-
Notifications
You must be signed in to change notification settings - Fork 62
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Stress Tests MCAD using KWOK #469
base: main
Are you sure you want to change the base?
Conversation
Thanks for the PR @vishakha-ramani ! @z103cb it looks like the Travis build is failing, can you point us to any of the Travis CI output? My view only shows limited information on why it failed. |
@asm582 Can you review this PR? Also, does the MCAD repo follow the practices of squashing commits before a merge into main? |
The build failed in quota management which is not related to this PR, I have restarted the build |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for the PR, most scripts use the same base template and run different kinds of experiments, can we create a single script for running different kinds of experiments? Also, I have provided a few comments on the markdown file.
test/stress-tests-kwok/gpu-tests.md
Outdated
@@ -0,0 +1,160 @@ | |||
## MCAD GPU Request Performance test with KWOK | |||
This experiment assumes that you have a KWOK controller as well as MCAD controller running inside a kind cluster. If not, follow [this](https://github.com/vishakha-ramani/multi-cluster-app-dispatcher/blob/main/test/perf-test/simulatingnodesandappwrappers.md) for installation instructions. | |||
The MCAD service for gpu requests is little weird . Here is the experiment I did and what I observed: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
for gpu requests is little weird
can we change the wording and give the user a little more insight into the use case we ought to test, please?
test/stress-tests-kwok/gpu-tests.md
Outdated
... | ||
``` | ||
|
||
7. The pod is scheduled to one of the fake nodes (which theoretically it shouldn't). |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Curious, did we ever print a histogram inside MCAD and see how it looks? also, can we mention which version of MCAD was used for testing?
@Sara-KS, |
@z103cb Thank you for the feedback! We can plan on integrating this work into the build. |
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
values: | ||
- kwok | ||
# A taints was added to an automatically created Node. | ||
# You can remove taints of Node or add this tolerations. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
# You can remove taints of Node or add this tolerations. | |
# You can remove the taint on the node or add this toleration. |
- A single taint was added to the node. A taint contains a list, but only one is specified here.
- A single toleration is added to the job. A toleration contains a list, but only one is specified here.
operator: In | ||
values: | ||
- kwok | ||
# A taints was added to an automatically created Node. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
# A taints was added to an automatically created Node. | |
# A taint was added automatically to the created Node. |
Though the node was created automatically in the shell script at Line 21, the taint was also added automatically.
``` | ||
For example, for this job spec, the timestamps captured are `creationTimestamp: "2023-08-03T13:07:27Z"` and `completionTime: "2023-08-03T13:09:52Z"`. The job completion time is then just `completionTime - creationTimestamp`. | ||
|
||
3. Dispatch Time: The interval between AppWrapper creation time and dispatch time |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
3. Dispatch Time: The interval between AppWrapper creation time and dispatch time | |
3. Dispatch Time: The interval between AppWrapper creation time and dispatch time. |
I added a period to show that the sentence has ended.
Dispatch performance of MCAD with many small jobs
The following experiment was performed to understand the overhead associated with MCAD, if any, when large number of small AW jobs, consisting ony of 1 pod, are inserted into a system. To do that, we use KWOK toolkit that enables us to simulate the lifecycle of fake nodes, pods and other Kubernetes API resources.