Compare commits

..
Author SHA1 Message Date
panxiao81 53a080b310 ci: 为测试显式选择 Pod runner
test / shell (pull_request) Successful in 1m34s
test / python (pull_request) Successful in 1m13s
test / go (pull_request) Successful in 3m14s
2026-09-21 09:34:30 +00:00
panxiao81 38e8d59541 feat: 增加 Runner 生命周期结构化日志
test / go (pull_request) Successful in 3m10s
test / shell (pull_request) Failing after 10m4s
test / python (pull_request) Failing after 10m4s
2026-09-21 09:31:46 +00:00
panxiao81 661b5e9218 Merge vm-dev 集成测试隔离标签
publish images / publish-images (push) Failing after 11m12s
test / python (push) Successful in 24s
test / shell (push) Successful in 28s
test / go (push) Successful in 6m44s
2026-09-21 08:51:17 +00:00
panxiao81 48b6b8038e feat: 隔离 VM 集成测试标签
test / python (pull_request) Failing after 11m44s
test / shell (pull_request) Successful in 21s
test / go (pull_request) Successful in 5m12s
2026-09-21 08:50:18 +00:00
panxiao81 70c5ff422f Merge assignment 重投预算修复
publish images / publish-images (push) Failing after 12m48s
test / python (push) Successful in 31s
test / shell (push) Successful in 34s
test / go (push) Successful in 5m40s
2026-09-21 08:44:42 +00:00
panxiao81 906e6a2e18 fix: 保持后端 assignment 可重试
test / shell (pull_request) Failing after 13m28s
test / python (pull_request) Failing after 13m29s
test / go (pull_request) Successful in 6m15s
2026-09-21 08:44:00 +00:00
panxiao81 ace84373f6 Merge OpenSandbox Runner workload 标记
test / python (push) Successful in 25s
test / shell (push) Successful in 30s
test / go (push) Successful in 4m30s
publish images / publish-images (push) Failing after 13m58s
2026-09-21 08:28:32 +00:00
panxiao81 87cf359ef6 fix: 标记 OpenSandbox Runner workload
test / shell (pull_request) Successful in 26s
test / go (pull_request) Successful in 2m58s
test / python (pull_request) Failing after 14m15s
2026-09-21 08:22:56 +00:00
panxiao81 ad3934e7e1 Merge OpenSandbox metadata 编码修复
test / python (push) Successful in 21s
test / shell (push) Successful in 24s
test / go (push) Successful in 6m52s
publish images / publish-images (push) Failing after 10m40s
2026-09-21 07:56:50 +00:00
panxiao81 d82687382a fix: 编码 OpenSandbox assignment metadata
test / go (pull_request) Successful in 4m45s
test / shell (pull_request) Failing after 13m27s
test / python (pull_request) Failing after 13m27s
2026-09-21 07:48:10 +00:00
panxiao81 7f52cf393f Merge Pod 与 VM 独立执行容量池
publish images / publish-images (push) Failing after 11m8s
test / python (push) Successful in 10s
test / shell (push) Successful in 26s
test / go (push) Successful in 5m51s
2026-09-21 07:26:10 +00:00
panxiao81 8a186dcd86 Merge Runner 无中断恢复与容量隔离
test / python (push) Successful in 25s
test / shell (push) Successful in 29s
test / go (push) Successful in 2m28s
publish images / publish-images (push) Failing after 11m13s
2026-09-21 06:48:11 +00:00
panxiao81 327c73e744 Merge Go Runner 与 OpenSandbox VM 集成
test / python (push) Successful in 15s
test / shell (push) Successful in 21s
publish images / publish-images (push) Failing after 13m58s
test / go (push) Failing after 15m46s
2026-09-21 06:11:40 +00:00
14 changed files with 220 additions and 26 deletions
+2 -2
View File
@@ -17,7 +17,7 @@ jobs:
- run: go vet ./...
