cert-manager/pkg/controller/certificates/trigger/trigger_controller.go
Tim Ramlot 085136068a
fix misspell linter
Signed-off-by: Tim Ramlot <42113979+inteon@users.noreply.github.com>
2024-04-29 15:21:07 +02:00

372 lines
16 KiB
Go

/*
Copyright 2020 The cert-manager 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
http://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.
*/
package trigger
import (
"context"
"fmt"
"math"
"time"
"github.com/go-logr/logr"
corev1 "k8s.io/api/core/v1"
apierrors "k8s.io/apimachinery/pkg/api/errors"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/labels"
"k8s.io/client-go/tools/cache"
"k8s.io/client-go/tools/record"
"k8s.io/client-go/util/workqueue"
"k8s.io/utils/clock"
internalcertificates "github.com/cert-manager/cert-manager/internal/controller/certificates"
"github.com/cert-manager/cert-manager/internal/controller/certificates/policies"
"github.com/cert-manager/cert-manager/internal/controller/feature"
internalinformers "github.com/cert-manager/cert-manager/internal/informers"
apiutil "github.com/cert-manager/cert-manager/pkg/api/util"
cmapi "github.com/cert-manager/cert-manager/pkg/apis/certmanager/v1"
cmmeta "github.com/cert-manager/cert-manager/pkg/apis/meta/v1"
cmclient "github.com/cert-manager/cert-manager/pkg/client/clientset/versioned"
cmlisters "github.com/cert-manager/cert-manager/pkg/client/listers/certmanager/v1"
controllerpkg "github.com/cert-manager/cert-manager/pkg/controller"
"github.com/cert-manager/cert-manager/pkg/controller/certificates"
logf "github.com/cert-manager/cert-manager/pkg/logs"
"github.com/cert-manager/cert-manager/pkg/scheduler"
utilfeature "github.com/cert-manager/cert-manager/pkg/util/feature"
"github.com/cert-manager/cert-manager/pkg/util/pki"
"github.com/cert-manager/cert-manager/pkg/util/predicate"
)
const (
ControllerName = "certificates-trigger"
// stopIncreaseBackoff is the number of issuance attempts after which the backoff period should stop to increase
stopIncreaseBackoff = 6 // 2 ^ (6 - 1) = 32 = maxDelay
// maxDelay is the maximum backoff period
maxDelay = 32 * time.Hour
)
// This controller observes the state of the certificate's currently
// issued `spec.secretName` and the rest of the `certificate.spec` fields to
// determine whether a re-issuance is required.
// It triggers re-issuance by adding the `Issuing` status condition when a new
// certificate is required.
type controller struct {
certificateLister cmlisters.CertificateLister
certificateRequestLister cmlisters.CertificateRequestLister
secretLister internalinformers.SecretLister
client cmclient.Interface
recorder record.EventRecorder
scheduledWorkQueue scheduler.ScheduledWorkQueue
// fieldManager is the string which will be used as the Field Manager on
// fields created or edited by the cert-manager Kubernetes client during
// Apply API calls.
fieldManager string
// The following are used for testing purposes.
clock clock.Clock
shouldReissue policies.Func
dataForCertificate func(context.Context, *cmapi.Certificate) (policies.Input, error)
}
func NewController(
log logr.Logger,
ctx *controllerpkg.Context,
shouldReissue policies.Func,
) (*controller, workqueue.RateLimitingInterface, []cache.InformerSynced) {
// create a queue used to queue up items to be processed
queue := workqueue.NewNamedRateLimitingQueue(workqueue.NewItemExponentialFailureRateLimiter(time.Second*1, time.Second*30), ControllerName)
// obtain references to all the informers used by this controller
certificateInformer := ctx.SharedInformerFactory.Certmanager().V1().Certificates()
certificateRequestInformer := ctx.SharedInformerFactory.Certmanager().V1().CertificateRequests()
secretsInformer := ctx.KubeSharedInformerFactory.Secrets()
certificateInformer.Informer().AddEventHandler(&controllerpkg.QueuingEventHandler{Queue: queue})
// When a CertificateRequest resource changes, enqueue the Certificate resource that owns it.
