Production checklist

Walk this list before you depend on Sazabi for a production system. Each item is a decision to make once, and each links to the guide that owns it.

This is a checklist, not a linear procedure. Do the items that apply to you, in any order — skipping one does not block the others.

Scope and rotate your keys

• Send logs with public keys only. They are project-scoped and meant for ingestion; keep them out of anything that also needs read access.
• Reserve secret keys for the API, SDKs, and the MCP server. They are organization-scoped and grant full read and write, so treat them like production credentials — store them in a secrets manager, never in source control.
• Have a rotation plan. Create a new key, cut traffic over, then deactivate the old one. Manage both key types under Administer.

Redact sensitive data

• Decide what must never leave your systems. Strip it before it reaches Sazabi: redact secrets and sensitive fields at your log source or collector — the point where you own the data — rather than after ingestion.

Estimate your volume

• Estimate what each source will send in production. Not only in your test — a source that is quiet in staging can be loud under real traffic. See Instrument for how ingestion and log streams work so your estimate matches what actually flows.

Check the limits

• Confirm your production sources fit within Sazabi's limits. Review the current limits in Limits before cutover and, if a source will exceed them, plan how to shape or split its traffic.

Confirm retention

• Confirm the retention window matches what you need. Investigations and any compliance requirements both depend on it, and logs older than the window are deleted automatically. Retention and regions are covered in Instrument.

Watch for silent failures

• Watch that your sources keep flowing. A source that silently stops sending leaves the agent blind, so check intake activity for each source under Settings > Log streams — each shows whether it is currently receiving data — and re-check after you change anything upstream.
• Route the disconnect notification. Sazabi notifies your team when a log source is disconnected in the project. Choose where that lands under notifications so the change reaches the people who own the source.
• Do not assume silence means healthy. A source that keeps its connection but quietly stops sending data will not raise that notification on its own today, so make checking intake activity part of your production monitoring routine.

Plan for cost

• Review your plan and usage before you scale up traffic. Sazabi charges two ways: token usage draws from your prepaid credit balance, and metered usage — log ingestion, issues, and automation runs — is billed on your cycle invoice. Scaling traffic up raises metered ingestion; running more investigations consumes more credits. Check Settings > Billing and Settings > Usage, and see Administer for how pricing and credits work; the actual numbers live in the product and on the pricing page.

Further reading