[SDK] Get Kubernetes Events for Job #1975

andreyvelich · 2024-01-05T23:23:52Z

Fixes: #1863.

This will allow users to get Kubernetes Events for Job and Job's pods via get_job_logs API and verbose parameter.
Initially, I didn't watch for the events similar to logs with follow=True parameter. This can be done later in the following PRs.

The events will be returned in this format:

{
  "test-kubeflow-worker-0": [
    "2024-01-05 22:58:20 Successfully assigned kubeflow-andrey/test-kubeflow-worker-0 to .....",
    "2024-01-05 22:58:21 Container image \"docker.io/pytorch/pytorch:1.12.1-cuda11.3-cudnn8-runtime\" already present on machine",
    "2024-01-05 22:58:21 Created container pytorch",
    "2024-01-05 22:58:21 Started container pytorch"
  ],
  "test-kubeflow-worker-1": [
    "2024-01-05 22:58:20 Successfully assigned kubeflow-andrey/test-kubeflow-worker-1 to ....",
    "2024-01-05 22:58:21 Container image \"docker.io/pytorch/pytorch:1.12.1-cuda11.3-cudnn8-runtime\" already present on machine",
    "2024-01-05 22:58:21 Created container pytorch",
    "2024-01-05 22:58:21 Started container pytorch"
  ],
  "test-kubeflow": [
    "2024-01-05 22:58:20 Created pod: test-kubeflow-worker-0",
    "2024-01-05 22:58:20 Created pod: test-kubeflow-worker-1",
    "2024-01-05 22:58:20 Created service: test-kubeflow-worker-0",
    "2024-01-05 22:58:20 Created service: test-kubeflow-worker-1",
    "2024-01-05 22:59:11 Pod: kubeflow-andrey.test-kubeflow-worker-1 exited with code 0",
    "2024-01-05 22:59:11 Pod: kubeflow-andrey.test-kubeflow-worker-0 exited with code 0",
    "2024-01-05 22:59:13 PyTorchJob kubeflow-andrey/test-kubeflow successfully completed."
  ]
}

I introduce a new get_job_pods API to get Job's pods data.
Also, I removed support for Python3 from our SDK. I believe, various Python typing capabilities wasn't there in Python 3.

/assign @droctothorpe @deepanker13 @johnugeorge @tenzen-y

google-oss-prow · 2024-01-05T23:23:55Z

@andreyvelich: GitHub didn't allow me to assign the following users: droctothorpe, deepanker13.

Note that only kubeflow members with read permissions, repo collaborators and people who have commented on this issue/PR can be assigned. Additionally, issues/PRs can only have 10 assignees at the same time.
For more information please see the contributor guide

In response to this:

Fixes: #1863.

This will allow users to get Kubernetes Events for Job and Job's pods via get_job_logs API and verbose parameter.
Initially, I didn't watch for the events similar to logs with follow=True parameter. This can be done later in the following PRs.

The events will be returned in this format:

{
 "test-kubeflow-worker-0": [
   "2024-01-05 22:58:20 Successfully assigned kubeflow-andrey/test-kubeflow-worker-0 to .....",
   "2024-01-05 22:58:21 Container image \"docker.io/pytorch/pytorch:1.12.1-cuda11.3-cudnn8-runtime\" already present on machine",
   "2024-01-05 22:58:21 Created container pytorch",
   "2024-01-05 22:58:21 Started container pytorch"
 ],
 "test-kubeflow-worker-1": [
   "2024-01-05 22:58:20 Successfully assigned kubeflow-andrey/test-kubeflow-worker-1 to ....",
   "2024-01-05 22:58:21 Container image \"docker.io/pytorch/pytorch:1.12.1-cuda11.3-cudnn8-runtime\" already present on machine",
   "2024-01-05 22:58:21 Created container pytorch",
   "2024-01-05 22:58:21 Started container pytorch"
 ],
 "test-kubeflow": [
   "2024-01-05 22:58:20 Created pod: test-kubeflow-worker-0",
   "2024-01-05 22:58:20 Created pod: test-kubeflow-worker-1",
   "2024-01-05 22:58:20 Created service: test-kubeflow-worker-0",
   "2024-01-05 22:58:20 Created service: test-kubeflow-worker-1",
   "2024-01-05 22:59:11 Pod: kubeflow-andrey.test-kubeflow-worker-1 exited with code 0",
   "2024-01-05 22:59:11 Pod: kubeflow-andrey.test-kubeflow-worker-0 exited with code 0",
   "2024-01-05 22:59:13 PyTorchJob kubeflow-andrey/test-kubeflow successfully completed."
 ]
}

I introduce a new get_job_pods API to get Job's pods data.
Also, I removed support for Python3 from our SDK. I believe, various Python typing capabilities wasn't there in Python 3.

/assign @droctothorpe @deepanker13 @johnugeorge @tenzen-y

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

google-oss-prow · 2024-01-05T23:24:06Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: andreyvelich

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

~~OWNERS~~ [andreyvelich]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

coveralls · 2024-01-05T23:26:59Z

Pull Request Test Coverage Report for Build 7491318987

0 of 0 changed or added relevant lines in 0 files are covered.
No unchanged relevant lines lost coverage.
Overall first build on issue-1863-sdk-job-events at 42.942%

Totals
Change from base Build 7478335417:	42.9%
Covered Lines:	3760
Relevant Lines:	8756

💛 - Coveralls

johnugeorge · 2024-01-09T19:11:12Z

sdk/python/kubeflow/training/api/training_client.py

+            events = self.core_api.list_namespaced_event(namespace=namespace)
+
+            # Get events for the Job and Job's pods.
+            for event in events.items:


How do we distinguish between job and pod events? Should we add a prefix?

