fix: stop a stalled docker daemon from hanging jobs and the e2e suite (#1404)

When the docker daemon stopped answering on https://gitea.com/gitea/runner/actions/runs/1204971, unbounded exec runs hung the e2e suite until the 20m `go test` panic, which skipped teardown. Exec runs and teardown are now bounded, and attached exec streams, including the runner's user ID probe, close with their context.

CLI daemons now stop with SIGTERM, so job volumes no longer leak.

_Written by Claude (Opus 5.5)._

---------

Co-authored-by: bircni <[email protected]>
Reviewed-on: https://gitea.com/gitea/runner/pulls/1404
Reviewed-by: bircni <[email protected]>
Co-authored-by: silverwind <[email protected]>
This commit is contained in:
silverwindandbircni authored and bircni committed 2026-11-07 08:07:41 +00:00
1 parent 8e999cd039
commit d34e4cf11a
9 files changed
+114 -91

No files matched your search

+29 -15
View File
@@ -9,6 +9,7 @@ import (
"bufio"
"bytes"
"cmp"
"context"
"encoding/json"
"errors"
"os"
@@ -138,14 +139,31 @@ func daemonWorkflow(t *testing.T, m mode, spec workflowSpec) *workflowResult {
return result
}
func execWorkflow(t *testing.T, m mode, spec workflowSpec) *workflowResult {
func runExec(t *testing.T, env []string, args ...string) (stdout, stderr []byte, err error) {
t.Helper()
ctx := t.Context()
bin, err := runnerBinary()
if err != nil {
t.Fatal(err)
}
ctx, cancel := context.WithTimeout(t.Context(), runTimeout)
defer cancel()
cmd := exec.CommandContext(ctx, bin, append([]string{"exec"}, args...)...)
cmd.Env = append(slices.DeleteFunc(os.Environ(), func(kv string) bool { return strings.HasPrefix(kv, "ACTIONS_") }), env...)
cmd.Cancel = func() error { return cmd.Process.Signal(os.Interrupt) }
cmd.WaitDelay = cleanupTimeout
var stdoutBuf, stderrBuf bytes.Buffer
cmd.Stdout, cmd.Stderr = &stdoutBuf, &stderrBuf
err = cmd.Run()
if ctx.Err() != nil {
t.Fatalf("exec did not finish within %s\n%s%s", runTimeout, stdoutBuf.Bytes(), stderrBuf.Bytes())
}
return stdoutBuf.Bytes(), stderrBuf.Bytes(), err
}
func execWorkflow(t *testing.T, m mode, spec workflowSpec) *workflowResult {
t.Helper()
ctx := t.Context()
workdir := t.TempDir()
workflows := filepath.Join(workdir, ".gitea", "workflows")
if err := os.MkdirAll(workflows, 0o755); err != nil {
@@ -170,12 +188,12 @@ func execWorkflow(t *testing.T, m mode, spec workflowSpec) *workflowResult {
}
}
args := []string{"exec", "--json", "-C", workdir, "-W", workflows, "--artifact-server-path", t.TempDir(), "--artifact-server-port", "0"}
env := slices.DeleteFunc(os.Environ(), func(kv string) bool { return strings.HasPrefix(kv, "ACTIONS_") })
args := []string{"--json", "-C", workdir, "-W", workflows, "--artifact-server-path", t.TempDir(), "--artifact-server-port", "0"}
var env []string
if m.kubernetes {
cluster := requireCluster(t)
args = append(args, "-i", "kubernetes://"+cluster.jobImage)
env = append(env, "KUBECONFIG="+cluster.kubernetes.Kubeconfig)
env = []string{"KUBECONFIG=" + cluster.kubernetes.Kubeconfig}
} else {
args = append(args, "-i", os.Getenv("E2E_JOB_IMAGE"), "--network", fixture.network)
}
@@ -185,21 +203,17 @@ func execWorkflow(t *testing.T, m mode, spec workflowSpec) *workflowResult {
args = append(args, "--input", name+"="+value)
}
}
cmd := exec.CommandContext(ctx, bin, append(args, spec.execArgs...)...)
cmd.Env = env
var stdout, stderr bytes.Buffer
cmd.Stdout, cmd.Stderr = &stdout, &stderr
runErr := cmd.Run()
stdout, stderr, runErr := runExec(t, env, append(args, spec.execArgs...)...)
var exitErr *exec.ExitError
if runErr != nil && !errors.As(runErr, &exitErr) {
t.Fatalf("exec: %v\n%s", runErr, stderr.String())
t.Fatalf("exec: %v\n%s", runErr, stderr)
}
result := &workflowResult{conclusion: "success", jobs: map[string]string{}}
if runErr != nil {
result.conclusion = "failure"
}
scanner := bufio.NewScanner(&stdout)
scanner := bufio.NewScanner(bytes.NewReader(stdout))
scanner.Buffer(nil, 1<<20)
for scanner.Scan() {
var entry struct {
@@ -220,8 +234,8 @@ func execWorkflow(t *testing.T, m mode, spec workflowSpec) *workflowResult {
}
}
}
if result.conclusion != "success" && len(result.jobs) == 0 {
t.Fatalf("exec ran no job: %v\n%s", runErr, stderr.String())
if len(result.jobs) == 0 {
t.Fatalf("exec ran no job: %v\n%s", runErr, stderr)
}
return result
}