python:
runs-on: self-hosted
runs-on: [self-hosted, pod]
steps:
- uses: actions/checkout@v4
- uses: actions/setup-python@v5
@@ -28,7 +28,7 @@ jobs:
- run: python -m compileall -q src
shell:
runs-on: self-hosted
runs-on: [self-hosted, pod]
steps:
- uses: actions/checkout@v4
- run: |
+1 -1
View File
@@ -5,7 +5,7 @@ on:
jobs:
kind:
runs-on: [self-hosted, vm]
runs-on: [self-hosted, vm-dev]
steps:
- name: Verify Docker
run: docker info
+4
View File
@@ -15,6 +15,10 @@ runs-on: [self-hosted, vm]
只执行一个 job,并在 job 结束后连同本地状态一起销毁。完整的设计约束见
[`docs/design-principles.md`](docs/design-principles.md)。
集成期间可将 `VM_RUNNER_LABEL=vm-dev`,只接取显式使用
`runs-on: [self-hosted, vm-dev]` 的测试任务;生产 `vm` job 将保持在 Gitea pending,
不会在 backend 修复过程中继续涌入。
目标 Go controller 组件:
- `scheduler`:以常驻 Gitea RunnerService 身份直接领取 task,并把完整 assignment
+28 -11
View File
@@ -4,7 +4,7 @@ import (
"context"
"errors"
"fmt"
"log"
"log/slog"
"net/http"
"os"
"slices"
@@ -49,7 +49,7 @@ type controllerConfig struct {
PodNamespace, PodImage, PodServiceAccount, SPIRECluster, SPIREClass string
SPIREAgentID string
PodExecutorUID, PodCapacity int
OpenSandboxURL, OpenSandboxAPIKey, OpenSandboxPool string
OpenSandboxURL, OpenSandboxAPIKey, OpenSandboxPool, VMRunnerLabel string
VMTimeout, VMCapacity int
}
@@ -131,7 +131,7 @@ func runController(ctx context.Context) error {
labels = append(labels, string(taskassignment.BackendPod))
}
if slices.Contains(config.Components, controller.VMWorker) {
labels = append(labels, string(taskassignment.BackendVM))
labels = append(labels, config.VMRunnerLabel)
}
poller := taskscheduler.Poller{
Client: giteaClient,
@@ -139,7 +139,7 @@ func runController(ctx context.Context) error {
JetStream: producerJS, SubjectBase: config.SubjectBase,
}},
Config: taskscheduler.PollerConfig{Version: "gitea-dynamic-runner/0.4", Labels: labels, Capacity: config.PodCapacity + config.VMCapacity},
OnError: func(err error) { log.Printf("scheduler: %v", err) },
OnError: func(err error) { slog.Error("scheduler error", "component", "scheduler", "error", err) },
}
kubernetesConfig, err := rest.InClusterConfig()
if err != nil {
@@ -190,11 +190,13 @@ func runController(ctx context.Context) error {
return fmt.Errorf("recover Pod facade claim %s: %w", assignment.ID, err)
}
}
component, err := workerComponent(ctx, workerJS, config, taskassignment.BackendPod, config.PodCapacity, taskworker.Worker{Backend: backend, Bootstrap: bootstrap}, registry, podPool)
component, err := workerComponent(ctx, workerJS, config, taskassignment.BackendPod, config.PodCapacity, taskworker.Worker{Backend: backend, Bootstrap: bootstrap, OnEvent: workerEventLogger(taskassignment.BackendPod)}, registry, podPool)
if err != nil {
return err
}
lifecycle := podbackend.Lifecycle{Backend: backend, OnError: func(err error) { log.Printf("pod lifecycle: %v", err) }}
lifecycle := podbackend.Lifecycle{Backend: backend, OnError: func(err error) {
slog.Error("backend lifecycle error", "component", "lifecycle", "backend", taskassignment.BackendPod, "error", err)
}}
components[controller.PodWorker] = runComponent(func(ctx context.Context) error {
