Compare commits

...

2 Commits

Author SHA1 Message Date
silverwind
b4a64b97dd feat: support --platform and --pull in container.options (#1104)
Fixes https://gitea.com/gitea/runner/issues/667
Fixes https://gitea.com/gitea/runner/pulls/1103

`container.options` is parsed with docker/cli's `addFlags`, which omits the flags docker/cli registers on the `create` and `run` commands themselves. Those were rejected as unknown — most visibly `--platform`.

| Flag | Behavior | Why |
| --- | --- | --- |
| `--platform` | overrides the runner-wide container architecture | the only way to pick an image architecture per job, e.g. across a matrix |
| `--pull always\|missing\|never` | `always` forces a pull, `never` skips it | `force_pull` is runner-wide config today, with no per-job control |

Both are resolved in `NewContainer` because the image pull runs before the container is created and the two have to agree.

`--name`, `--quiet` and `--disable-content-trust` are now accepted and ignored, and `--use-api-socket` is rejected pointing at `container.docker_host`, so a valid `docker create` line no longer fails outright.

<sub>Written by Claude (Opus 4.8).</sub>

---------

Co-authored-by: bircni <bircni@icloud.com>
Reviewed-on: https://gitea.com/gitea/runner/pulls/1104
Reviewed-by: bircni <bircni@icloud.com>
Co-authored-by: silverwind <me@silverwind.io>
2026-07-24 16:42:30 +00:00
bircni
26f9fb12af fix: clean service containers after failed job setup (#1066)
Fixes https://gitea.com/gitea/runner/issues/659

Docker teardown was chained with `Then`, which stops on the previous step's error and on a cancelled context, so the first failure orphaned everything downstream. A volume that is still in use, or a flaky daemon, left the service containers and the job network behind — matching the reports in the issue, where the leaks show up after failed or overloaded jobs.

Cleanup is now straight-line and best-effort: every step runs regardless of what failed before it, and the errors are joined. Two things follow from that:

- Service containers are removed before the job volumes, since a service can hold one via `--volumes-from` in its options. The network stays last, once every container has detached.
- Only job container and volume errors are returned. Service and network errors are logged, as before, because `stopJobContainer()` also runs as the pre-flight step of job start, where a network cleanup error must not abort the job.

The nil check on `rc.JobContainer` is kept, but it is not the leak reporters are hitting: `container.NewContainer` never returns nil in the docker build, so cleanup with no job container is only reachable in the `WITHOUT_DOCKER` stub.

Addresses https://gitea.com/gitea/runner/pulls/1066#issuecomment-1239073.

Not covered here: the `buildx_buildkit_*` volumes reported in the issue are created by buildx inside the job, not by the runner, so no runner-side teardown removes them.

---------

Co-authored-by: silverwind <2021+silverwind@noreply.gitea.com>
Co-authored-by: silverwind <me@silverwind.io>
Reviewed-on: https://gitea.com/gitea/runner/pulls/1066
Reviewed-by: silverwind <2021+silverwind@noreply.gitea.com>
2026-07-24 16:36:09 +00:00
5 changed files with 248 additions and 46 deletions

