Files
act_runner/internal/app/run/runner_test.go
Lunny Xiao 2e72cf902a enhance(kubernetes): fetch job pod copies from the runner over the pod network (#1354)
The Kubernetes backend copies actions and workspaces into job containers with `tar` over `pods/exec`, so every byte goes through the API server. On managed clusters that often sits in other zones than the nodes, so each byte can cross zones twice, and the copies load the API server.

With `kubernetes.transfer.mode: auto` or `direct`, an init container copies `gitea-runner` into the job pod, and the job container fetches directories and copies of 64 KiB or more gzipped from the runner over the pod network. `auto` falls back to the API server for a job pod that cannot reach the runner, `direct` fails the job.

Copies can hold secrets, so each is served once, over TLS 1.3 pinned to a certificate the runner generates, only to the job pod's IPs.

Copying setup-go@v5, cache@v4, setup-java@v4 and build-push-action@v6 sends 11.6 MB through the API server on main and 532 B with this PR.

Co-authored-by: silverwind <2021+silverwind@noreply.gitea.com>
Co-authored-by: silverwind <me@silverwind.io>
Reviewed-on: https://gitea.com/gitea/runner/pulls/1354
Reviewed-by: silverwind <2021+silverwind@noreply.gitea.com>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
2026-10-31 08:00:36 +00:00

462 lines
19 KiB
Go

// Copyright 2026 The Gitea Authors. All rights reserved.
// SPDX-License-Identifier: MIT
package run
import (
"context"
"encoding/base64"
"errors"
"io"
"net/http"
"net/http/httptest"
"os"
"path/filepath"
"runtime"
"strings"
"sync/atomic"
"testing"
"time"
"gitea.com/gitea/runner/act/artifactcache"
"gitea.com/gitea/runner/act/common"
"gitea.com/gitea/runner/act/common/logging"
"gitea.com/gitea/runner/act/container"
"gitea.com/gitea/runner/act/runner"
clientmocks "gitea.com/gitea/runner/internal/pkg/client/mocks"
"gitea.com/gitea/runner/internal/pkg/config"
"gitea.com/gitea/runner/internal/pkg/labels"
"gitea.com/gitea/runner/internal/pkg/ver"
"connectrpc.com/connect/v2"
"gitea.dev/actionslib/pkg/model"
runnerv1 "gitea.dev/actionslib/runner/v1"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/mock"
"github.com/stretchr/testify/require"
"google.golang.org/protobuf/types/known/structpb"
)
func TestRunnerCapabilitiesAndDeclare(t *testing.T) {
require.Equal(t, []string{CapabilityCancelling}, RunnerCapabilities())
cli := clientmocks.NewClient(t)
cli.On("Declare", mock.Anything, mock.MatchedBy(func(req *runnerv1.DeclareRequest) bool {
return req.Version == ver.Version() &&
len(req.Labels) == 1 &&
req.Labels[0] == "ubuntu" &&
len(req.Capabilities) == 1 &&
req.Capabilities[0] == CapabilityCancelling
})).Return(&runnerv1.DeclareResponse{}, nil).Run(func(args mock.Arguments) {
info, ok := connect.CallInfoForClientContext(args.Get(0).(context.Context))
require.True(t, ok)
info.ResponseHeader().Set("X-Gitea-Actions-Capabilities", " cancelling,cache-v2 ")
})
r := &Runner{client: cli}
_, err := r.Declare(t.Context(), []string{"ubuntu"})
require.NoError(t, err)
require.Equal(t, "cancelling,cache-v2", r.capabilities)
}
func TestRunnerDefaultActionsURLUsesMirrorOnlyForGithub(t *testing.T) {
r := &Runner{cfg: &config.Config{}}
r.cfg.Runner.GithubMirror = "https://mirror.example"
task := taskWithDefaultActionsURL("https://github.com")
require.Equal(t, "https://mirror.example", r.getDefaultActionsURL(task))
