perf(actions): debounce runner heartbeat writes and throttle task picks (#38281)

3 reductions in the DB load generated by many runners polling
`FetchTask`:

**1. Debounce runner heartbeat writes**
Every poll wrote `last_online`, and every `UpdateTask`/`UpdateLog` wrote
`last_active` — while a runner streams logs that is many writes per
second per runner. These are now persisted only when stale enough to
actually affect the active/offline status (`ShouldPersistLastOnline` /
`ShouldPersistLastActive`), using the existing columns.

**2. Throttle concurrent task picks**
A new in-process semaphore (`MAX_CONCURRENT_TASK_PICKS`) bounds how many
runners run the task-assignment transaction at once, so a fleet polling
together cannot stampede the query. Throttled polls retry on their next
poll without advancing the runner's tasks version.

**3. Paginate the task-pick query**
`CreateTaskForRunner` previously loaded every waiting job in the
runner's scope into memory on each poll (no `LIMIT`). Now it pages
through the waiting backlog oldest-first with `LIMIT`, claiming the
first label-matching job.

---------

Co-authored-by: Zettat123 <zettat123@gmail.com>
This commit is contained in:
bircni
2026-07-07 21:16:20 +02:00
committed by GitHub
parent 97078b96cf
commit 308a6f12ae
10 changed files with 317 additions and 30 deletions
+20
View File
@@ -75,8 +75,28 @@ type ActionRunner struct {
const (
RunnerOfflineTime = time.Minute
RunnerIdleTime = 10 * time.Second
// RunnerHeartbeatInterval is how often last_online is persisted on poll.
// Must stay well below RunnerOfflineTime so runners don't flap to offline.
RunnerHeartbeatInterval = 30 * time.Second
// RunnerActiveInterval is how often last_active is persisted while a runner
// streams task updates and logs. Must stay well below RunnerIdleTime so a
// busy runner keeps showing ACTIVE without a DB write on every RPC.
RunnerActiveInterval = 5 * time.Second
)
// ShouldPersistLastOnline reports whether last_online is stale enough to be
// worth writing back. Avoids a DB write on every runner poll.
func ShouldPersistLastOnline(last timeutil.TimeStamp, now time.Time) bool {
return now.Sub(last.AsTime()) >= RunnerHeartbeatInterval
}
// ShouldPersistLastActive reports whether last_active is stale enough to be
// worth writing back. Avoids a DB write on every UpdateTask/UpdateLog RPC while
// a runner is actively streaming logs.
func ShouldPersistLastActive(last timeutil.TimeStamp, now time.Time) bool {
return now.Sub(last.AsTime()) >= RunnerActiveInterval
}
// BelongsToOwnerName before calling, should guarantee that all attributes are loaded
func (r *ActionRunner) BelongsToOwnerName() string {
if r.RepoID != 0 {