0
0
mirror of https://github.com/go-gitea/gitea.git synced 2026-07-09 19:44:19 +02:00
gitea/models/actions/task_test.go
bircni 308a6f12ae
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>
2026-07-07 19:16:20 +00:00

445 lines
13 KiB
Go

// Copyright 2026 The Gitea Authors. All rights reserved.
// SPDX-License-Identifier: MIT
package actions
import (
"strings"
"testing"
runnerv1 "gitea.dev/actions-proto-go/runner/v1"
"gitea.dev/models/db"
"gitea.dev/models/unittest"
"gitea.dev/modules/actions/jobparser"
"gitea.dev/modules/timeutil"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
"google.golang.org/protobuf/types/known/timestamppb"
)
func TestMakeTaskStepDisplayName(t *testing.T) {
tests := []struct {
name string
jobStep *jobparser.Step
expected string
}{
{
name: "explicit name",
jobStep: &jobparser.Step{
Name: "Test Step",
},
expected: "Test Step",
},
{
name: "uses step",
jobStep: &jobparser.Step{
Uses: "actions/checkout@v4",
},
expected: "Run actions/checkout@v4",
},
{
name: "single-line run",
jobStep: &jobparser.Step{
Run: "echo hello",
},
expected: "Run echo hello",
},
{
name: "multi-line run block scalar",
jobStep: &jobparser.Step{
Run: "\n echo hello \r\n echo world \n ",
},
expected: "Run echo hello",
},
{
name: "fallback to id",
jobStep: &jobparser.Step{
ID: "step-id",
},
expected: "Run step-id",
},
{
name: "very long name truncated",
jobStep: &jobparser.Step{
Name: strings.Repeat("a", 300),
},
expected: strings.Repeat("a", 252) + "…",
},
{
name: "very long run truncated",
jobStep: &jobparser.Step{
Run: strings.Repeat("a", 300),
},
expected: "Run " + strings.Repeat("a", 248) + "…",
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
result := makeTaskStepDisplayName(tt.jobStep, 255)
assert.Equal(t, tt.expected, result)
})
}
}
func TestTaskCancellingFinalizesToCancelled(t *testing.T) {
newRunningTask := func(t *testing.T) (*ActionTask, *ActionRunJob) {
t.Helper()
run := &ActionRun{
Title: "cancelling-test-run",
RepoID: 1,
OwnerID: 2,
WorkflowID: "test.yaml",
Index: 999,
TriggerUserID: 2,
Ref: "refs/heads/master",
CommitSHA: "c2d72f548424103f01ee1dc02889c1e2bff816b0",
Event: "push",
TriggerEvent: "push",
Status: StatusRunning,
Started: timeutil.TimeStampNow(),
}
require.NoError(t, db.Insert(t.Context(), run))
job := &ActionRunJob{
RunID: run.ID,
RepoID: run.RepoID,
OwnerID: run.OwnerID,
CommitSHA: run.CommitSHA,
Name: "cancelling-finalization-job",
Attempt: 1,
JobID: "cancelling-finalization-job",
Status: StatusRunning,
}
require.NoError(t, db.Insert(t.Context(), job))
runner := &ActionRunner{
UUID: "runner-cancelling-supported",
Name: "runner-cancelling-supported",
HasCancellingSupport: true,
}
require.NoError(t, db.Insert(t.Context(), runner))
task := &ActionTask{
JobID: job.ID,
Attempt: 1,
RunnerID: runner.ID,
Status: StatusRunning,
Started: timeutil.TimeStampNow(),
RepoID: run.RepoID,
OwnerID: run.OwnerID,
CommitSHA: run.CommitSHA,
}
require.NoError(t, db.Insert(t.Context(), task))
job.TaskID = task.ID
_, err := UpdateRunJob(t.Context(), job, nil, "task_id")
require.NoError(t, err)
return task, job
}
testResult := func(t *testing.T, result runnerv1.Result) {
t.Helper()
require.NoError(t, unittest.PrepareTestDatabase())
task, job := newRunningTask(t)
require.NoError(t, StopTask(t.Context(), task.ID, StatusCancelling))
taskAfterStop := unittest.AssertExistsAndLoadBean(t, &ActionTask{ID: task.ID})
assert.Equal(t, StatusCancelling, taskAfterStop.Status)
updatedTask, err := UpdateTaskByState(t.Context(), task.RunnerID, &runnerv1.TaskState{
Id: task.ID,
Result: result,
StoppedAt: timestamppb.Now(),
})
require.NoError(t, err)
assert.Equal(t, StatusCancelled, updatedTask.Status)
taskAfterUpdate := unittest.AssertExistsAndLoadBean(t, &ActionTask{ID: task.ID})
assert.Equal(t, StatusCancelled, taskAfterUpdate.Status)
jobAfterUpdate := unittest.AssertExistsAndLoadBean(t, &ActionRunJob{ID: job.ID})
assert.Equal(t, StatusCancelled, jobAfterUpdate.Status)