certificateRequestInformer.Informer().AddEventHandler(&controllerpkg.BlockingEventHandler{
WorkFunc: certificates.EnqueueCertificatesForResourceUsingPredicates(log, queue, certificateInformer.Lister(), labels.Everything(), predicate.ResourceOwnerOf),
})
// When a Secret resource changes, enqueue any Certificate resources that name it as spec.secretName.
secretsInformer.Informer().AddEventHandler(&controllerpkg.BlockingEventHandler{
// Trigger reconciles on changes to the Secret named `spec.secretName`
WorkFunc: certificates.EnqueueCertificatesForResourceUsingPredicates(log, queue, certificateInformer.Lister(), labels.Everything(),
predicate.ExtractResourceName(predicate.CertificateSecretName)),
})
// build a list of InformerSynced functions that will be returned by the Register method.
// the controller will only begin processing items once all of these informers have synced.
mustSync := []cache.InformerSynced{
certificateRequestInformer.Informer().HasSynced,
secretsInformer.Informer().HasSynced,
certificateInformer.Informer().HasSynced,
}
return &controller{
certificateLister: certificateInformer.Lister(),
certificateRequestLister: certificateRequestInformer.Lister(),
secretLister: secretsInformer.Lister(),
client: ctx.CMClient,
recorder: ctx.Recorder,
scheduledWorkQueue: scheduler.NewScheduledWorkQueue(ctx.Clock, queue.Add),
fieldManager: ctx.FieldManager,
// The following are used for testing purposes.
clock: ctx.Clock,
shouldReissue: shouldReissue,
dataForCertificate: (&policies.Gatherer{
CertificateRequestLister: certificateRequestInformer.Lister(),
SecretLister: secretsInformer.Lister(),
}).DataForCertificate,
}, queue, mustSync
}
func (c *controller) ProcessItem(ctx context.Context, key string) error {
log := logf.FromContext(ctx).WithValues("key", key)
ctx = logf.NewContext(ctx, log)
namespace, name, err := cache.SplitMetaNamespaceKey(key)
if err != nil {
log.Error(err, "invalid resource key passed to ProcessItem")
return nil
}
crt, err := c.certificateLister.Certificates(namespace).Get(name)
if apierrors.IsNotFound(err) {
log.V(logf.DebugLevel).Info("certificate not found for key", "error", err.Error())
return nil
}
if err != nil {
return err
}
if apiutil.CertificateHasCondition(crt, cmapi.CertificateCondition{
Type: cmapi.CertificateConditionIssuing,
Status: cmmeta.ConditionTrue,
}) {
// Do nothing if an issuance is already in progress.
return nil
}
// It is possible for multiple Certificates to reference the same Secret. In that case, without this check,
// the duplicate Certificates would each be issued and store their version of the X.509 certificate in the
// target Secret, triggering the re-issuance of the other Certificate resources who's spec no longer matches
// what is in the Secret. This would cause a flood of re-issuance attempts and overloads the Kubernetes API
// and the API server of the issuing CA.
isOwner, duplicates, err := internalcertificates.CertificateOwnsSecret(ctx, c.certificateLister, c.secretLister, crt)
if err != nil {
return err
}
if !isOwner {
log.V(logf.DebugLevel).Info("Certificate.Spec.SecretName refers to the same Secret as other Certificates in the same namespace, skipping trigger.", "duplicates", duplicates)
// If the Certificate is not the owner of the Secret, we requeue the Certificate and wait for the
// Certificate to become the owner of the Secret. This can happen if the Certificate is updated to
// reference a different Secret, or if the conflicting Certificate is deleted or updated to no longer
// reference the Secret.
c.scheduledWorkQueue.Add(key, 3*time.Minute)
return nil
}
input, err := c.dataForCertificate(ctx, crt)
if err != nil {
return err
}
// Don't trigger issuance if we need to back off due to previous failures and Certificate's spec has not changed.