task = taskWithDefaultActionsURL("https://gitea.example")
require.Equal(t, "https://gitea.example", r.getDefaultActionsURL(task))
}
func TestRunnerRunningCountAndNullLogger(t *testing.T) {
r := &Runner{}
require.Equal(t, int64(0), r.RunningCount())
r.runningCount.Add(2)
require.Equal(t, int64(2), r.RunningCount())
handler := NullLogger{}.WithJobLogger()
require.NotNil(t, handler)
require.False(t, handler.Enabled(t.Context(), logging.LevelError))
}
func TestRunnerReclaimsVolumesAfterReporting(t *testing.T) {
for _, mode := range []string{"deferred", "post-task script"} {
t.Run(mode, func(t *testing.T) {
cfg := &config.Config{
Cache: config.Cache{Enabled: new(false)},
Runner: config.Runner{Timeout: time.Minute, LogReportInterval: time.Minute, StateReportInterval: time.Minute},
Container: config.Container{Network: "host", DockerHost: "-"},
}
if mode == "post-task script" {
if runtime.GOOS == "windows" {
t.Skip("uses a POSIX script")
}
cfg.Runner.PostTaskScript = filepath.Join(t.TempDir(), "post-task.sh")
require.NoError(t, os.WriteFile(cfg.Runner.PostTaskScript, []byte("#!/bin/sh\n: > \"$0.done\"\n"), 0o700))
}
cli := clientmocks.NewClient(t)
cli.AddressValue = "https://gitea.example/"
r := NewRunner(cfg, &config.Registration{UUID: "runner-1", Labels: []string{"ubuntu:docker://node:20"}}, cli)
var reported atomic.Bool
var removed atomic.Int64
cli.On("UpdateLog", mock.Anything, mock.Anything).Return(func(_ context.Context, req *runnerv1.UpdateLogRequest) (*runnerv1.UpdateLogResponse, error) {
assert.False(t, reported.Load())
return &runnerv1.UpdateLogResponse{AckIndex: req.Index + int64(len(req.Rows))}, nil
})
cli.On("UpdateTask", mock.Anything, mock.Anything).Return(func(_ context.Context, req *runnerv1.UpdateTaskRequest) (*runnerv1.UpdateTaskResponse, error) {
if req.State.Result != runnerv1.Result_RESULT_UNSPECIFIED {
assert.Equal(t, runnerv1.Result_RESULT_SUCCESS, req.State.Result)
assert.Equal(t, int64(1), r.RunningCount())
if mode == "post-task script" {
assert.FileExists(t, cfg.Runner.PostTaskScript+".done")
}
reported.Store(true)
}
return &runnerv1.UpdateTaskResponse{State: req.State}, nil
})
var volumesCreated atomic.Bool
daemon := httptest.NewServer(http.HandlerFunc(func(writer http.ResponseWriter, request *http.Request) {
writer.Header().Set("Content-Type", "application/json")
writer.Header().Set("API-Version", "1.47")
path := strings.TrimPrefix(request.URL.Path, "/v1.47")
if response, ok := map[string]string{
"/_ping": "OK",
"/info": `{"Architecture":"amd64","OSType":"linux"}`,
"/containers/json": "[]",
"/networks": "[]",
"/volumes": `{"Volumes":[]}`,
"/containers/create": `{"Id":"job-id"}`,
"/containers/job-id/json": `{"Id":"job-id","Config":{},"State":{"Status":"running"}}`,
}[path]; ok {
_, _ = io.WriteString(writer, response)
return
}
switch {
case strings.HasPrefix(path, "/images/"):
_, _ = io.WriteString(writer, `{"Id":"image-id","Config":{},"Os":"linux","Architecture":"amd64"}`)
case path == "/volumes/create":
volumesCreated.Store(true)
_, _ = io.WriteString(writer, "{}")
case strings.HasSuffix(path, "/exec"):
_, _ = io.WriteString(writer, `{"Id":"exec-id"}`)
case strings.HasPrefix(path, "/exec/"):
_, _ = io.WriteString(writer, `{"Running":false,"ExitCode":0}`)
case request.Method == http.MethodDelete || strings.HasSuffix(path, "/start") || strings.HasSuffix(path, "/kill") || strings.HasSuffix(path, "/archive"):