group, groupContext := errgroup.WithContext(ctx)
group.Go(func() error { return component.Run(groupContext) })
@@ -220,11 +222,13 @@ func runController(ctx context.Context) error {
return fmt.Errorf("recover VM facade claim %s: %w", assignment.ID, err)
}
}
component, err := workerComponent(ctx, workerJS, config, taskassignment.BackendVM, config.VMCapacity, taskworker.Worker{Backend: backend, Bootstrap: bootstrap}, registry, vmPool)
component, err := workerComponent(ctx, workerJS, config, taskassignment.BackendVM, config.VMCapacity, taskworker.Worker{Backend: backend, Bootstrap: bootstrap, OnEvent: workerEventLogger(taskassignment.BackendVM)}, registry, vmPool)
if err != nil {
return err
}
lifecycleReconciler := opensandboxbackend.LifecycleReconciler{Backend: backend, OnError: func(err error) { log.Printf("VM lifecycle: %v", err) }}
lifecycleReconciler := opensandboxbackend.LifecycleReconciler{Backend: backend, OnError: func(err error) {
slog.Error("backend lifecycle error", "component", "lifecycle", "backend", taskassignment.BackendVM, "error", err)
}}
components[controller.VMWorker] = runComponent(func(ctx context.Context) error {
group, groupContext := errgroup.WithContext(ctx)
group.Go(func() error { return component.Run(groupContext) })
@@ -294,11 +298,21 @@ func workerComponent(ctx context.Context, js jetstream.JetStream, config control
}
return assignmentqueue.ConsumerComponent{
Consumer: consumer, Capacity: capacity,
Processor: assignmentqueue.Processor{TrustDomain: config.TrustDomain, Accepter: accepter, Claims: claims, Admission: admission},
OnError: func(err error) { log.Printf("%s worker: %v", backend, err) },
Processor: assignmentqueue.Processor{TrustDomain: config.TrustDomain, Accepter: accepter, Claims: claims, Admission: admission, OnEvent: func(event assignmentqueue.Event) {
slog.Info("assignment transition", "component", "worker", "event", event.Name, "backend", event.Backend, "assignment", event.AssignmentID, "retry_delay", event.RetryDelay)
}},
OnError: func(err error) {
slog.Error("assignment processing error", "component", "worker", "backend", backend, "error", err)
},
}, nil
}
func workerEventLogger(backend taskassignment.Backend) func(taskworker.Event) {
return func(event taskworker.Event) {
slog.Info("executor transition", "component", "worker", "event", event.Name, "backend", backend, "assignment", event.AssignmentID, "executor", event.Executor, "phase", event.Phase)
}
}
func loadControllerConfig() (controllerConfig, error) {
selection, err := controller.ParseSelection(os.Getenv("COMPONENTS"))
if err != nil {
@@ -345,7 +359,7 @@ func loadControllerConfig() (controllerConfig, error) {
FacadeListen: env("RUNNER_FACADE_LISTEN", ":8443"), FacadeURL: os.Getenv("RUNNER_FACADE_URL"), FacadeSPIFFEID: os.Getenv("RUNNER_FACADE_SPIFFE_ID"), CapabilityKey: []byte(capabilityKey),
PodNamespace: env("POD_NAMESPACE", "gitea-actions"), PodImage: os.Getenv("POD_EXECUTOR_IMAGE"), PodServiceAccount: env("POD_SERVICE_ACCOUNT", "gitea-task-executor"),