backoff, delay := shouldBackoffReissuingOnFailure(log, c.clock, input.Certificate, input.NextRevisionRequest)
if backoff {
nextIssuanceRetry := c.clock.Now().Add(delay)
message := fmt.Sprintf("Backing off from issuance due to previously failed issuance(s). Issuance will next be attempted at %v", nextIssuanceRetry)
log.V(logf.InfoLevel).Info(message)
c.scheduleRecheckOfCertificateIfRequired(log, key, delay)
return nil
}
if crt.Status.RenewalTime != nil {
// ensure a resync is scheduled in the future so that we re-check
// Certificate resources and trigger them near expiry time
c.scheduleRecheckOfCertificateIfRequired(log, key, crt.Status.RenewalTime.Time.Sub(c.clock.Now()))
}
reason, message, reissue := c.shouldReissue(input)
if !reissue {
// no re-issuance required, return early
return nil
}
// Although the below recorder.Event already logs the event, the log
// line is quite unreadable (very long). Since this information is very
// important for the user and the operator, we log the following
// message.
log.V(logf.InfoLevel).Info("Certificate must be re-issued", "reason", reason, "message", message)
crt = crt.DeepCopy()
apiutil.SetCertificateCondition(crt, crt.Generation, cmapi.CertificateConditionIssuing, cmmeta.ConditionTrue, reason, message)
if err := c.updateOrApplyStatus(ctx, crt); err != nil {
return err
}
c.recorder.Event(crt, corev1.EventTypeNormal, "Issuing", message)
return nil
}
// updateOrApplyStatus will update the controller status. If the
// ServerSideApply feature is enabled, the managed fields will instead get
// applied using the relevant Patch API call.
func (c *controller) updateOrApplyStatus(ctx context.Context, crt *cmapi.Certificate) error {
if utilfeature.DefaultFeatureGate.Enabled(feature.ServerSideApply) {
var conditions []cmapi.CertificateCondition
if cond := apiutil.GetCertificateCondition(crt, cmapi.CertificateConditionIssuing); cond != nil {
conditions = []cmapi.CertificateCondition{*cond}
}
return internalcertificates.ApplyStatus(ctx, c.client, c.fieldManager, &cmapi.Certificate{
ObjectMeta: metav1.ObjectMeta{Namespace: crt.Namespace, Name: crt.Name},
Status: cmapi.CertificateStatus{Conditions: conditions},
})
} else {
_, err := c.client.CertmanagerV1().Certificates(crt.Namespace).UpdateStatus(ctx, crt, metav1.UpdateOptions{})
return err
}
}
// shouldBackOffReissuingOnFailure returns true if an issuance needs to be
// delayed and the required delay after calculating the exponential backoff.
// The backoff periods are 1h, 2h, 4h, 8h, 16h and 32h counting from when the last
// failure occurred,
// so the returned delay will be backoff_period - (current_time - last_failure_time)
//
// Notably, it returns no back-off when the certificate doesn't
// match the "next" certificate (since a mismatch means that this certificate
// gets re-issued immediately).
//
// Note that the request can be left nil: in that case, the returned back-off
// will be 0 since it means the CR must be created immediately.
func shouldBackoffReissuingOnFailure(log logr.Logger, c clock.Clock, crt *cmapi.Certificate, nextCR *cmapi.CertificateRequest) (bool, time.Duration) {
if crt.Status.LastFailureTime == nil {
return false, 0
}
// We want to immediately trigger a re-issuance when the certificate
// changes. In order to detect a "change", we compare the "next" CR with the
// certificate spec and reissue if there is a mismatch. To understand this
// mechanism, take a look at the diagram of the scenario C at the top of the
// gatherer.go file.
