Skip to content

K
k3s
Commit 5f2846ee authored by Kubernetes Submit Queue's avatar Kubernetes Submit Queue Committed by GitHub
Browse files
Options

Merge pull request #31973 from soltysh/job_e2e_slowness

Automatic merge from submit-queue Speed up job's e2e when waiting for failure **What this PR does / why we need it**: Job controller synchronizes objects only when job itself or underlying pod changes. Or, when full resync is performed once 10 mins. This leads e2e test to unnecessarily wait that longer timeout, sometimes at least. I've added job modification action which triggers resync, if the job wasn't terminated within shorter period of time. @ixdy ptal @janetkuo @erictune fyi
parents 0cd24992 393d5513
Hide whitespace changes
Inline Side-by-side
Showing with 49 additions and 11 deletions
test/e2e/batch_v1_jobs.go
View file @ 5f2846ee
......@@ -108,7 +108,7 @@ var _ = framework.KubeDescribe("V1Job", func() {
By("Ensuring job shows many failures")
err = wait.Poll(framework.Poll, v1JobTimeout, func() (bool, error) {
curr, err := f.Client.Batch().Jobs(f.Namespace.Name).Get(job.Name)
curr, err := getV1Job(f.Client, f.Namespace.Name, job.Name)
if err != nil {
return false, err
}
......@@ -184,12 +184,12 @@ var _ = framework.KubeDescribe("V1Job", func() {
Expect(err).NotTo(HaveOccurred())
By("Ensuring job was deleted")
_, err = f.Client.Batch().Jobs(f.Namespace.Name).Get(job.Name)
_, err = getV1Job(f.Client, f.Namespace.Name, job.Name)
Expect(err).To(HaveOccurred())
Expect(errors.IsNotFound(err)).To(BeTrue())
})
It("should fail a job [Slow]", func() {
It("should fail a job", func() {
By("Creating a job")
job := newTestV1Job("notTerminate", "foo", api.RestartPolicyNever, parallelism, completions)
activeDeadlineSeconds := int64(10)
......@@ -198,7 +198,18 @@ var _ = framework.KubeDescribe("V1Job", func() {
Expect(err).NotTo(HaveOccurred())
By("Ensuring job was failed")
err = waitForV1JobFail(f.Client, f.Namespace.Name, job.Name)
err = waitForV1JobFail(f.Client, f.Namespace.Name, job.Name, 20*time.Second)
if err == wait.ErrWaitTimeout {
job, err = getV1Job(f.Client, f.Namespace.Name, job.Name)
Expect(err).NotTo(HaveOccurred())
// the job stabilized and won't be synced until modification or full
// resync happens, we don't want to wait for the latter so we force
// sync modifying it
job.Spec.Parallelism = &completions
job, err = updateV1Job(f.Client, f.Namespace.Name, job)
Expect(err).NotTo(HaveOccurred())
err = waitForV1JobFail(f.Client, f.Namespace.Name, job.Name, v1JobTimeout)
}
Expect(err).NotTo(HaveOccurred())
})
})
......@@ -265,10 +276,18 @@ func newTestV1Job(behavior, name string, rPol api.RestartPolicy, parallelism, co
return job
}
func getV1Job(c *client.Client, ns, name string) (*batch.Job, error) {
return c.Batch().Jobs(ns).Get(name)
}
func createV1Job(c *client.Client, ns string, job *batch.Job) (*batch.Job, error) {
return c.Batch().Jobs(ns).Create(job)
}
func updateV1Job(c *client.Client, ns string, job *batch.Job) (*batch.Job, error) {
return c.Batch().Jobs(ns).Update(job)
}
func deleteV1Job(c *client.Client, ns, name string) error {
return c.Batch().Jobs(ns).Delete(name, api.NewDeleteOptions(0))
}
......@@ -304,8 +323,8 @@ func waitForV1JobFinish(c *client.Client, ns, jobName string, completions int32)
}
// Wait for job fail.
func waitForV1JobFail(c *client.Client, ns, jobName string) error {
return wait.Poll(framework.Poll, v1JobTimeout, func() (bool, error) {
func waitForV1JobFail(c *client.Client, ns, jobName string, timeout time.Duration) error {
return wait.Poll(framework.Poll, timeout, func() (bool, error) {
curr, err := c.Batch().Jobs(ns).Get(jobName)
if err != nil {
return false, err
......
test/e2e/job.go
View file @ 5f2846ee
......@@ -104,7 +104,7 @@ var _ = framework.KubeDescribe("Job", func() {
By("Ensuring job shows many failures")
err = wait.Poll(framework.Poll, jobTimeout, func() (bool, error) {
curr, err := f.Client.Extensions().Jobs(f.Namespace.Name).Get(job.Name)
curr, err := getJob(f.Client, f.Namespace.Name, job.Name)
if err != nil {
return false, err
}
......@@ -180,7 +180,7 @@ var _ = framework.KubeDescribe("Job", func() {
Expect(err).NotTo(HaveOccurred())
By("Ensuring job was deleted")
_, err = f.Client.Extensions().Jobs(f.Namespace.Name).Get(job.Name)
_, err = getJob(f.Client, f.Namespace.Name, job.Name)
Expect(err).To(HaveOccurred())
Expect(errors.IsNotFound(err)).To(BeTrue())
})
......@@ -194,7 +194,18 @@ var _ = framework.KubeDescribe("Job", func() {
Expect(err).NotTo(HaveOccurred())
By("Ensuring job was failed")
err = waitForJobFail(f.Client, f.Namespace.Name, job.Name)
err = waitForJobFail(f.Client, f.Namespace.Name, job.Name, 20*time.Second)
if err == wait.ErrWaitTimeout {
job, err = getJob(f.Client, f.Namespace.Name, job.Name)
Expect(err).NotTo(HaveOccurred())
// the job stabilized and won't be synced until modification or full
// resync happens, we don't want to wait for the latter so we force
// sync modifying it
job.Spec.Parallelism = &completions
job, err = updateJob(f.Client, f.Namespace.Name, job)
Expect(err).NotTo(HaveOccurred())
err = waitForJobFail(f.Client, f.Namespace.Name, job.Name, jobTimeout)
}
Expect(err).NotTo(HaveOccurred())
})
})
......@@ -262,10 +273,18 @@ func newTestJob(behavior, name string, rPol api.RestartPolicy, parallelism, comp
return job
}
func getJob(c *client.Client, ns, name string) (*batch.Job, error) {
return c.Extensions().Jobs(ns).Get(name)
}
func createJob(c *client.Client, ns string, job *batch.Job) (*batch.Job, error) {
return c.Extensions().Jobs(ns).Create(job)
}
func updateJob(c *client.Client, ns string, job *batch.Job) (*batch.Job, error) {
return c.Extensions().Jobs(ns).Update(job)
}
func deleteJob(c *client.Client, ns, name string) error {
return c.Extensions().Jobs(ns).Delete(name, nil)
}
......@@ -301,8 +320,8 @@ func waitForJobFinish(c *client.Client, ns, jobName string, completions int32) e
}
// Wait for job fail.
func waitForJobFail(c *client.Client, ns, jobName string) error {
return wait.Poll(framework.Poll, jobTimeout, func() (bool, error) {
func waitForJobFail(c *client.Client, ns, jobName string, timeout time.Duration) error {
return wait.Poll(framework.Poll, timeout, func() (bool, error) {
curr, err := c.Extensions().Jobs(ns).Get(jobName)
if err != nil {
return false, err
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment