Skip to content

chore: acquire lock for individual workspace transition #15859

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 5 commits into from
Dec 13, 2024
Merged

chore: acquire lock for individual workspace transition #15859

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
1213769
chore: attempt to fix race
DanielleMaywood Dec 10, 2024
c45817b
chore: use `log` instead of `e.log`
DanielleMaywood Dec 13, 2024
46fa216
chore: add check for err != context.Canceled
DanielleMaywood Dec 13, 2024
ae7fca9
chore: use Cian's suggestion
DanielleMaywood Dec 13, 2024
6b133e7
chore: ignore last commit, misread the comment
DanielleMaywood Dec 13, 2024
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
12 changes: 11 additions & 1 deletion coderd/autobuild/lifecycle_executor.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@ package autobuild
import (
"context"
"database/sql"
"fmt"
"net/http"
"sync"
"sync/atomic"
Expand Down Expand Up @@ -177,6 +178,15 @@ func (e *Executor) runOnce(t time.Time) Stats {
err := e.db.InTx(func(tx database.Store) error {
var err error

ok, err := tx.TryAcquireLock(e.ctx, database.GenLockID(fmt.Sprintf("lifecycle-executor:%s", wsID)))
if err != nil {
return xerrors.Errorf("try acquire lifecycle executor lock: %w", err)
}
if !ok {
log.Debug(e.ctx, "unable to acquire lock for workspace, skipping")
return nil
}

// Re-check eligibility since the first check was outside the
// transaction and the workspace settings may have changed.
ws, err = tx.GetWorkspaceByID(e.ctx, wsID)
Expand Down Expand Up @@ -389,7 +399,7 @@ func (e *Executor) runOnce(t time.Time) Stats {
}
return nil
}()
if err != nil {
if err != nil && !xerrors.Is(err, context.Canceled) {
log.Error(e.ctx, "failed to transition workspace", slog.Error(err))
statsMu.Lock()
stats.Errors[wsID] = err
Expand Down
71 changes: 71 additions & 0 deletions coderd/autobuild/lifecycle_executor_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,7 @@ import (
"github.com/coder/coder/v2/coderd/coderdtest"
"github.com/coder/coder/v2/coderd/database"
"github.com/coder/coder/v2/coderd/database/dbauthz"
"github.com/coder/coder/v2/coderd/database/dbtestutil"
"github.com/coder/coder/v2/coderd/notifications"
"github.com/coder/coder/v2/coderd/notifications/notificationstest"
"github.com/coder/coder/v2/coderd/schedule"
Expand Down Expand Up @@ -72,6 +73,76 @@ func TestExecutorAutostartOK(t *testing.T) {
require.Equal(t, template.AutostartRequirement.DaysOfWeek, []string{"monday", "tuesday", "wednesday", "thursday", "friday", "saturday", "sunday"})
}

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

db, ps := dbtestutil.NewDB(t)

var (
sched = mustSchedule(t, "CRON_TZ=UTC 0 * * * *")
// Create our first client
tickCh = make(chan time.Time, 2)
statsChA = make(chan autobuild.Stats)
clientA = coderdtest.New(t, &coderdtest.Options{
IncludeProvisionerDaemon: true,
AutobuildTicker: tickCh,
AutobuildStats: statsChA,
Database: db,
Pubsub: ps,
})
// ... And then our second client
statsChB = make(chan autobuild.Stats)
_ = coderdtest.New(t, &coderdtest.Options{
IncludeProvisionerDaemon: true,
AutobuildTicker: tickCh,
AutobuildStats: statsChB,
Database: db,
Pubsub: ps,
})
// Now create a workspace (we can use either client, it doesn't matter)
workspace = mustProvisionWorkspace(t, clientA, func(cwr *codersdk.CreateWorkspaceRequest) {
cwr.AutostartSchedule = ptr.Ref(sched.String())
})
)

// Have the workspace stopped so we can perform an autostart
workspace = coderdtest.MustTransitionWorkspace(t, clientA, workspace.ID, database.WorkspaceTransitionStart, database.WorkspaceTransitionStop)

// Get both clients to perform a lifecycle execution tick
next := sched.Next(workspace.LatestBuild.CreatedAt)

startCh := make(chan struct{})
go func() {
<-startCh
tickCh <- next
}()
go func() {
<-startCh
tickCh <- next
}()
Copy link
Member

Choose a reason for hiding this comment

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

Is this potentially racy? We're testing that the lock acquire works but theoretically that might not happen if the first coderd grabs the job, completes it, and then the second one does.

I doubt it matters as I suppose we're happy even if the try acquire is hit only a faction of the time, but thought I'd flag it anyway.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Looking again, you're probably right. I ran the test with verbose logging and it looks like this all occurs within 0.05s.

If the test doesn't hit the lock, then we are likely to hit a flake. I'll have a go at increasing this time buffer.

Copy link
Member

@johnstcn johnstcn Dec 13, 2024 •
edited
Loading

Choose a reason for hiding this comment

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

I think you might be able to reduce (but not eliminate) racyness by having a second chan struct{} that you then close after starting both goroutines, making them both wait until it's closed to start.

e.g.

    startCh := make(chan struct{})
	go func() {
        <-startCh
		tickChA <- next
		close(tickChA)
	}()
	go func() {
        <-startCh
		tickChB <- next
		close(tickChB)
	}()
    close(startCh)

You might also be able to get both of them to tick very closely in time by sharing the same tick channel, and making it buffered with size 2. (Of course then you'd want to avoid closing the channel twice to avoid a panic)

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I've gone with your proposal @johnstcn.

It looks like for testing we just use an echo provisioner job, so getting that to take artificially longer for this specific test may not be a trivial task.

close(startCh)

// Now we want to check the stats for both clients
statsA := <-statsChA
statsB := <-statsChB

// We expect there to be no errors
assert.Len(t, statsA.Errors, 0)
assert.Len(t, statsB.Errors, 0)

// We also expect there to have been only one transition
require.Equal(t, 1, len(statsA.Transitions)+len(statsB.Transitions))

stats := statsA
if len(statsB.Transitions) == 1 {
stats = statsB
}

// And we expect this transition to have been a start transition
assert.Contains(t, stats.Transitions, workspace.ID)
assert.Equal(t, database.WorkspaceTransitionStart, stats.Transitions[workspace.ID])
}

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

Expand Down
Loading