Right now, users can use names to get events for Job or Pods.
E.g. for PyTorchJob with name train-mnist and 2 workers:

{"train-mnist": "Events", "train-mnist-worker-0": "Events", "train-mnist-worker-1": "Events"}

Users can understand the all pod names by running TrainingClient().get_pod_names()

Do you think we should add the prefix that identify type of the object ?
e.g.

{"pytorchjob-train-mnist": "Events", "pod-train-mnist-worker-0": "Events", "pod-train-mnist-worker-1": "Events"}

cc @droctothorpe @tenzen-y

sdk/python/kubeflow/training/api/training_client.py

droctothorpe · 2024-01-09T22:59:26Z

sdk/python/kubeflow/training/api/training_client.py

+                    and pod.status.phase != constants.POD_PHASE_PENDING
+                ):
+                    log_streams.append(
+                        watch.Watch().stream(


Neat! Didn't know that the Python client supported this.

droctothorpe · 2024-01-09T23:00:40Z

sdk/python/kubeflow/training/api/training_client.py

@@ -861,27 +924,58 @@ def get_job_logs(
                                break

                            # Print logs to the StdOut
-                            print(f"[Pod {pods[index]}]: {logline}")
+                            print(f"[Pod {pods[index].metadata.name}]: {logline}")


Thoughts on colorizing by colorizing by pod name? Probably overkill / can be a separate PR.

@droctothorpe Can you elaborate here ? What are your ideas on printing logs for pods?

Here is an example of output from https://github.com/stern/stern.

Each pod gets a distinct color, which helps parse a giant wall of interleaved logs. Happy to tackle it in a discrete PR.

I see, thanks for sharing.
Can you create an issue to discuss this ?
We might need to discuss how various output will produce the results while running this (e.g. Jupyter Notebook, VSCode, local terminal).

andreyvelich · 2024-01-11T14:31:11Z

@droctothorpe @johnugeorge I added object kind to the event message for now, so it is easier to parse data from the event messaged:

{
  "pytorchjob test-kubeflow": [
    "2024-01-05 22:58:20 Created pod: test-kubeflow-worker-0"
  ],
  "pod test-kubeflow-worker-0": [
    "2024-01-05 22:58:20 Successfully assigned kubeflow-andrey/test-kubeflow-worker-1 to ...."
  ],
 }

johnugeorge · 2024-01-11T16:29:27Z

/lgtm
Thanks @andreyvelich

google-oss-prow bot assigned johnugeorge Jan 5, 2024

google-oss-prow bot assigned tenzen-y Jan 5, 2024

google-oss-prow bot requested review from jinchihe and kuizhiqing January 5, 2024 23:24

google-oss-prow bot added approved size/L labels Jan 5, 2024

johnugeorge reviewed Jan 9, 2024

View reviewed changes

droctothorpe reviewed Jan 9, 2024

View reviewed changes

sdk/python/kubeflow/training/api/training_client.py Show resolved Hide resolved

droctothorpe reviewed Jan 9, 2024

View reviewed changes

andreyvelich added 2 commits January 10, 2024 20:25

[SDK] Get Kubernetes Events for Job

a04756b

Move class at the top

54b6269

andreyvelich force-pushed the issue-1863-sdk-job-events branch from e35a25b to 54b6269 Compare January 10, 2024 20:28

Add object kind to event message

b2d9c36

Add example for events

cbaf066

google-oss-prow bot added the lgtm label Jan 11, 2024

google-oss-prow bot merged commit 778f555 into kubeflow:master Jan 11, 2024
31 checks passed

andreyvelich deleted the issue-1863-sdk-job-events branch January 11, 2024 16:42

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[SDK] Get Kubernetes Events for Job #1975

[SDK] Get Kubernetes Events for Job #1975

andreyvelich commented Jan 5, 2024

google-oss-prow bot commented Jan 5, 2024

google-oss-prow bot commented Jan 5, 2024

coveralls commented Jan 5, 2024 •

edited

Loading

johnugeorge Jan 9, 2024

andreyvelich Jan 10, 2024 •

edited

Loading

droctothorpe Jan 9, 2024

droctothorpe Jan 9, 2024

andreyvelich Jan 10, 2024

droctothorpe Jan 10, 2024

andreyvelich Jan 10, 2024

droctothorpe Jan 10, 2024

andreyvelich commented Jan 11, 2024

johnugeorge commented Jan 11, 2024

[SDK] Get Kubernetes Events for Job #1975

[SDK] Get Kubernetes Events for Job #1975

Conversation

andreyvelich commented Jan 5, 2024

google-oss-prow bot commented Jan 5, 2024

google-oss-prow bot commented Jan 5, 2024

coveralls commented Jan 5, 2024 • edited Loading

Pull Request Test Coverage Report for Build 7491318987

💛 - Coveralls

johnugeorge Jan 9, 2024

Choose a reason for hiding this comment

andreyvelich Jan 10, 2024 • edited Loading

Choose a reason for hiding this comment

droctothorpe Jan 9, 2024

Choose a reason for hiding this comment

droctothorpe Jan 9, 2024

Choose a reason for hiding this comment

andreyvelich Jan 10, 2024

Choose a reason for hiding this comment

droctothorpe Jan 10, 2024

Choose a reason for hiding this comment

andreyvelich Jan 10, 2024

Choose a reason for hiding this comment

droctothorpe Jan 10, 2024

Choose a reason for hiding this comment

andreyvelich commented Jan 11, 2024

johnugeorge commented Jan 11, 2024

coveralls commented Jan 5, 2024 •

edited

Loading

andreyvelich Jan 10, 2024 •

edited

Loading