}
t.Run("runner reports success", func(t *testing.T) {
testResult(t, runnerv1.Result_RESULT_SUCCESS)
})
t.Run("runner reports failure", func(t *testing.T) {
testResult(t, runnerv1.Result_RESULT_FAILURE)
})
}
func TestStopTaskCancellingFallsBackForLegacyRunner(t *testing.T) {
require.NoError(t, unittest.PrepareTestDatabase())
run := &ActionRun{
Title: "cancelling-test-run",
RepoID: 1,
OwnerID: 2,
WorkflowID: "test.yaml",
Index: 999,
TriggerUserID: 2,
Ref: "refs/heads/master",
CommitSHA: "c2d72f548424103f01ee1dc02889c1e2bff816b0",
Event: "push",
TriggerEvent: "push",
Status: StatusRunning,
Started: timeutil.TimeStampNow(),
}
require.NoError(t, db.Insert(t.Context(), run))
job := &ActionRunJob{
RunID: run.ID,
RepoID: run.RepoID,
OwnerID: run.OwnerID,
CommitSHA: run.CommitSHA,
Name: "legacy-cancelling-job",
Attempt: 1,
JobID: "legacy-cancelling-job",
Status: StatusRunning,
}
require.NoError(t, db.Insert(t.Context(), job))
runner := &ActionRunner{
UUID: "runner-legacy-no-cancelling",
Name: "runner-legacy-no-cancelling",
HasCancellingSupport: false,
}
require.NoError(t, db.Insert(t.Context(), runner))
task := &ActionTask{
JobID: job.ID,
Attempt: 1,
RunnerID: runner.ID,
Status: StatusRunning,
Started: timeutil.TimeStampNow(),
RepoID: run.RepoID,
OwnerID: run.OwnerID,
CommitSHA: run.CommitSHA,
}
require.NoError(t, db.Insert(t.Context(), task))
job.TaskID = task.ID
_, err := UpdateRunJob(t.Context(), job, nil, "task_id")
require.NoError(t, err)
require.NoError(t, StopTask(t.Context(), task.ID, StatusCancelling))
taskAfterStop := unittest.AssertExistsAndLoadBean(t, &ActionTask{ID: task.ID})
assert.Equal(t, StatusCancelled, taskAfterStop.Status)
assert.NotZero(t, taskAfterStop.Stopped)
jobAfterStop := unittest.AssertExistsAndLoadBean(t, &ActionRunJob{ID: job.ID})
assert.Equal(t, StatusCancelled, jobAfterStop.Status)
assert.NotZero(t, jobAfterStop.Stopped)
}
func TestStopTaskCancellingFallsBackForMissingRunner(t *testing.T) {
require.NoError(t, unittest.PrepareTestDatabase())
run := &ActionRun{
Title: "cancelling-test-run",
RepoID: 1,
OwnerID: 2,
WorkflowID: "test.yaml",
Index: 999,
TriggerUserID: 2,
Ref: "refs/heads/master",
CommitSHA: "c2d72f548424103f01ee1dc02889c1e2bff816b0",
Event: "push",
TriggerEvent: "push",
Status: StatusRunning,
Started: timeutil.TimeStampNow(),
}
require.NoError(t, db.Insert(t.Context(), run))
job := &ActionRunJob{
RunID: run.ID,
RepoID: run.RepoID,
OwnerID: run.OwnerID,
CommitSHA: run.CommitSHA,
Name: "missing-runner-cancelling-job",
Attempt: 1,
JobID: "missing-runner-cancelling-job",
Status: StatusRunning,
}
require.NoError(t, db.Insert(t.Context(), job))
runner := &ActionRunner{
UUID: "runner-cleaned-up-before-cancel",
Name: "runner-cleaned-up-before-cancel",
HasCancellingSupport: true,
}
require.NoError(t, db.Insert(t.Context(), runner))
task := &ActionTask{
JobID: job.ID,
Attempt: 1,
RunnerID: runner.ID,
Status: StatusRunning,
Started: timeutil.TimeStampNow(),
RepoID: run.RepoID,
OwnerID: run.OwnerID,
CommitSHA: run.CommitSHA,
}
require.NoError(t, db.Insert(t.Context(), task))
job.TaskID = task.ID
_, err := UpdateRunJob(t.Context(), job, nil, "task_id")
require.NoError(t, err)
_, err = db.DeleteByID[ActionRunner](t.Context(), runner.ID)
require.NoError(t, err)
require.NoError(t, StopTask(t.Context(), task.ID, StatusCancelling))
taskAfterStop := unittest.AssertExistsAndLoadBean(t, &ActionTask{ID: task.ID})
assert.Equal(t, StatusCancelled, taskAfterStop.Status)
assert.NotZero(t, taskAfterStop.Stopped)
jobAfterStop := unittest.AssertExistsAndLoadBean(t, &ActionRunJob{ID: job.ID})
assert.Equal(t, StatusCancelled, jobAfterStop.Status)
assert.NotZero(t, jobAfterStop.Stopped)
}
// TestReleaseTaskForRunner verifies that releasing a freshly-claimed task returns
// its job to the waiting queue and deletes the task and its steps, so a failure
// while assembling the runner response cannot strand the job in running state.
