
logger := slogtest.Make(
t, &slogtest.Options{IgnoreErrors: true},
).Leveled(slog.LevelDebug)
db, pubSub := dbtestutil.NewDB(t)
if useBrokenPubsub {
pubSub = &brokenPublisher{Pubsub: pubSub}
}
})
clock := quartz.NewMock(t)
ctx := testutil.Context(t, testutil.WaitShort)
cfg := codersdk.PrebuildsConfig{}
logger := slogtest.Make(
t, &slogtest.Options{IgnoreErrors: true},
).Leveled(slog.LevelDebug)
db, pubSub := dbtestutil.NewDB(t)
controller := prebuilds.NewStoreReconciler(db, pubSub, cfg, logger, quartz.NewMock(t), prometheus.NewRegistry())

ownerID := uuid.New()
dbgen.User(t, db, database.User{
ID: ownerID,
})
org, template := setupTestDBTemplate(t, db, ownerID, templateDeleted)
templateVersionID := setupTestDBTemplateVersion(
ctx,
t,
clock,
db,
pubSub,
org.ID,
ownerID,
template.ID,
)
preset := setupTestDBPreset(
t,
db,
templateVersionID,
1,
uuid.New().String(),
)
prebuild := setupTestDBPrebuild(
t,
clock,
db,
pubSub,
prebuildLatestTransition,
prebuildJobStatus,
org.ID,
preset,
template.ID,
templateVersionID,
)

if !templateVersionActive {
// Create a new template version and mark it as active
// This marks the template version that we care about as inactive
setupTestDBTemplateVersion(ctx, t, clock, db, pubSub, org.ID, ownerID, template.ID)
}

// Run the reconciliation multiple times to ensure idempotency
// 8 was arbitrary, but large enough to reasonably trust the result
for i := 1; i <= 8; i++ {
require.NoErrorf(t, controller.ReconcileAll(ctx), "failed on iteration %d", i)

if tc.shouldCreateNewPrebuild != nil {
newPrebuildCount := 0
workspaces, err := db.GetWorkspacesByTemplateID(ctx, template.ID)
require.NoError(t, err)
for _, workspace := range workspaces {
if workspace.ID != prebuild.ID {
newPrebuildCount++

controller := prebuilds.NewStoreReconciler(db, pubSub, cfg, logger, quartz.NewMock(t), prometheus.NewRegistry())

ownerID := uuid.New()
dbgen.User(t, db, database.User{
ID: ownerID,
})
org, template := setupTestDBTemplate(t, db, ownerID, templateDeleted)
templateVersionID := setupTestDBTemplateVersion(
ctx,
t,
clock,
db,
pubSub,
org.ID,
ownerID,
template.ID,
)
preset := setupTestDBPreset(
t,
db,
templateVersionID,
1,
uuid.New().String(),
)
prebuild := setupTestDBPrebuild(
t,
clock,
db,
pubSub,
prebuildLatestTransition,
prebuildJobStatus,
org.ID,
preset,
template.ID,
templateVersionID,
)

if !templateVersionActive {
// Create a new template version and mark it as active
// This marks the template version that we care about as inactive
setupTestDBTemplateVersion(ctx, t, clock, db, pubSub, org.ID, ownerID, template.ID)
}

// Run the reconciliation multiple times to ensure idempotency
// 8 was arbitrary, but large enough to reasonably trust the result
for i := 1; i <= 8; i++ {
require.NoErrorf(t, controller.ReconcileAll(ctx), "failed on iteration %d", i)

if tc.shouldCreateNewPrebuild != nil {
newPrebuildCount := 0
workspaces, err := db.GetWorkspacesByTemplateID(ctx, template.ID)
require.NoError(t, err)
for _, workspace := range workspaces {
if workspace.ID != prebuild.ID {
newPrebuildCount++
}
}
// This test configures a preset that desires one prebuild.
// In cases where new prebuilds should be created, there should be exactly one.
require.Equal(t, *tc.shouldCreateNewPrebuild, newPrebuildCount == 1)
}
// This test configures a preset that desires one prebuild.
// In cases where new prebuilds should be created, there should be exactly one.
require.Equal(t, *tc.shouldCreateNewPrebuild, newPrebuildCount == 1)
}

if tc.shouldDeleteOldPrebuild != nil {
builds, err := db.GetWorkspaceBuildsByWorkspaceID(ctx, database.GetWorkspaceBuildsByWorkspaceIDParams{
WorkspaceID: prebuild.ID,
})
require.NoError(t, err)
if *tc.shouldDeleteOldPrebuild {
require.Equal(t, 2, len(builds))
require.Equal(t, database.WorkspaceTransitionDelete, builds[0].Transition)
} else {
require.Equal(t, 1, len(builds))
require.Equal(t, prebuildLatestTransition, builds[0].Transition)
if tc.shouldDeleteOldPrebuild != nil {
builds, err := db.GetWorkspaceBuildsByWorkspaceID(ctx, database.GetWorkspaceBuildsByWorkspaceIDParams{
WorkspaceID: prebuild.ID,
})
require.NoError(t, err)
if *tc.shouldDeleteOldPrebuild {
require.Equal(t, 2, len(builds))
require.Equal(t, database.WorkspaceTransitionDelete, builds[0].Transition)
} else {
require.Equal(t, 1, len(builds))
require.Equal(t, prebuildLatestTransition, builds[0].Transition)
}
}
}
}
})
})
}
}
}
}
}
}
}

// brokenPublisher is used to validate that Publish() calls which always fail do not affect the reconciler's behaviour,
// since the messages published are not essential but merely advisory.
type brokenPublisher struct {
pubsub.Pubsub
}

func (b *brokenPublisher) Publish(event string, _ []byte) error {

Check failure on line 420 in enterprise/coderd/prebuilds/reconcile_test.go

View workflow job for this annotation

GitHub Actions / lint 

unused-receiver: method receiver 'b' is not referenced in method's body, consider removing or renaming it as _ (revive)
// I'm explicitly _not_ checking for EventJobPosted (coderd/database/provisionerjobs/provisionerjobs.go) since that
// required too much knowledge of the underlying implementation.
return xerrors.Errorf("refusing to publish %q", event)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should we fail slowly instead?

e.g. <-time.After(testutil.IntervalFast)

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Addressed in 9ddad0b

}

func TestMultiplePresetsPerTemplateVersion(t *testing.T) {
t.Parallel()

Expand Down
Loading