SPIRECluster: env("SPIRE_CLUSTER", "homelab"), SPIREClass: env("SPIRE_CLASS", "spire-mgmt-spire"), SPIREAgentID: os.Getenv("SPIRE_AGENT_ID"), PodExecutorUID: envInt("POD_EXECUTOR_UID", 2000), PodCapacity: envInt("POD_CAPACITY", 4),
OpenSandboxURL: os.Getenv("OPENSANDBOX_API"), OpenSandboxPool: env("OPENSANDBOX_POOL", "ci-vm"), VMTimeout: envInt("VM_TIMEOUT_SECONDS", 14400), VMCapacity: envInt("VM_CAPACITY", 1),
OpenSandboxURL: os.Getenv("OPENSANDBOX_API"), OpenSandboxPool: env("OPENSANDBOX_POOL", "ci-vm"), VMRunnerLabel: env("VM_RUNNER_LABEL", "vm"), VMTimeout: envInt("VM_TIMEOUT_SECONDS", 14400), VMCapacity: envInt("VM_CAPACITY", 1),
}
if config.WorkloadAPIAddr == "" || config.FacadeURL == "" || config.FacadeSPIFFEID == "" {
return controllerConfig{}, errors.New("SPIFFE_ENDPOINT_SOCKET, RUNNER_FACADE_URL, and RUNNER_FACADE_SPIFFE_ID are required")
@@ -357,6 +371,9 @@ func loadControllerConfig() (controllerConfig, error) {
return controllerConfig{}, errors.New("SPIRE_AGENT_ID is required for pod-worker")
}
if slices.Contains(selection, controller.VMWorker) {
if config.VMRunnerLabel != "vm" && config.VMRunnerLabel != "vm-dev" {
return controllerConfig{}, errors.New("VM_RUNNER_LABEL must be vm or vm-dev")
}
if config.OpenSandboxURL == "" {
return controllerConfig{}, errors.New("OPENSANDBOX_API is required for vm-worker")
}
@@ -56,6 +56,7 @@ func TestLoadControllerConfigRequiresOpenSandboxSecretOnlyForVM(t *testing.T) {
t.Setenv("RUNNER_FACADE_URL", "https://facade:8443")
t.Setenv("RUNNER_FACADE_SPIFFE_ID", "spiffe://ddupan.top/controller")
t.Setenv("OPENSANDBOX_API", "http://opensandbox.internal")
t.Setenv("VM_RUNNER_LABEL", "vm-dev")
if _, err := loadControllerConfig(); err == nil {
t.Fatal("expected missing OpenSandbox API key file error")
}
+3 -1
View File
@@ -4,6 +4,7 @@ import (
"context"
"errors"
"fmt"
"log/slog"
"os"
"os/signal"
"syscall"
@@ -12,8 +13,9 @@ import (
)
func main() {
slog.SetDefault(slog.New(slog.NewJSONHandler(os.Stderr, nil)))
if err := run(); err != nil {
fmt.Fprintln(os.Stderr, err)
slog.Error("runner stopped", "error", err)
os.Exit(1)
}
}
+26 -1
View File
@@ -78,6 +78,22 @@ type Processor struct {
Admission Admission
RetryDelay time.Duration
ClaimTimeout time.Duration
OnEvent func(Event)
}
// Event describes a non-sensitive assignment handoff transition. It never
// contains task payloads, credentials, capabilities, or workload identities.
type Event struct {
Name string
AssignmentID string
Backend taskassignment.Backend
RetryDelay time.Duration
}
func (p Processor) event(name string, assignment taskassignment.Assignment, retryDelay time.Duration) {
if p.OnEvent != nil {
p.OnEvent(Event{Name: name, AssignmentID: assignment.ID, Backend: assignment.Backend, RetryDelay: retryDelay})
}
}
func (p Processor) Process(ctx context.Context, message Message) error {
@@ -88,6 +104,7 @@ func (p Processor) Process(ctx context.Context, message Message) error {
if err != nil {
return errors.Join(err, message.TermWithReason("invalid assignment"))
}
p.event("received", assignment, 0)
if _, err := p.Claims.Offer(assignment); err != nil {
return errors.Join(err, message.TermWithReason("conflicting assignment"))
}