func TestReleaseTaskForRunner(t *testing.T) {
require.NoError(t, unittest.PrepareTestDatabase())
run := &ActionRun{
Title: "release-task-test-run",
RepoID: 1,
OwnerID: 2,
WorkflowID: "test.yaml",
Index: 9902,
TriggerUserID: 2,
Ref: "refs/heads/main",
CommitSHA: "c2d72f548424103f01ee1dc02889c1e2bff816b0",
Event: "push",
TriggerEvent: "push",
Status: StatusWaiting,
}
require.NoError(t, db.Insert(t.Context(), run))
job := &ActionRunJob{
RunID: run.ID,
RepoID: run.RepoID,
OwnerID: run.OwnerID,
CommitSHA: run.CommitSHA,
Name: "release-job",
Attempt: 1,
JobID: "release-job",
Status: StatusWaiting,
RunsOn: []string{"ubuntu-latest"},
WorkflowPayload: []byte("on: push\njobs:\n release-job:\n runs-on: ubuntu-latest\n steps:\n - run: echo hi\n"),
}
require.NoError(t, db.Insert(t.Context(), job))
runner := &ActionRunner{
UUID: "release-runner-uuid",
Name: "release-runner",
AgentLabels: []string{"ubuntu-latest"},
}
runner.GenerateAndFillToken()
require.NoError(t, db.Insert(t.Context(), runner))
task, ok, err := CreateTaskForRunner(t.Context(), runner)
require.NoError(t, err)
require.True(t, ok)
require.NotNil(t, task)
claimed := unittest.AssertExistsAndLoadBean(t, &ActionRunJob{ID: job.ID})
require.Equal(t, StatusRunning, claimed.Status)
require.Equal(t, task.ID, claimed.TaskID)
require.NoError(t, ReleaseTaskForRunner(t.Context(), task))
// Job is back in the waiting queue with no task assigned.
released := unittest.AssertExistsAndLoadBean(t, &ActionRunJob{ID: job.ID})
assert.Equal(t, StatusWaiting, released.Status)
assert.Zero(t, released.TaskID)
assert.Zero(t, released.Started)
// The task and its steps are gone.
unittest.AssertNotExistsBean(t, &ActionTask{ID: task.ID})
unittest.AssertNotExistsBean(t, &ActionTaskStep{TaskID: task.ID})
}
// TestCreateTaskForRunnerPagination verifies that a job sitting beyond the first page is still claimed
func TestCreateTaskForRunnerPagination(t *testing.T) {
require.NoError(t, unittest.PrepareTestDatabase())
defer func(orig int) { pickTaskBatchSize = orig }(pickTaskBatchSize)
pickTaskBatchSize = 2
run := &ActionRun{
Title: "pagination-test-run",
RepoID: 1,
OwnerID: 2,
WorkflowID: "test.yaml",
Index: 9903,
TriggerUserID: 2,
Ref: "refs/heads/main",
CommitSHA: "c2d72f548424103f01ee1dc02889c1e2bff816b0",
Event: "push",
TriggerEvent: "push",
Status: StatusWaiting,
}
require.NoError(t, db.Insert(t.Context(), run))
// Five waiting jobs the runner cannot run, then one it can.
// With a page size of 2 the matching job only appears on the third page.
for i := range 5 {
mismatch := &ActionRunJob{
RunID: run.ID,
RepoID: run.RepoID,
OwnerID: run.OwnerID,
CommitSHA: run.CommitSHA,
Name: "mismatch-" + string(rune('a'+i)),
Attempt: 1,
JobID: "mismatch-" + string(rune('a'+i)),
Status: StatusWaiting,
RunsOn: []string{"windows-latest"},
}
require.NoError(t, db.Insert(t.Context(), mismatch))
}
target := &ActionRunJob{
RunID: run.ID,
RepoID: run.RepoID,
OwnerID: run.OwnerID,
CommitSHA: run.CommitSHA,
Name: "target-job",
Attempt: 1,
JobID: "target-job",
Status: StatusWaiting,
RunsOn: []string{"ubuntu-latest"},
WorkflowPayload: []byte("on: push\njobs:\n target-job:\n runs-on: ubuntu-latest\n steps:\n - run: echo hi\n"),
}
require.NoError(t, db.Insert(t.Context(), target))
runner := &ActionRunner{
UUID: "pagination-runner-uuid",
Name: "pagination-runner",
AgentLabels: []string{"ubuntu-latest"},
}
runner.GenerateAndFillToken()
require.NoError(t, db.Insert(t.Context(), runner))
task, ok, err := CreateTaskForRunner(t.Context(), runner)
require.NoError(t, err)
require.True(t, ok)
require.NotNil(t, task)
claimed := unittest.AssertExistsAndLoadBean(t, &ActionRunJob{ID: target.ID})
assert.Equal(t, StatusRunning, claimed.Status)
assert.Equal(t, task.ID, claimed.TaskID)
}