//
// Note that the "next" CR is the only CR that matters when looking at
// whether the certificate still matches its CR. The "current" CR matches
// the previous spec of the certificate, so we don't want to be looking at
// the current CR.
if nextCR == nil {
log.V(logf.InfoLevel).Info("next CertificateRequest not available, skipping checking if Certificate matches the CertificateRequest")
} else {
mismatches, err := pki.RequestMatchesSpec(nextCR, crt.Spec)
if err != nil {
log.V(logf.InfoLevel).Info("next CertificateRequest cannot be decoded, skipping checking if Certificate matches the CertificateRequest")
return false, 0
}
if len(mismatches) > 0 {
log.V(logf.ExtendedInfoLevel).WithValues("mismatches", mismatches).Info("Certificate is failing but the Certificate differs from CertificateRequest, backoff is not required")
return false, 0
}
}
now := c.Now()
durationSinceFailure := now.Sub(crt.Status.LastFailureTime.Time)
initialDelay := time.Hour
delay := initialDelay
failedIssuanceAttempts := 0
// It is possible that crt.Status.LastFailureTime != nil &&
// crt.Status.FailedIssuanceAttempts == nil (in case of the Certificate having
// failed for an installation of cert-manager before the issuance
// attempts were introduced). In such case delay = initialDelay.
if crt.Status.FailedIssuanceAttempts != nil {
failedIssuanceAttempts = *crt.Status.FailedIssuanceAttempts
delay = time.Hour * time.Duration(math.Pow(2, float64(failedIssuanceAttempts-1)))
}
// Ensure that maximum returned delay is 32 hours
// delay cannot be calculated for large issuance numbers, so we
// cannot reliably check if delay > maxDelay directly
// (see i.e the result of time.Duration(math.Pow(2, 99)))
if failedIssuanceAttempts > stopIncreaseBackoff {
delay = maxDelay
}
// Ensure that minimum returned delay is 1 hour. This is here to guard
// against an edge case where the delay duration got messed
// up as a result of maths misuse in the previous calculations
if delay < initialDelay {
delay = initialDelay
}
if durationSinceFailure >= delay {
log.V(logf.ExtendedInfoLevel).WithValues("since_failure", durationSinceFailure).Info("Certificate has been in failure state long enough, no need to back off")
return false, 0
}
return true, delay - durationSinceFailure
}
// scheduleRecheckOfCertificateIfRequired will schedule the resource with the
// given key to be re-queued for processing after the given amount of time
// has elapsed.
// If the 'durationUntilRenewalTime' is less than zero, it will not be
// queued again.
func (c *controller) scheduleRecheckOfCertificateIfRequired(log logr.Logger, key string, durationUntilRenewalTime time.Duration) {
// don't schedule a re-queue if the time is in the past.
// if it is in the past, the resource will be triggered during the
// current call to the ProcessItem method. If we added the item to the
// queue with a duration of <=0, we would otherwise continually re-queue
// in a tight loop whilst we wait for the caching listers to observe
// the 'Triggered' status condition changing to 'True'.
if durationUntilRenewalTime < 0 {
return
}
log.V(logf.DebugLevel).Info("scheduling renewal", "duration_until_renewal", durationUntilRenewalTime.String())
c.scheduledWorkQueue.Add(key, durationUntilRenewalTime)
}
// controllerWrapper wraps the `controller` structure to make it implement
// the controllerpkg.queueingController interface
type controllerWrapper struct {
*controller
}
func (c *controllerWrapper) Register(ctx *controllerpkg.Context) (workqueue.RateLimitingInterface, []cache.InformerSynced, error) {
// construct a new named logger to be reused throughout the controller
log := logf.FromContext(ctx.RootContext, ControllerName)
ctrl, queue, mustSync := NewController(log,
ctx,
policies.NewTriggerPolicyChain(ctx.Clock).Evaluate,
)
c.controller = ctrl
return queue, mustSync, nil
}
func init() {
controllerpkg.Register(ControllerName, func(ctx *controllerpkg.ContextFactory) (controllerpkg.Interface, error) {
return controllerpkg.NewBuilder(ctx, ControllerName).
For(&controllerWrapper{}).
Complete()
})
}