if request.Method == http.MethodDelete && strings.HasPrefix(path, "/volumes/") && volumesCreated.Load() {
assert.Equal(t, mode != "post-task script", reported.Load())
if mode == "post-task script" {
assert.NoFileExists(t, cfg.Runner.PostTaskScript+".done")
}
assert.Equal(t, int64(1), r.RunningCount())
removed.Add(1)
}
writer.WriteHeader(http.StatusNoContent)
case strings.HasPrefix(path, "/containers/") && strings.HasSuffix(path, "/json"):
http.NotFound(writer, request)
default:
t.Errorf("unexpected Docker request: %s %s", request.Method, request.URL)
http.NotFound(writer, request)
}
}))
t.Cleanup(daemon.Close)
t.Setenv("DOCKER_HOST", daemon.URL)
require.NoError(t, r.Run(t.Context(), &runnerv1.Task{
Context: &structpb.Struct{},
WorkflowPayload: []byte("jobs:\n job:\n runs-on: ubuntu\n steps:\n - run: exit 0\n if: false\n"),
}))
assert.True(t, reported.Load())
assert.Equal(t, int64(2), removed.Load())
assert.Zero(t, r.RunningCount())
})
}
}
func TestCleanupJobVolumesJoinsErrorsAfterCancellation(t *testing.T) {
ctx, cancel := context.WithCancel(t.Context())
cancel()
err := cleanupJobVolumes(ctx, []common.Executor{
func(ctx context.Context) error {
require.NoError(t, ctx.Err())
deadline, ok := ctx.Deadline()
assert.True(t, ok)
assert.InDelta(t, time.Minute.Seconds(), time.Until(deadline).Seconds(), 1)
return io.EOF
},
func(context.Context) error { return io.ErrClosedPipe },
})
require.ErrorIs(t, err, io.EOF)
require.ErrorIs(t, err, io.ErrClosedPipe)
}
func TestRunnerReturnsMaskedFinalReportingError(t *testing.T) {
cfg, err := config.LoadDefault("")
require.NoError(t, err)
cfg.Cache.Enabled = new(false)
cfg.Runner.ReportCloseTimeout = time.Millisecond
cli := clientmocks.NewClient(t)
cli.AddressValue = "https://gitea.example/"
reportErr := errors.New("reporting unavailable for hunter2")
cli.On("UpdateLog", mock.Anything, mock.Anything).Return(nil, reportErr)
cli.On("UpdateTask", mock.Anything, mock.Anything).Return(nil, reportErr)
r := NewRunner(cfg, &config.Registration{}, cli)
err = r.Run(t.Context(), &runnerv1.Task{
Context: &structpb.Struct{},
WorkflowPayload: []byte("["),
Secrets: map[string]string{"PASSWORD": "hunter2"},
})
require.ErrorContains(t, err, "reporting unavailable for ***")
assert.NotContains(t, err.Error(), "hunter2")
assert.Zero(t, r.RunningCount())
}
func TestNewRunnerInitializesLabelsAndEnvironment(t *testing.T) {
cacheEnabled := false
cfg := &config.Config{}
cfg.Cache.Enabled = &cacheEnabled
cfg.Runner.Envs = map[string]string{"EXISTING": "value"}
reg := &config.Registration{
Name: "runner",
Labels: []string{"ubuntu:host", "", "pool:e57e18d4"},
}
cli := clientmocks.NewClient(t)
cli.AddressValue = "https://gitea.example/"
r := NewRunner(cfg, reg, cli)
require.Equal(t, "runner", r.name)
require.Len(t, r.labels, 2)
require.Equal(t, []string{"ubuntu", "pool:e57e18d4"}, r.labels.Names())
require.Equal(t, "value", r.envs["EXISTING"])
require.Equal(t, "https://gitea.example/api/actions_pipeline/", r.envs["ACTIONS_RUNTIME_URL"])
require.Equal(t, "https://gitea.example", r.envs["ACTIONS_RESULTS_URL"])
require.Equal(t, "true", r.envs["GITEA_ACTIONS"])
require.NotEmpty(t, r.envs["GITEA_ACTIONS_RUNNER_VERSION"])
require.Nil(t, r.cacheHandler)
require.Empty(t, r.envs[runner.CacheServiceV2Env], "no cache server, nothing to serve v2 from")
}
func TestRunnerFallbackPlatform(t *testing.T) {
tests := []struct {
name string
label string
dockerRunning bool
