Skip to content

Commit

Permalink
wait for a given duration in case of imagePullBackOff
Browse files Browse the repository at this point in the history
We have implemented imagePullBackOff as fail fast. The issue with this approach
is, the node where the pod is scheduled often experiences registry rate limit.
The image pull failure because of the rate limit returns the same warning
(reason: Failed and message: ImagePullBackOff). The pod can potentially recover
after waiting for enough time until the cap is expired. Kubernetes can then
successfully pull the image and bring the pod up.

Introducing a default configuration to specify cluster level timeout to allow
the imagePullBackOff to retry for a given duration. Once that duration has
passed, return a permanent failure.

#5987
#7184

Signed-off-by: Priti Desai <[email protected]>

wait for a given duration in case of imagePullBackOff

Signed-off-by: Priti Desai <[email protected]>
  • Loading branch information
pritidesai authored and tekton-robot committed Feb 26, 2024
1 parent 9be03e2 commit a40423b
Show file tree
Hide file tree
Showing 9 changed files with 247 additions and 22 deletions.
5 changes: 5 additions & 0 deletions config/config-defaults.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -87,6 +87,11 @@ data:
# no default-resolver-type is specified by default
default-resolver-type:
# default-imagepullbackoff-timeout contains the default duration to wait
# before requeuing the TaskRun to retry, specifying 0 here is equivalent to fail fast
# possible values could be 1m, 5m, 10s, 1h, etc
# default-imagepullbackoff-timeout: "5m"
# default-container-resource-requirements allow users to update default resource requirements
# to a init-containers and containers of a pods create by the controller
# Onet: All the resource requirements are applied to init-containers and containers
Expand Down
21 changes: 21 additions & 0 deletions docs/additional-configs.md
Original file line number Diff line number Diff line change
Expand Up @@ -31,6 +31,7 @@ installation.
- [Verify the transparency logs using `rekor-cli`](#verify-the-transparency-logs-using-rekor-cli)
- [Verify Tekton Resources](#verify-tekton-resources)
- [Pipelinerun with Affinity Assistant](#pipelineruns-with-affinity-assistant)
- [TaskRuns with `imagePullBackOff` Timeout](#taskruns-with-imagepullbackoff-timeout)
- [Next steps](#next-steps)


Expand Down Expand Up @@ -672,6 +673,26 @@ please take a look at [Trusted Resources](./trusted-resources.md).
The cluster operators can review the [guidelines](developers/affinity-assistant.md) to `cordon` a node in the cluster
with the tekton controller and the affinity assistant is enabled.
## TaskRuns with `imagePullBackOff` Timeout
Tekton pipelines has adopted a fail fast strategy with a taskRun failing with `TaskRunImagePullFailed` in case of an
`imagePullBackOff`. This can be limited in some cases, and it generally depends on the infrastructure. To allow the
cluster operators to decide whether to wait in case of an `imagePullBackOff`, a setting is available to configure
the wait time in minutes such that the controller will wait for the specified duration before declaring a failure.
For example, with the following `config-defaults`, the controller does not mark the taskRun as failure for 5 minutes since
the pod is scheduled in case the image pull fails with `imagePullBackOff`.
See issue https://github.com/tektoncd/pipeline/issues/5987 for more details.
```yaml
apiVersion: v1
kind: ConfigMap
metadata:
name: config-defaults
namespace: tekton-pipelines
data:
default-imagepullbackoff-timeout: "5"
```
## Next steps
To get started with Tekton check the [Introductory tutorials][quickstarts],
Expand Down
14 changes: 14 additions & 0 deletions pkg/apis/config/default.go
Original file line number Diff line number Diff line change
Expand Up @@ -49,6 +49,8 @@ const (
// default resource requirements, will be applied to all the containers, which has empty resource requirements
ResourceRequirementDefaultContainerKey = "default"

DefaultImagePullBackOffTimeout = 0 * time.Minute

defaultTimeoutMinutesKey = "default-timeout-minutes"
defaultServiceAccountKey = "default-service-account"
defaultManagedByLabelValueKey = "default-managed-by-label-value"
Expand All @@ -60,6 +62,7 @@ const (
defaultForbiddenEnv = "default-forbidden-env"
defaultResolverTypeKey = "default-resolver-type"
defaultContainerResourceRequirementsKey = "default-container-resource-requirements"
defaultImagePullBackOffTimeout = "default-imagepullbackoff-timeout"
)

