Skip to content

study: run the AIOpsLab audit against a proposed fix that records the deployed workloads - #27

Merged
TheDeadcoder merged 2 commits into
mainfrom
study/aiopslab-fix
Oct 16, 2026
Merged

TheDeadcoder merged 2 commits into
mainfrom
study/aiopslab-fix

Conversation

@TheDeadcoder

Copy link
Copy Markdown
Owner

Adds workload_check.patch, a proposed fix for AIOpsLab's mitigation verifiers, and lets the audit run against it.

  • With the patch, a pod without container statuses counts as unhealthy unless it has succeeded.
  • AIOpsLab records the application's Deployments and StatefulSets after deploy. eval() then fails if one of them
    no longer exists or is scaled to zero.
  • The four problems with their own pod checks use the shared one. The Kafka problem also requires its feature flag
    to be off.
  • The runner takes --runs, so the patched runs go to runs-patched/ with their own ledger. It stops when an
    episode saves no result.
  • The README has the predictions for the patched runs, fixed before they ran.

The patch makes the same changes as the upstream proposal, microsoft/AIOpsLab#209. The results come in a separate PR.

Sign up for free to 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.

1 participant