want string
}{
{"a docker label needs no daemon probe", "ubuntu:docker://node:18", false, "mirror.example/ci:noble"},
{"host labels keep the image where docker runs", "ubuntu:host", true, "mirror.example/ci:noble"},
{"host labels without docker run on the host", "ubuntu:host", false, labels.SelfHostedPlatform},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
reachable := dockerReachable
dockerReachable = func(context.Context) bool { return tt.dockerRunning }
t.Cleanup(func() { dockerReachable = reachable })
label, err := labels.Parse(tt.label)
require.NoError(t, err)
cfg := &config.Config{}
cfg.Runner.DefaultImage = "mirror.example/ci:noble"
r := &Runner{cfg: cfg, labels: labels.Labels{label}}
require.Equal(t, tt.want, r.fallbackPlatform(t.Context()))
})
}
}
func TestRunnerKubernetesOptionsLayerTheLabelPodTemplatesInRunsOnOrder(t *testing.T) {
cfg := &config.Config{Kubernetes: config.Kubernetes{
PodTemplate: map[string]any{"base": true},
PodTemplates: map[string]map[string]any{"gpu": {"gpu": true}, "large": {"large": true}, "other": {"other": true}},
}}
r := &Runner{cfg: cfg, labels: labels.Labels{{Name: "linux"}, {Name: "gpu"}, {Name: "large"}}}
require.Equal(t, []map[string]any{{"base": true}, {"large": true}, {"gpu": true}}, r.kubernetesOptions([]string{"linux", "large", "other", "gpu", "large"}).PodTemplates)
}
func TestRunnerKubernetesOptionsTransfer(t *testing.T) {
kubernetesRunner := &Runner{cfg: &config.Config{}}
kubernetesRunner.cfg.Kubernetes.Transfer = config.KubernetesTransfer{Mode: config.KubernetesTransferModeAPI, Host: "10.0.0.5", Port: 4000, Image: "gitea/runner:latest"}
require.Nil(t, kubernetesRunner.kubernetesOptions(nil).Transfer)
kubernetesRunner.cfg.Kubernetes.Transfer.Mode = config.KubernetesTransferModeAuto
require.Equal(t, &container.KubernetesTransfer{Host: "10.0.0.5", Port: 4000, Image: "gitea/runner:latest", Fallback: true}, kubernetesRunner.kubernetesOptions(nil).Transfer)
kubernetesRunner.cfg.Kubernetes.Transfer.Mode = config.KubernetesTransferModeDirect
require.Equal(t, &container.KubernetesTransfer{Host: "10.0.0.5", Port: 4000, Image: "gitea/runner:latest"}, kubernetesRunner.kubernetesOptions(nil).Transfer)
}
// Proxy variables are assembled per task, because a job's service containers have to be
// reached directly and they are only known once the workflow is parsed.
func TestNewRunnerLeavesProxyToTheTask(t *testing.T) {
clearProxyEnv(t)
t.Setenv("http_proxy", "http://proxy:3128")
cfg := &config.Config{}
cfg.Cache.ExternalServer = "http://cache.local:8088/"
cfg.Cache.Host = "runner.local"
reg := &config.Registration{Name: "runner"}
cli := clientmocks.NewClient(t)
cli.AddressValue = "https://gitea.example/"
r := NewRunner(cfg, reg, cli)
t.Cleanup(func() { _ = r.Close() })
require.NotContains(t, r.envs, "http_proxy")
require.NotContains(t, r.envs, "no_proxy")
proxyEnv := JobProxyEnv(r.envs, r.builtInCacheURL(), nil)
assert.Contains(t, proxyEnv["no_proxy"], "runner.local")
assert.NotContains(t, proxyEnv["no_proxy"], "cache.local")
}
func taskWithDefaultActionsURL(url string) *runnerv1.Task {
return &runnerv1.Task{
Context: &structpb.Struct{
Fields: map[string]*structpb.Value{
"gitea_default_actions_url": structpb.NewStringValue(url),
},
},
}
}
// The results service is decided per task, because that is where the job's token and its
// instance are both known. NewRunner only leaves the address Gitea serves.