// DefaultConfig holds all the default configurations for the config.
Expand All @@ -79,6 +82,7 @@ type Defaults struct {
DefaultForbiddenEnv []string
DefaultResolverType string
DefaultContainerResourceRequirements map[string]corev1.ResourceRequirements
DefaultImagePullBackOffTimeout time.Duration
}

// GetDefaultsConfigName returns the name of the configmap containing all
Expand Down Expand Up @@ -109,6 +113,7 @@ func (cfg *Defaults) Equals(other *Defaults) bool {
other.DefaultTaskRunWorkspaceBinding == cfg.DefaultTaskRunWorkspaceBinding &&
other.DefaultMaxMatrixCombinationsCount == cfg.DefaultMaxMatrixCombinationsCount &&
other.DefaultResolverType == cfg.DefaultResolverType &&
other.DefaultImagePullBackOffTimeout == cfg.DefaultImagePullBackOffTimeout &&
reflect.DeepEqual(other.DefaultForbiddenEnv, cfg.DefaultForbiddenEnv)
}

Expand All @@ -121,6 +126,7 @@ func NewDefaultsFromMap(cfgMap map[string]string) (*Defaults, error) {
DefaultCloudEventsSink: DefaultCloudEventSinkValue,
DefaultMaxMatrixCombinationsCount: DefaultMaxMatrixCombinationsCount,
DefaultResolverType: DefaultResolverTypeValue,
DefaultImagePullBackOffTimeout: DefaultImagePullBackOffTimeout,
}

if defaultTimeoutMin, ok := cfgMap[defaultTimeoutMinutesKey]; ok {
Expand Down Expand Up @@ -191,6 +197,14 @@ func NewDefaultsFromMap(cfgMap map[string]string) (*Defaults, error) {
tc.DefaultContainerResourceRequirements = resourceRequirementsValue
}

if defaultImagePullBackOff, ok := cfgMap[defaultImagePullBackOffTimeout]; ok {
timeout, err := time.ParseDuration(defaultImagePullBackOff)
if err != nil {
return nil, fmt.Errorf("failed parsing tracing config %q", defaultImagePullBackOffTimeout)
}
tc.DefaultImagePullBackOffTimeout = timeout
}

return &tc, nil
}

Expand Down
31 changes: 31 additions & 0 deletions pkg/apis/config/default_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,7 @@ package config_test

import (
"testing"
"time"