@@ -96,8 +113,10 @@ func (p Processor) Process(ctx context.Context, message Message) error {
if delay <= 0 {
delay = 2 * time.Second
}
p.event("capacity_wait", assignment, delay)
return message.NakWithDelay(delay)
}
p.event("capacity_acquired", assignment, 0)
accepted, err := p.Accepter.Accept(ctx, assignment)
if err != nil {
p.Admission.Release(assignment.ID)
@@ -105,9 +124,11 @@ func (p Processor) Process(ctx context.Context, message Message) error {
if delay <= 0 {
delay = 15 * time.Second
}
p.event("backend_retry", assignment, delay)
return errors.Join(err, message.NakWithDelay(delay))
}
if accepted {
p.event("backend_ready", assignment, 0)
timeout := p.ClaimTimeout
if timeout <= 0 {
timeout = 4 * time.Minute
@@ -120,17 +141,21 @@ func (p Processor) Process(ctx context.Context, message Message) error {
if delay <= 0 {
delay = 2 * time.Second
}
p.event("claim_timeout", assignment, delay)
return errors.Join(err, message.NakWithDelay(delay))
}
p.event("runner_claimed", assignment, 0)
if err := message.DoubleAck(ctx); err != nil {
return fmt.Errorf("ack assignment %s: %w", assignment.ID, err)
}
p.event("acked", assignment, 0)
return nil
}
delay := p.RetryDelay
if delay <= 0 {
delay = 2 * time.Second
}
p.event("backend_pending", assignment, delay)
return message.NakWithDelay(delay)
}
@@ -167,7 +192,7 @@ func OpenConsumer(ctx context.Context, manager consumerManager, stream, subjectB
AckPolicy: jetstream.AckExplicitPolicy,
AckWait: 5 * time.Minute,
MaxAckPending: capacity,
MaxDeliver: 20,
MaxDeliver: 1000,
})
if err != nil {
return nil, fmt.Errorf("open %s assignment consumer: %w", backend, err)
+12 -2
View File
@@ -126,13 +126,23 @@ func encodedAssignment(t *testing.T) []byte {
func TestProcessorAcknowledgesPersistedHandoff(t *testing.T) {
message := &fakeMessage{data: encodedAssignment(t)}
processor := Processor{TrustDomain: "ddupan.top", Accepter: &fakeAccepter{accepted: true}, Claims: &fakeClaims{claimed: true}, Admission: &fakeAdmission{allowed: true}}
var events []Event
processor := Processor{TrustDomain: "ddupan.top", Accepter: &fakeAccepter{accepted: true}, Claims: &fakeClaims{claimed: true}, Admission: &fakeAdmission{allowed: true}, OnEvent: func(event Event) { events = append(events, event) }}
if err := processor.Process(context.Background(), message); err != nil {
t.Fatal(err)
}
if message.acked != 1 || message.nacked != 0 {
t.Fatalf("message = %#v", message)
}
want := []string{"received", "capacity_acquired", "backend_ready", "runner_claimed", "acked"}
if len(events) != len(want) {
t.Fatalf("events = %#v", events)
}
for index := range want {
if events[index].Name != want[index] || events[index].AssignmentID != "gitea-task-42" || events[index].Backend != taskassignment.BackendPod {
t.Fatalf("event[%d] = %#v", index, events[index])
}
}
}
func TestProcessorRetriesUntilBackendHandoffIsDurable(t *testing.T) {
@@ -205,7 +215,7 @@ func TestOpenConsumerUsesIndependentDurablePerBackend(t *testing.T) {
if _, err := OpenConsumer(context.Background(), manager, "CI_RUNNER", "ci.assignment", taskassignment.BackendPod, 4); err != nil {
t.Fatal(err)
