Revert production ML deployments during incidents using Argo Rollouts or Kubernetes.
npx skills add https://github.com/DuqueOM/ML-MLOps-Portfolio --skill rollback-duqueom
Please help me install this Agent Skill. Skill: rollback Source: https://github.com/DuqueOM/ML-MLOps-Portfolio/tree/main/.devin/skills/rollback Command: npx skills add https://github.com/DuqueOM/ML-MLOps-Portfolio --skill rollback-duqueom
Emergency rollback of production ML deployments to a known-good state during incidents, preventing user impact and data loss by aborting in-flight changes and reverting deployments and models.
Initiate an emergency rollback by identifying the affected service, selecting the last healthy revision, executing the rollback plan (Argo Rollouts or kubectl), reverting the MLflow model if needed, silencing alerts, and validating pod readiness
High-intent search queries and answers about installing and using this skill.
To rollback a production ML deployment during an incident, you need an approved plan to abort canaries, undo the deployment, revert the MLflow model stage, silence alerts, and audit the rollback. This restores service stability by reverting to the last healthy revision.
Reverting an MLflow model stage during a Kubernetes incident involves applying a STOP-approved plan that aborts in-flight changes and rolls back model artifacts. This targeted recovery ensures your production ML service returns to a known-good state without data loss.
Yes, emergency rollback works with both Argo Rollouts and plain Kubernetes Deployments. It executes the rollback plan using the respective controller mechanisms to revert to the last healthy revision and halt bad releases effectively.
To abort in-flight canary releases and stop a bad ML model deployment, execute an emergency rollback plan that halts ongoing rollout processes. This action immediately stops bad releases and prevents further user impact across your production environment.
Validating pod readiness after a production ML rollback is the final step in the recovery process. Once you revert the deployment and MLflow model stage, you must verify that the pods stabilize to ensure the service is fully restored.
You should silence alerts during an incident rollback after undoing the deployment and reverting the MLflow model stage. This prevents alert fatigue while you validate pod readiness and complete the audit trail for the emergency rollback.
CompSci-SquadOrchestrate multi-agent MLOps workflows for ML pipelines with MLflow, Feast, and KServe.
brainbrewlabsRoll back Kubernetes, Docker Swarm, and managed hosting deployments to a previous stable version.
rohitg00Automate progressive Kubernetes deployments with Argo Rollouts and Flagger.
rohitg00Automate Kubernetes resource management by executing kubectl commands.