"github.com/google/go-cmp/cmp"
"github.com/tektoncd/pipeline/pkg/apis/config"
Expand All @@ -43,6 +44,7 @@ func TestNewDefaultsFromConfigMap(t *testing.T) {
DefaultManagedByLabelValue: "something-else",
DefaultMaxMatrixCombinationsCount: 256,
DefaultResolverType: "git",
DefaultImagePullBackOffTimeout: time.Duration(5) * time.Second,
},
fileName: config.GetDefaultsConfigName(),
},
Expand All @@ -62,12 +64,16 @@ func TestNewDefaultsFromConfigMap(t *testing.T) {
},
},
DefaultMaxMatrixCombinationsCount: 256,
DefaultImagePullBackOffTimeout: 0,
},
fileName: "config-defaults-with-pod-template",
},
{
expectedError: true,
fileName: "config-defaults-timeout-err",
}, {
expectedError: true,
fileName: "config-defaults-imagepullbackoff-timeout-err",
},
// Previously the yaml package did not support UnmarshalStrict, though
// it's supported now however it may introduce incompatibility, so we decide
Expand All @@ -81,6 +87,7 @@ func TestNewDefaultsFromConfigMap(t *testing.T) {
DefaultManagedByLabelValue: config.DefaultManagedByLabelValue,
DefaultPodTemplate: &pod.Template{},
DefaultMaxMatrixCombinationsCount: 256,
DefaultImagePullBackOffTimeout: 0,
},
},
{
Expand All @@ -92,6 +99,7 @@ func TestNewDefaultsFromConfigMap(t *testing.T) {
DefaultManagedByLabelValue: config.DefaultManagedByLabelValue,
DefaultAAPodTemplate: &pod.AffinityAssistantTemplate{},
DefaultMaxMatrixCombinationsCount: 256,
DefaultImagePullBackOffTimeout: 0,
},
},
{
Expand All @@ -106,6 +114,7 @@ func TestNewDefaultsFromConfigMap(t *testing.T) {
DefaultTimeoutMinutes: 60,
DefaultServiceAccount: "default",
DefaultManagedByLabelValue: config.DefaultManagedByLabelValue,
DefaultImagePullBackOffTimeout: 0,
},
},
{
Expand All @@ -117,6 +126,7 @@ func TestNewDefaultsFromConfigMap(t *testing.T) {
DefaultMaxMatrixCombinationsCount: 256,
DefaultManagedByLabelValue: "tekton-pipelines",
DefaultForbiddenEnv: []string{"TEKTON_POWER_MODE", "TEST_ENV", "TEST_TEKTON"},
DefaultImagePullBackOffTimeout: time.Duration(15) * time.Second,
},
},
{
Expand All @@ -128,6 +138,7 @@ func TestNewDefaultsFromConfigMap(t *testing.T) {
DefaultManagedByLabelValue: "tekton-pipelines",
DefaultMaxMatrixCombinationsCount: 256,
DefaultContainerResourceRequirements: map[string]corev1.ResourceRequirements{},
DefaultImagePullBackOffTimeout: 0,
},
},
{
Expand All @@ -142,6 +153,7 @@ func TestNewDefaultsFromConfigMap(t *testing.T) {
DefaultServiceAccount: "default",
DefaultManagedByLabelValue: "tekton-pipelines",
DefaultMaxMatrixCombinationsCount: 256,
DefaultImagePullBackOffTimeout: 0,
DefaultContainerResourceRequirements: map[string]corev1.ResourceRequirements{
config.ResourceRequirementDefaultContainerKey: {
Requests: corev1.ResourceList{
Expand Down Expand Up @@ -197,6 +209,7 @@ func TestNewDefaultsFromEmptyConfigMap(t *testing.T) {
DefaultManagedByLabelValue: "tekton-pipelines",
DefaultServiceAccount: "default",
DefaultMaxMatrixCombinationsCount: 256,
DefaultImagePullBackOffTimeout: 0,
}
verifyConfigFileWithExpectedConfig(t, DefaultsConfigEmptyName, expectedConfig)
}
Expand Down Expand Up @@ -345,6 +358,24 @@ func TestEquals(t *testing.T) {
DefaultForbiddenEnv: []string{"TEST_ENV", "TEKTON_POWER_MODE"},
},
expected: true,
}, {
name: "different default ImagePullBackOff timeout",
left: &config.Defaults{
DefaultImagePullBackOffTimeout: 10,
},
right: &config.Defaults{
DefaultImagePullBackOffTimeout: 20,
},
expected: false,
}, {
name: "same default ImagePullBackOff timeout",
left: &config.Defaults{
DefaultImagePullBackOffTimeout: 20,
},
right: &config.Defaults{
DefaultImagePullBackOffTimeout: 20,
},
expected: true,
},
}

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -21,3 +21,4 @@ data:
default-timeout-minutes: "50"
default-service-account: "tekton"
default-forbidden-env: "TEST_TEKTON, TEKTON_POWER_MODE,TEST_ENV,TEST_TEKTON"
default-imagepullbackoff-timeout: "15s"
Original file line number Diff line number Diff line change
@@ -0,0 +1,21 @@
# Copyright 2019 The Tekton Authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

apiVersion: v1
kind: ConfigMap
metadata:
name: config-defaults
namespace: tekton-pipelines
data:
default-imagepullbackoff-timeout: "not-a-timeout"
1 change: 1 addition & 0 deletions pkg/apis/config/testdata/config-defaults.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -22,3 +22,4 @@ data:
default-service-account: "tekton"
default-managed-by-label-value: "something-else"
default-resolver-type: "git"
default-imagepullbackoff-timeout: "5s"
48 changes: 45 additions & 3 deletions pkg/reconciler/taskrun/taskrun.go
Original file line number Diff line number Diff line change
Expand Up @@ -93,13 +93,15 @@ type Reconciler struct {
tracerProvider trace.TracerProvider
}

const ImagePullBackOff = "ImagePullBackOff"

var (
// Check that our Reconciler implements taskrunreconciler.Interface
_ taskrunreconciler.Interface = (*Reconciler)(nil)

