Skip to content

Move upstream - #39

Closed
Jackson Clark (HacksonClark) wants to merge 141 commits into
microsoft:mainfrom
xlab-uiuc:main
Closed

Move upstream#39
Jackson Clark (HacksonClark) wants to merge 141 commits into
microsoft:mainfrom
xlab-uiuc:main

Conversation

@HacksonClark

Copy link
Copy Markdown
Contributor

No description provided.

Handle traces/metrics file not found errors
Fix the issue of getting log of the workload pod
Fix issue 41 (Fix key error with TTA)
Changed the key "TTR" to "TTA" in aiopslab\orchestrator\tasks\analysis.py
Injected faults can lead to failure of starting the app next time when the program exits before fault recovery.
Add codes to ensure the happening of fault recovery before exit if faults have been injected (exceptions or ctrl+c)
hotfix: Fix revoke_auth localization oracle
hotfix: loadgenerator fault not working
Fix localization oracle for storage_user_unregistered
…remotely; fix vLLM agent parameter assignment; align max_tokens default in SimulationRequest with vLLM agent
Oracle polls pod status for 1 minute to handle k8s restart policy
Add AIOpsLab agent implementations and enhancements

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this might be useful. Do you mind documenting its usecase in a few lines?

print(f"Container {container_status.name} is not ready")
all_normal = False
# Polling for 1 minute to check if all services are back to normal
for _ in range(12): # 5 seconds interval, 12 times, total 1 minute

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we might already be checking pod status somewhere. Would it make more sense to create a function and move it to a module?

Copy link
Copy Markdown
ContributorAuthor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, we have a function for this called wait_for_pod_ready in kubectl.py. I can update it to use that.

payload_script=self.payload_script,
url=f"{frontend_url}",
)
wrk = Wrk(rate=10, dist="exp", connections=2, duration=10, threads=2)

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thinking out loud: If someone wants to change wrk2 parameters, they would have to edit a lot of files. We can add it as a todo for now, but can we think of something that makes it easy to run wrk2 with different parameters? There is also the question of how to determine the 'right' parameters for a problem/fault type.

Copy link
Copy Markdown
ContributorAuthor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sure, we generalized the workload interface in SREArena so that way it's easier to use. We could do something similar here.

@HacksonClark

Copy link
Copy Markdown
ContributorAuthor

I seemed to have done something wrong with git when trying to resolve the merge conflicts. It doesn't let me do it in this PR, so I moved to #52

Sign up for freeto join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

8 participants

@HacksonClark@gaganso@yinfangchen@daklqw@ChuanweiQu@Flemington8@yimingsu01@rMaxiQp