}
if manager.config.Durable != "pod" || manager.config.FilterSubject != "ci.assignment.pod" || manager.config.AckPolicy != jetstream.AckExplicitPolicy || manager.config.MaxAckPending != 4 {
if manager.config.Durable != "pod" || manager.config.FilterSubject != "ci.assignment.pod" || manager.config.AckPolicy != jetstream.AckExplicitPolicy || manager.config.MaxAckPending != 4 || manager.config.MaxDeliver != 1000 {
t.Fatalf("config = %#v", manager.config)
}
}
+69 -4
View File
@@ -3,9 +3,13 @@ package opensandboxbackend
import (
"context"
"encoding/base64"
"encoding/json"
"errors"
"fmt"
"net/http"
"sort"
"strings"
"time"
opensandbox "github.com/alibaba/OpenSandbox/sdks/sandbox/go"
@@ -17,6 +21,8 @@ import (
const assignmentMetadata = "ci.ddupan.top/assignment-id"
const terminalMetadata = "ci.ddupan.top/terminal"
const annotationsMetadataPrefix = "ci.ddupan.top/annotations-"
const metadataValueLimit = 63
type Lifecycle interface {
ListSandboxes(context.Context, opensandbox.ListOptions) (*opensandbox.ListSandboxesResponse, error)
@@ -123,7 +129,11 @@ func (b Backend) RecoverAssignments(ctx context.Context, trustDomain string) ([]
if sandbox.Metadata[terminalMetadata] == "true" {
continue
}
assignment, err := taskassignment.FromMetadata(sandbox.Metadata, sandbox.Metadata, trustDomain)
annotations, err := decodeAnnotations(sandbox.Metadata)
if err != nil {
return nil, fmt.Errorf("decode sandbox %s annotations: %w", sandbox.ID, err)
}
assignment, err := taskassignment.FromMetadata(sandbox.Metadata, annotations, trustDomain)
if err != nil {
return nil, fmt.Errorf("recover sandbox %s: %w", sandbox.ID, err)
}
@@ -169,8 +179,12 @@ func (b Backend) Create(ctx context.Context, assignment taskassignment.Assignmen
for key, value := range launch.Environment {
environment[key] = value
}
sandboxMetadata := clone(launch.Metadata.Annotations)
for key, value := range launch.Metadata.Labels {
sandboxMetadata := clone(launch.Metadata.Labels)
annotations, err := encodeAnnotations(launch.Metadata.Annotations)
if err != nil {
return nil, fmt.Errorf("encode sandbox annotations: %w", err)
}
for key, value := range annotations {
sandboxMetadata[key] = value
}
request := opensandbox.CreateSandboxRequest{
@@ -194,7 +208,11 @@ func (b Backend) BindIdentity(ctx context.Context, executor *taskworker.Executor
if err != nil {
return fmt.Errorf("verify sandbox identity metadata: %w", err)
}
if sandbox.Metadata["ci.ddupan.top/spiffe-id"] != identity.SPIFFEID {
annotations, err := decodeAnnotations(sandbox.Metadata)
if err != nil {
return fmt.Errorf("decode sandbox identity metadata: %w", err)
}
if annotations["ci.ddupan.top/spiffe-id"] != identity.SPIFFEID {
return fmt.Errorf("sandbox %s has inconsistent SPIFFE identity metadata", executor.Name)
}
return nil
@@ -243,3 +261,50 @@ func clone(source map[string]string) map[string]string {
}
return result
}
// OpenSandbox metadata follows Kubernetes label-value constraints, unlike Pod
// annotations. Store the annotation map as deterministic URL-safe base64
// chunks so repository paths and SPIFFE IDs remain lossless and recoverable.