// Pod failure reasons that trigger failure of the TaskRun
podFailureReasons = map[string]struct{}{
"ImagePullBackOff": {},
ImagePullBackOff: {},
"InvalidImageName": {},
}
)
Expand Down Expand Up @@ -170,7 +172,7 @@ func (c *Reconciler) ReconcileKind(ctx context.Context, tr *v1.TaskRun) pkgrecon
}

// Check for Pod Failures
if failed, reason, message := c.checkPodFailed(tr); failed {
if failed, reason, message := c.checkPodFailed(ctx, tr); failed {
err := c.failTaskRun(ctx, tr, reason, message)
return c.finishReconcileUpdateEmitEvents(ctx, tr, before, err)
}
Expand Down Expand Up @@ -221,10 +223,30 @@ func (c *Reconciler) ReconcileKind(ctx context.Context, tr *v1.TaskRun) pkgrecon
return nil
}

func (c *Reconciler) checkPodFailed(tr *v1.TaskRun) (bool, v1.TaskRunReason, string) {
func (c *Reconciler) checkPodFailed(ctx context.Context, tr *v1.TaskRun) (bool, v1.TaskRunReason, string) {
for _, step := range tr.Status.Steps {
if step.Waiting != nil {
if _, found := podFailureReasons[step.Waiting.Reason]; found {
if step.Waiting.Reason == ImagePullBackOff {
imagePullBackOffTimeOut := config.FromContextOrDefaults(ctx).Defaults.DefaultImagePullBackOffTimeout
// only attempt to recover from the imagePullBackOff if specified
if imagePullBackOffTimeOut.Seconds() != 0 {
p, err := c.KubeClientSet.CoreV1().Pods(tr.Namespace).Get(ctx, tr.Status.PodName, metav1.GetOptions{})
if err != nil {
message := fmt.Sprintf(`The step %q in TaskRun %q failed to pull the image %q and the pod with error: "%s."`, step.Name, tr.Name, step.ImageID, err)
return true, v1.TaskRunReasonImagePullFailed, message
}
for _, condition := range p.Status.Conditions {
// check the pod condition to get the time when the pod was scheduled
// keep trying until the pod schedule time has exceeded the specified imagePullBackOff timeout duration
if condition.Type == corev1.PodScheduled {
if c.Clock.Since(condition.LastTransitionTime.Time) < imagePullBackOffTimeOut {
return false, "", ""
}
}
}
}
}
image := step.ImageID
message := fmt.Sprintf(`The step %q in TaskRun %q failed to pull the image %q. The pod errored with the message: "%s."`, step.Name, tr.Name, image, step.Waiting.Message)
return true, v1.TaskRunReasonImagePullFailed, message
Expand All @@ -234,6 +256,26 @@ func (c *Reconciler) checkPodFailed(tr *v1.TaskRun) (bool, v1.TaskRunReason, str
for _, sidecar := range tr.Status.Sidecars {
if sidecar.Waiting != nil {
if _, found := podFailureReasons[sidecar.Waiting.Reason]; found {
if sidecar.Waiting.Reason == ImagePullBackOff {
imagePullBackOffTimeOut := config.FromContextOrDefaults(ctx).Defaults.DefaultImagePullBackOffTimeout
// only attempt to recover from the imagePullBackOff if specified
if imagePullBackOffTimeOut.Seconds() != 0 {
p, err := c.KubeClientSet.CoreV1().Pods(tr.Namespace).Get(ctx, tr.Status.PodName, metav1.GetOptions{})
if err != nil {
message := fmt.Sprintf(`The sidecar %q in TaskRun %q failed to pull the image %q and the pod with error: "%s."`, sidecar.Name, tr.Name, sidecar.ImageID, err)
return true, v1.TaskRunReasonImagePullFailed, message
}
for _, condition := range p.Status.Conditions {
// check the pod condition to get the time when the pod was scheduled
// keep trying until the pod schedule time has exceeded the specified imagePullBackOff timeout duration
if condition.Type == corev1.PodScheduled {
if c.Clock.Since(condition.LastTransitionTime.Time) < imagePullBackOffTimeOut {
return false, "", ""
}
}
}
}
}
image := sidecar.ImageID
message := fmt.Sprintf(`The sidecar %q in TaskRun %q failed to pull the image %q. The pod errored with the message: "%s."`, sidecar.Name, tr.Name, image, sidecar.Waiting.Message)
return true, v1.TaskRunReasonImagePullFailed, message
Expand Down
Loading

0 comments on commit a40423b

Please sign in to comment.