View File

@@ -0,0 +1,86 @@
// Copyright 2026 The Gitea Authors. All rights reserved.
// SPDX-License-Identifier: MIT
//go:build !(WITHOUT_DOCKER || !(linux || darwin || windows || netbsd))
package container
import (
"errors"
"fmt"
"io"
"slices"
"github.com/kballard/go-shellquote"
"github.com/spf13/pflag"
)
const (
pullPolicyAlways = "always"
pullPolicyMissing = "missing"
pullPolicyNever = "never"
)
var pullPolicies = []string{pullPolicyAlways, pullPolicyMissing, pullPolicyNever}
// createFlags are the flags docker/cli registers on the `create` and `run` commands
// instead of in addFlags, so they are not part of containerOptions.
type createFlags struct {
platform string
pull string
name string
useAPISocket bool
}
func registerCreateFlags(flags *pflag.FlagSet) *createFlags {
cf := new(createFlags)
flags.StringVar(&cf.platform, "platform", "", "Set platform if server is multi-platform capable")
flags.StringVar(&cf.pull, "pull", pullPolicyMissing, `Pull image before creating ("always", "missing", "never")`)
flags.StringVar(&cf.name, "name", "", "Assign a name to the container")
flags.BoolVar(&cf.useAPISocket, "use-api-socket", false, "Bind mount Docker API socket and required auth")
// Accepted without effect: pull progress is only logged at debug level, and docker
// no longer implements content trust.
flags.BoolP("quiet", "q", false, "Suppress the pull output")
flags.Bool("disable-content-trust", true, "Skip image verification (deprecated)")
return cf
}
// parseContainerOptions parses a container options string. The flags are returned even
// on error, holding whatever was read before the failure.
func parseContainerOptions(options string) (*pflag.FlagSet, *containerOptions, *createFlags, error) {
flags := pflag.NewFlagSet("container_flags", pflag.ContinueOnError)
flags.SetOutput(io.Discard)
copts := addFlags(flags)
cf := registerCreateFlags(flags)
args, err := shellquote.Split(options)
if err != nil {
return flags, copts, cf, fmt.Errorf("Cannot split container options: '%s': '%w'", options, err)
}
if err := flags.Parse(args); err != nil {
return flags, copts, cf, fmt.Errorf("Cannot parse container options: '%s': '%w'", options, err)
}
return flags, copts, cf, nil
}
// createFlagsFromOptions reads the create-level flags that have to be known before the
// container is created. Malformed options keep the defaults here and are reported by
// mergeContainerConfigs at create time.
func createFlagsFromOptions(options string) *createFlags {
_, _, cf, _ := parseContainerOptions(options)
return cf
}
func (cf *createFlags) validate() error {
if !slices.Contains(pullPolicies, cf.pull) {
return fmt.Errorf("invalid --pull option %q: must be one of %q", cf.pull, pullPolicies)
}
if cf.useAPISocket {
return errors.New("--use-api-socket is not supported, use the runner's container.docker_host setting to expose a docker socket")
}
return nil
}

View File

@@ -0,0 +1,61 @@
// Copyright 2026 The Gitea Authors. All rights reserved.
// SPDX-License-Identifier: MIT
package container
import (
"testing"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
)
func TestCreateFlagsFromOptions(t *testing.T) {
for _, tc := range []struct {
options string
platform string
pull string
}{
{"", "", pullPolicyMissing},
{"-v /a:/b --platform=linux/arm64 --pull always", "linux/arm64", pullPolicyAlways},
{"--platform linux/arm/v7 --pull never", "linux/arm/v7", pullPolicyNever},
{`--platform "linux/amd64`, "", pullPolicyMissing}, // malformed, defaults kept
} {
t.Run(tc.options, func(t *testing.T) {
cf := createFlagsFromOptions(tc.options)
assert.Equal(t, tc.platform, cf.platform)
assert.Equal(t, tc.pull, cf.pull)
})
}
}
func TestCreateFlagsValidate(t *testing.T) {
for _, tc := range []struct {
options string
wantErr string
}{
{"--quiet --disable-content-trust --name mine", ""},
{"--pull sometimes", `invalid --pull option "sometimes"`},
{"--use-api-socket", "--use-api-socket is not supported"},
} {
t.Run(tc.options, func(t *testing.T) {
err := createFlagsFromOptions(tc.options).validate()
if tc.wantErr == "" {
require.NoError(t, err)
return
}
require.ErrorContains(t, err, tc.wantErr)
})
}
}
func TestNewContainerAppliesCreateFlags(t *testing.T) {
input := &NewContainerInput{Platform: "linux/amd64", Options: "--platform linux/arm64 --pull never"}
cr := NewContainer(input).(*containerReference)
assert.Equal(t, "linux/arm64", input.Platform)
assert.Equal(t, pullPolicyNever, cr.pullPolicy)
kept := &NewContainerInput{Platform: "linux/amd64", Options: "--privileged"}
NewContainer(kept)
assert.Equal(t, "linux/amd64", kept.Platform)
}