func encodeAnnotations(annotations map[string]string) (map[string]string, error) {
data, err := json.Marshal(annotations)
if err != nil {
return nil, err
}
encoded := base64.RawURLEncoding.EncodeToString(data)
result := make(map[string]string, (len(encoded)+metadataValueLimit-1)/metadataValueLimit)
for index := 0; len(encoded) > 0; index++ {
length := min(metadataValueLimit, len(encoded))
result[fmt.Sprintf("%s%03d", annotationsMetadataPrefix, index)] = encoded[:length]
encoded = encoded[length:]
}
return result, nil
}
func decodeAnnotations(metadata map[string]string) (map[string]string, error) {
keys := make([]string, 0)
for key := range metadata {
if strings.HasPrefix(key, annotationsMetadataPrefix) {
keys = append(keys, key)
}
}
if len(keys) == 0 {
return nil, errors.New("sandbox annotation metadata is missing")
}
sort.Strings(keys)
var encoded strings.Builder
for index, key := range keys {
if key != fmt.Sprintf("%s%03d", annotationsMetadataPrefix, index) {
return nil, errors.New("sandbox annotation metadata chunks are incomplete")
}
encoded.WriteString(metadata[key])
}
data, err := base64.RawURLEncoding.DecodeString(encoded.String())
if err != nil {
return nil, err
}
var annotations map[string]string
if err := json.Unmarshal(data, &annotations); err != nil {
return nil, err
}
return annotations, nil
}
+29 -2
View File
@@ -31,7 +31,8 @@ func (f *fakeLifecycle) GetSandbox(_ context.Context, id string) (*opensandbox.S
return &item, nil
}
}
return &opensandbox.SandboxInfo{ID: id, Metadata: map[string]string{"ci.ddupan.top/spiffe-id": assignment().Identity.SPIFFEID}}, nil
metadata, _ := encodeAnnotations(map[string]string{"ci.ddupan.top/spiffe-id": assignment().Identity.SPIFFEID})
return &opensandbox.SandboxInfo{ID: id, Metadata: metadata}, nil
}
func (f *fakeLifecycle) PatchSandboxMetadata(_ context.Context, id string, patch opensandbox.MetadataPatch) (*opensandbox.SandboxInfo, error) {
for index := range f.items {
@@ -89,12 +90,38 @@ func TestCreateUsesPoolAndPersistsRecoveryMetadata(t *testing.T) {
if err != nil {
t.Fatal(err)
}
if lifecycle.created.Extensions["poolRef"] != "ci-vm" || lifecycle.created.Metadata[assignmentMetadata] != assignment().ID || lifecycle.created.Env["CI_SPIFFE_ID"] != assignment().Identity.SPIFFEID || executor.Name != "sandbox-42" {
if lifecycle.created.Extensions["poolRef"] != "ci-vm" || lifecycle.created.Metadata[assignmentMetadata] != assignment().ID || lifecycle.created.Metadata["ci.ddupan.top/runner"] != "true" || lifecycle.created.Env["CI_SPIFFE_ID"] != assignment().Identity.SPIFFEID || executor.Name != "sandbox-42" {
t.Fatalf("request=%#v executor=%#v", lifecycle.created, executor)
}
if lifecycle.created.Env["CI_RUNNER_CAPABILITY"] != "capability" {
t.Fatalf("environment = %#v", lifecycle.created.Env)
}
annotations, err := decodeAnnotations(lifecycle.created.Metadata)
if err != nil || annotations["ci.ddupan.top/repository"] != "owner/repo" || annotations["ci.ddupan.top/spiffe-id"] != assignment().Identity.SPIFFEID {
t.Fatalf("annotations=%#v err=%v", annotations, err)
}
for _, value := range lifecycle.created.Metadata {
if len(value) > metadataValueLimit {
t.Fatalf("metadata value exceeds %d characters: %q", metadataValueLimit, value)
}
}
}
func TestAnnotationMetadataRoundTripPreservesSlashValues(t *testing.T) {
want := taskworker.BackendMetadata(assignment()).Annotations
encoded, err := encodeAnnotations(want)
if err != nil {
t.Fatal(err)
}
got, err := decodeAnnotations(encoded)
if err != nil {
t.Fatal(err)
}
for key, value := range want {
if got[key] != value {
t.Fatalf("%s=%q, want %q", key, got[key], value)
}
}
}
func TestBindIdentityVerifiesPersistedMetadata(t *testing.T) {
+1 -1
View File
@@ -102,7 +102,7 @@ func backendFromTask(task *runnerv1.Task) (Backend, error) {
return "", fmt.Errorf("task runs-on labels must include self-hosted: %v", labels)
}
hasPod := slices.Contains(labels, string(BackendPod))
hasVM := slices.Contains(labels, string(BackendVM))
hasVM := slices.Contains(labels, string(BackendVM)) || slices.Contains(labels, "vm-dev")
if hasPod == hasVM {
return "", fmt.Errorf("task runs-on labels must select exactly one of pod or vm: %v", labels)
}
@@ -28,6 +28,7 @@ func TestNewSelectsBackendFromRunsOn(t *testing.T) {
}{
{"[self-hosted, pod]", BackendPod},
{"[self-hosted, vm]", BackendVM},
{"[self-hosted, vm-dev]", BackendVM},
} {
assignment, err := New(task(t, test.labels), "ddupan.top")
if err != nil {
+29
View File
@@ -71,6 +71,28 @@ type Worker struct {
Backend Backend
Tasks TaskState
Bootstrap Bootstrap
OnEvent func(Event)
}
// Event describes a backend lifecycle transition without exposing launch
// environment values or other credentials.
