Auto-Rollback vs Manual Rollback: Which Algolia Deployment Mode Preserves Root-Cause Visibility?

Direct answer

Manual rollback is more likely to keep Algolia deployment failure logs actionable for post-incident analysis. Because it requires an explicit request, the original failed-deployment event—including its HTTP status, response body, and timestamp—remains in the log without an automatic republish step that can add or overwrite recovery events. Auto-rollback, documented as enabled by default on paid plans, immediately republishes the previous version. That recovery step can produce its own log entry, and if the rollback itself fails, the visible outcome may emphasize the rollback failure instead of the original 400 or 403 cause.

Likely explanation versus confirmed facts

Item	Status
Auto-rollback default on paid plans	Documented behavior; treat as release- and plan-dependent
Manual rollback requires explicit request	Documented behavior
HTTP 400/403 alone identifies root cause	Not confirmed; inspect full response body and request context
rollbackMode field exists	Version-sensitive; may be absent on older releases or free tiers
Manual rollback is the only reliable option across all tiers	Not confirmed; separate original-failure and rollback metadata can make auto-rollback actionable

Verification steps for this case

1. Collect the complete deployment event sequence around the incident: failed, rollback, and republish events, with timestamps.
2. Confirm whether the log or API response retains the original failed-deployment payload, error code and message, deployment ID, rollback mode or equivalent, and rollback success/failure flag.
3. For auto-rollback, verify that the original failure event remains after rollback completes and that rollback failure is logged separately from the original deployment failure.
4. If the original failure payload and correlation ID are absent, capture the failure response before requesting rollback, then use manual rollback.

Version-specific fields and payload settings to verify

• Original failure event: deployment ID, HTTP status, error message, request payload, timestamp.
• Rollback metadata: rollback mode or equivalent field, rollback success/failure flag, rollback timestamp.
• Correlation: a stable deployment/rollback correlation ID that links the rollback event to the original failure without merging them into a single entry.

These field names are not guaranteed across Algolia releases. Confirm them in your running release before building dashboards or alerts.

Missing diagnostic detail

Does Algolia retain the original failure payload and a rollback correlation ID separate from the recovery or republish event? This detail changes the recommendation: if yes, auto-rollback can remain enabled with log-based correlation; if no, use manual rollback and capture the failure response first.

Assumptions and uncertainty

This answer treats Algolia documentation and plan defaults as unverified for your specific release and tier. HTTP 400 or 403 is not proof of a mapping conflict, permissions mismatch, or network timeout. Free-tier deployments may not expose auto-rollback behavior at all. Treat rollback behavior as contract- and release-dependent until verified against your Algolia console or API logs.