View File

@@ -35,7 +35,6 @@ import (
"github.com/go-git/go-git/v5/plumbing/format/gitignore"
"github.com/gobwas/glob"
"github.com/joho/godotenv"
"github.com/kballard/go-shellquote"
"github.com/moby/moby/api/pkg/stdcopy"
"github.com/moby/moby/api/types/container"
"github.com/moby/moby/api/types/mount"
@@ -43,7 +42,6 @@ import (
"github.com/moby/moby/api/types/system"
"github.com/moby/moby/client"
specs "github.com/opencontainers/image-spec/specs-go/v1"
"github.com/spf13/pflag"
)
// drainGracePeriod bounds how long we wait for an output-copy goroutine to
@@ -57,6 +55,12 @@ const drainGracePeriod = 2 * time.Second
func NewContainer(input *NewContainerInput) ExecutionsEnvironment {
cr := new(containerReference)
cr.input = input
// Resolved up front because the image pull runs before the container is created.
cf := createFlagsFromOptions(input.Options)
if cf.platform != "" {
cr.input.Platform = cf.platform
}
cr.pullPolicy = cf.pull
return cr
}
@@ -137,6 +141,11 @@ func (cr *containerReference) Start(attach bool) common.Executor {
}
func (cr *containerReference) Pull(forcePull bool) common.Executor {
if cr.pullPolicy == pullPolicyNever {
return common.NewInfoExecutor("docker pull skipped image=%s, --pull=never in the options", cr.input.Image)
}
forcePull = forcePull || cr.pullPolicy == pullPolicyAlways
return common.
NewInfoExecutor("docker pull image=%s platform=%s username=%s forcePull=%t", cr.input.Image, cr.input.Platform, cr.input.Username, forcePull).
Then(
@@ -232,11 +241,12 @@ func (cr *containerReference) ReplaceLogWriter(stdout, stderr io.Writer) (io.Wri
}
type containerReference struct {
cli client.APIClient
id string
input *NewContainerInput
UID int
GID int
cli client.APIClient
id string
input *NewContainerInput
pullPolicy string
UID int
GID int
// attachDone is closed by the attach() streaming goroutine once it has
// drained and flushed the container's output. wait() blocks on it so the
// tail of the log lands before the step proceeds.
@@ -411,17 +421,13 @@ func (cr *containerReference) mergeContainerConfigs(ctx context.Context, config
}
// parse configuration from CLI container.options
flags := pflag.NewFlagSet("container_flags", pflag.ContinueOnError)
copts := addFlags(flags)
optionsArgs, err := shellquote.Split(input.Options)
flags, copts, cf, err := parseContainerOptions(input.Options)
if err != nil {
return nil, nil, fmt.Errorf("Cannot split container options: '%s': '%w'", input.Options, err)
return nil, nil, err
}
err = flags.Parse(optionsArgs)
if err != nil {
return nil, nil, fmt.Errorf("Cannot parse container options: '%s': '%w'", input.Options, err)
if err := cf.validate(); err != nil {
return nil, nil, fmt.Errorf("Cannot process container options: '%s': '%w'", input.Options, err)
}
// FIXME: If everything is fine after gitea/act v0.260.0, remove the following comment.
@@ -476,6 +482,9 @@ func (cr *containerReference) mergeContainerConfigs(ctx context.Context, config
}
hostConfig.Binds = binds
hostConfig.Mounts = mounts
if cf.name != "" {
logger.Warn("--name in the options will be ignored.")
}
if len(copts.netMode.Value()) > 0 {
logger.Warn("--network and --net in the options will be ignored.")
}