type Event struct {
Name string
AssignmentID string
Executor string
Phase Phase
}
func (w Worker) event(name string, assignment taskassignment.Assignment, executor *Executor) {
if w.OnEvent == nil {
return
}
event := Event{Name: name, AssignmentID: assignment.ID}
if executor != nil {
event.Executor = executor.Name
event.Phase = executor.Phase
}
w.OnEvent(event)
}
// Accept completes the durable handoff from JetStream to the backend. Once it
@@ -88,6 +110,7 @@ func (w Worker) Accept(ctx context.Context, assignment taskassignment.Assignment
return false, err
}
if executor == nil {
w.event("executor_absent", assignment, nil)
launch, launchErr := w.launchSpec(assignment)
if launchErr != nil {
return false, launchErr
@@ -96,13 +119,18 @@ func (w Worker) Accept(ctx context.Context, assignment taskassignment.Assignment
if err != nil {
return false, err
}
w.event("executor_created", assignment, executor)
} else {
w.event("executor_found", assignment, executor)
}
if executor.IdentityTarget == "" {
w.event("identity_target_pending", assignment, executor)
return false, nil
}
if err := w.Backend.BindIdentity(ctx, executor, assignment.Identity); err != nil {
return false, err
}
w.event("identity_bound", assignment, executor)
return true, nil
}
@@ -178,6 +206,7 @@ func (w Worker) launchSpec(assignment taskassignment.Assignment) (LaunchSpec, er
func BackendMetadata(assignment taskassignment.Assignment) Metadata {
return Metadata{
Labels: map[string]string{
"ci.ddupan.top/runner": "true",
"ci.ddupan.top/assignment-id": assignment.ID,
"ci.ddupan.top/task-id": strconv.FormatInt(assignment.Task.GetId(), 10),
"ci.ddupan.top/backend": string(assignment.Backend),
+14 -1
View File
@@ -79,7 +79,8 @@ func TestHandleRecoversExistingExecutorWithoutCreatingAnother(t *testing.T) {
func TestAcceptAcknowledgesAfterBackendAndIdentityAreDurable(t *testing.T) {
backend := &fakeBackend{}
worker := Worker{Backend: backend, Bootstrap: fakeBootstrap{}}
var events []Event
worker := Worker{Backend: backend, Bootstrap: fakeBootstrap{}, OnEvent: func(event Event) { events = append(events, event) }}
accepted, err := worker.Accept(context.Background(), assignment())
if err != nil || !accepted {
@@ -88,6 +89,18 @@ func TestAcceptAcknowledgesAfterBackendAndIdentityAreDurable(t *testing.T) {
if backend.created != 1 || backend.bound != 1 || backend.deleted != 0 {
t.Fatalf("created=%d bound=%d deleted=%d", backend.created, backend.bound, backend.deleted)
}
want := []string{"executor_absent", "executor_created", "identity_bound"}
if len(events) != len(want) {
t.Fatalf("events = %#v", events)
}
for index := range want {
if events[index].Name != want[index] || events[index].AssignmentID != "gitea-task-42" {
t.Fatalf("event[%d] = %#v", index, events[index])
}
}
if events[1].Executor != "executor" || events[1].Phase != PhaseRunning {
t.Fatalf("created event = %#v", events[1])
}
}
func TestAcceptRetriesWhileBackendIdentityTargetIsUnavailable(t *testing.T) {