Ikibazo gihuza ibibazo · Hagati Urwego
GitOps ya ML hamwe na Argo CD Ikibazo
Sobanukirwa na leta yifuzwa, ubwiyunge, gutembera no gusubira muri CD ya Argo, kandi utandukanye ubuzima bwoherejwe no kwemeza icyitegererezo.
Ikibazo 1 Bya 8
What does GitOps treat as the declared desired deployment state?
Reba buri kibazo8 ibibazo
Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.
- What does GitOps treat as the declared desired deployment state?
- Version-controlled configuration in a Git repository
- The current state of a developer's terminal
- The most recent prediction log
- The model's training accuracy only
- What does Argo CD do when live cluster state differs from Git?
- Automatically retrains the model
- Reports drift and may reconcile according to configured sync policy
- Changes the source repository to match the cluster
- Rewrites the model artifact
- What does self-heal generally do in an Argo CD setup?
- Restore deleted training data
- Repair model accuracy after drift
- Reconcile manual live changes back toward declared Git state
- Rotate every cloud secret
- Why can automated sync be risky with an unreviewed manifest change?
- It guarantees no rollback is possible
- It disables Kubernetes health checks
- It changes Git history automatically
- It can propagate an incorrect desired state quickly
- How should a model image be referenced for traceable deployment?
- By a mutable latest tag only
- By an immutable digest or version tied to validation evidence
- By the training dataset's filename
- By an unrecorded local path
- What does Kubernetes health status establish in this context?
- That data are free of bias
- That the model is calibrated
- Resource health, not prediction quality
- That the candidate beat its baseline
- Why may a Git revert not fully restore an ML deployment?
- Kubernetes ignores image versions
- Git cannot track text manifests
- Argo CD requires retraining to sync
- External state such as registry aliases or data migrations may not roll back
- Where should deployment credentials be stored?
- In a secret-management mechanism referenced by configuration
- As plaintext in a committed manifest
- Inside model weights
- In the public image tag