View File

@@ -454,37 +454,7 @@ func (rc *RunContext) startJobContainer() common.Executor {
rc.ServiceContainers = append(rc.ServiceContainers, c)
}
rc.cleanUpJobContainer = func(ctx context.Context) error {
reuseJobContainer := func(ctx context.Context) bool {
return rc.Config.ReuseContainers
}
if rc.JobContainer != nil {
return rc.JobContainer.Remove().IfNot(reuseJobContainer).
Then(container.NewDockerVolumeRemoveExecutor(rc.jobContainerName(), false)).IfNot(reuseJobContainer).
Then(container.NewDockerVolumeRemoveExecutor(rc.jobContainerName()+"-env", false)).IfNot(reuseJobContainer).
Then(func(ctx context.Context) error {
if len(rc.ServiceContainers) > 0 {
logger.Infof("Cleaning up services for job %s", rc.JobName)
if err := rc.stopServiceContainers()(ctx); err != nil {
logger.Errorf("Error while cleaning services: %v", err)
}
}
if createAndDeleteNetwork {
// clean network if it has been created by act
// if using service containers
// it means that the network to which containers are connecting is created by `runner`,
// so, we should remove the network at last.
logger.Infof("Cleaning up network for job %s, and network name is: %s", rc.JobName, networkName)
if err := container.NewDockerNetworkRemoveExecutor(networkName)(ctx); err != nil {
logger.Errorf("Error while cleaning network: %v", err)
}
}
return nil
})(ctx)
}
return nil
}
rc.cleanUpJobContainer = rc.cleanupJobResources(networkName, createAndDeleteNetwork)
// For Gitea, `jobContainerNetwork` should be the same as `networkName`
jobContainerNetwork := networkName
@@ -539,6 +509,41 @@ func (rc *RunContext) startJobContainer() common.Executor {
}
}
// cleanupJobResources removes everything the job created, continuing past failures.
// Only job container and volume errors are returned, the rest are logged.
func (rc *RunContext) cleanupJobResources(networkName string, createAndDeleteNetwork bool) common.Executor {
return func(ctx context.Context) error {
logger := common.Logger(ctx)
removeJobContainer := rc.JobContainer != nil && !rc.Config.ReuseContainers
var errs []error
if removeJobContainer {
errs = append(errs, rc.JobContainer.Remove()(ctx))
}
if len(rc.ServiceContainers) > 0 {
logger.Infof("Cleaning up services for job %s", rc.JobName)
if err := rc.stopServiceContainers()(ctx); err != nil {
logger.Errorf("Error while cleaning services: %v", err)
}
}
if removeJobContainer {
// after the containers using them, services can hold these via `--volumes-from`
name := rc.jobContainerName()
errs = append(errs,
container.NewDockerVolumeRemoveExecutor(name, false)(ctx),
container.NewDockerVolumeRemoveExecutor(name+"-env", false)(ctx))
}
if createAndDeleteNetwork {
// last, once every container has detached
logger.Infof("Cleaning up network for job %s, and network name is: %s", rc.JobName, networkName)
if err := container.NewDockerNetworkRemoveExecutor(networkName)(ctx); err != nil {
logger.Errorf("Error while cleaning network: %v", err)
}
}
return errors.Join(errs...)
}
}
func (rc *RunContext) execJobContainer(cmd []string, env map[string]string, user, workdir string) common.Executor { //nolint:unparam // pre-existing issue from nektos/act
return func(ctx context.Context) error {
return rc.JobContainer.Exec(cmd, env, user, workdir)(ctx)

View File

@@ -7,6 +7,7 @@ package runner
import (
"bytes"
"context"
"errors"
"fmt"
"os"
"runtime"
@@ -451,6 +452,46 @@ func TestRunContextValidVolumes(t *testing.T) {
assert.Len(t, rc.validVolumes(), len(got), "repeated calls must be stable, not accumulate")
}
func TestCleanupJobResourcesCleansServicesWithoutJobContainer(t *testing.T) {
service := &containerMock{}
service.On("Remove").Return(func(context.Context) error { return nil }).Once()
service.On("Close").Return(func(context.Context) error { return nil }).Once()
rc := &RunContext{
Config: &Config{},
ServiceContainers: []container.ExecutionsEnvironment{service},
}
err := rc.cleanupJobResources("external-network", false)(context.Background())
require.NoError(t, err)
service.AssertExpectations(t)
}
// cleanup used to bail out on a previous step's error and on a cancelled context
func TestCleanupJobResourcesContinuesAfterFailure(t *testing.T) {
t.Setenv("DOCKER_HOST", "unix:///nonexistent.sock")
jobContainer := &containerMock{}
jobContainer.On("Remove").Return(func(context.Context) error { return errors.New("removal failed") }).Once()
service := &containerMock{}
service.On("Remove").Return(func(context.Context) error { return nil }).Once()
service.On("Close").Return(func(context.Context) error { return nil }).Once()
rc := &RunContext{
Name: "job",
Config: &Config{},
Run: &model.Run{Workflow: &model.Workflow{Name: "wf"}, JobID: "job"},
JobContainer: jobContainer,
ServiceContainers: []container.ExecutionsEnvironment{service},
}
ctx, cancel := context.WithCancel(context.Background())
cancel()
require.Error(t, rc.cleanupJobResources("job-network", true)(ctx))
jobContainer.AssertExpectations(t)
service.AssertExpectations(t)
}
// TestInterpolateOutputsIsPerMatrixCombo guards the matrix-output fix: combinations share one
// *model.Job, so each must interpolate from its own pristine snapshot. Otherwise the first
// combo's resolved value freezes the shared template and later combos can't resolve their own.