func TestNewRunnerCacheServiceV2(t *testing.T) {
cfg := &config.Config{}
cfg.Cache.Dir, cfg.Cache.Host = t.TempDir(), "127.0.0.1"
cli := clientmocks.NewClient(t)
cli.AddressValue = "https://gitea.example/"
r := NewRunner(cfg, &config.Registration{Name: "runner"}, cli)
t.Cleanup(func() { _ = r.Close() })
const token = "task-token"
assert.Equal(t, "https://gitea.example", r.envs["ACTIONS_RESULTS_URL"])
assert.Empty(t, r.envs[runner.CacheServiceV2Env], "a promise the runner has not made yet")
// The registration is what makes it true: the cache server takes the results service over,
// having been told which instance to forward the artifact half to.
revoke, resultsURL := r.registerCacheForTask(token, artifactcache.JobCredential{Repo: "owner/repo", PublicURL: r.cacheHandler.ExternalURL()}, nil)
defer revoke()
require.Equal(t, r.cacheHandler.ExternalURL(), resultsURL)
// And what is advertised has to answer the cache service. A client without the GHES escape
// hatch, docker buildx among them, posts its cache calls at exactly this URL and nowhere else,
// so a 404 here is the failure this whole change is about.
req, err := http.NewRequestWithContext(t.Context(), http.MethodPost,
resultsURL+"/twirp/github.actions.results.api.v1.CacheService/GetCacheEntryDownloadURL",
strings.NewReader(`{"key":"k","version":"v"}`))
require.NoError(t, err)
req.Header.Set("Authorization", "Bearer "+token)
resp, err := http.DefaultClient.Do(req)
require.NoError(t, err)
defer resp.Body.Close()
assert.Equal(t, http.StatusOK, resp.StatusCode, "the advertised results service serves no cache service")
assert.Equal(t, r.envs["ACTIONS_CACHE_URL"], r.builtInCacheURL(), "the address only the runner knows is bypassed for the operator")
envs := r.cloneEnvs()
r.setResultsService(envs, resultsURL)
assert.Equal(t, resultsURL, envs["ACTIONS_RESULTS_URL"])
assert.Equal(t, "true", envs[runner.CacheServiceV2Env])
cfg.Cache.V2 = new(bool)
envs = r.cloneEnvs()
envs[runner.CacheServiceV2Env] = "true"
r.setResultsService(envs, resultsURL)
assert.Equal(t, "https://gitea.example", envs["ACTIONS_RESULTS_URL"])
assert.Empty(t, envs[runner.CacheServiceV2Env], "a runner.envs entry would promise v2 at an origin not serving it")
envs = r.cloneEnvs()
envs[runner.CacheServiceV2Env] = "true"
envs["ACTIONS_RESULTS_URL"] = "https://gitea.example/sub"
r.setResultsService(envs, "")
assert.Equal(t, "https://gitea.example/sub", envs["ACTIONS_RESULTS_URL"], "with no cache server there is nothing to front with")
assert.Empty(t, envs[runner.CacheServiceV2Env])
for instance, insecure := range map[string]bool{
"https://gitea.example/sub": false,
"https://self-signed.example": true,
} {
cfg.Runner.Insecure = insecure
envs = r.cloneEnvs()
envs["ACTIONS_RESULTS_URL"] = instance
r.setResultsService(envs, resultsURL)
assert.Equal(t, resultsURL, envs["ACTIONS_RESULTS_URL"], instance)
assert.Empty(t, envs[runner.CacheServiceV2Env])
}
workflow, err := model.ReadWorkflow(strings.NewReader(`jobs: {native: {runs-on: native}, linux: {runs-on: linux}, containerized: {runs-on: native, container: alpine}, empty: {runs-on: native, container: ""}}`))
require.NoError(t, err)
r.isolatedCacheContainer = func() string { return "a1b2c3d4e5f6" }
pickPlatform := func(runsOn []string) string { return map[string]string{"native": labels.SelfHostedPlatform}[runsOn[0]] }
for job, isolated := range map[string]bool{"native": false, "linux": true, "containerized": true, "empty": false} {
cacheURL, cacheContainer := r.cacheForJob(workflow.GetJob(job), pickPlatform(workflow.GetJob(job).RunsOn()))
if isolated {
assert.Equal(t, "a1b2c3d4e5f6", cacheContainer, job)
assert.Equal(t, strings.Replace(r.cacheHandler.ExternalURL(), "127.0.0.1", cacheContainer, 1), cacheURL, job)
} else {
assert.Empty(t, cacheContainer, job)
assert.Equal(t, r.cacheHandler.ExternalURL(), cacheURL, job)
}
}
}
func TestRestoreCalledWorkflowContext(t *testing.T) {
newTask := func(workflowCall map[string]any) *runnerv1.Task {
values := map[string]any{
"event_name": "workflow_call",
"event": map[string]any{"ref": "refs/heads/main", "inputs": map[string]any{"env": "leaf"}},
}
if workflowCall != nil {
values["gitea_workflow_call"] = workflowCall
}
taskCtx, err := structpb.NewStruct(values)
require.NoError(t, err)
return &runnerv1.Task{Context: taskCtx}
}
t.Run("dispatched run", func(t *testing.T) {
task := newTask(map[string]any{
"original_event_name": "workflow_dispatch",
"original_event_inputs": map[string]any{"target": "prod"},
"inputs": map[string]any{"target": "prod", "env": "leaf"},
})
assert.Equal(t, map[string]any{"target": "prod", "env": "leaf"}, restoreCalledWorkflowContext(task))
taskCtx := task.Context.AsMap()
assert.Equal(t, "workflow_dispatch", taskCtx["event_name"])
assert.Equal(t, map[string]any{"ref": "refs/heads/main", "inputs": map[string]any{"target": "prod"}}, taskCtx["event"])
assert.Equal(t, []string{"::group::Inputs", "env: leaf", "target: prod", "::endgroup::"}, inputLines(task.Context.Fields))
})
t.Run("event without inputs", func(t *testing.T) {
task := newTask(map[string]any{
"original_event_name": "push",
"inputs": map[string]any{"env": "leaf"},
})
assert.Equal(t, map[string]any{"env": "leaf"}, restoreCalledWorkflowContext(task))
taskCtx := task.Context.AsMap()
assert.Equal(t, "push", taskCtx["event_name"])
assert.Equal(t, map[string]any{"ref": "refs/heads/main"}, taskCtx["event"])
})
t.Run("not a called workflow's job", func(t *testing.T) {
task := newTask(nil)
assert.Nil(t, restoreCalledWorkflowContext(task))
assert.Equal(t, "workflow_call", task.Context.AsMap()["event_name"])
})
}
func TestTaskLogFields(t *testing.T) {
fields := func(token string) []any {
taskContext, err := structpb.NewStruct(map[string]any{"repository": "owner/repo", "run_id": "159", "run_number": "14", "gitea_runtime_token": token})
require.NoError(t, err)
return taskLogFields(&runnerv1.Task{Id: 72, Context: taskContext}, "build")
}
want := []any{"task_id", int64(72), "job", "build", "repository", "owner/repo", "run_id", "159", "run_number", "14"}
assert.Equal(t, want, fields("malformed"))
assert.Equal(t, append(want, "job_id", int64(221)), fields("header."+base64.RawURLEncoding.EncodeToString([]byte(`{"TaskID":72,"JobID":221}`))+".signature"))
}