Incident, on-call and postmortem records

Production incident records: alerts and pages, incident timelines, responder chat, status updates and written postmortems with root causes and follow-up actions. Labs use them to train and evaluate SRE and operations agents on how teams diagnose and fix real outages.

Last updated October 3, 2026

What a record contains

One incident: triggering alerts, a timeline of actions and status changes, responder discussion, customer-facing updates, and the postmortem with root cause and action items.

FieldWhat it holds
incident_id, severityPseudonymous ID and level
alerts[]What paged, and when
timeline[]Actions and status changes in order
channel_messages[]Responder discussion
status_updates[]Customer-facing updates
postmortemRoot cause, contributing factors and action items
Illustrative record. The values are made up to show the shape of the data.
{
  "incident_id": "INC-311",
  "severity": "SEV2",
  "started_at": "2024-11-02T03:05Z",
  "resolved_at": "2024-11-02T04:22Z",
  "timeline": [
    {"at": "03:05", "event": "alert: settlement_job_duration > 30m"},
    {"at": "03:16", "event": "rollback of batch size change started"}
  ],
  "postmortem": {"root_cause": "unbounded query after merchant growth",
                 "action_items": ["paginate settlement query", "add duration SLO"]}
}

How AI labs use it

SRE and on-call agents
Learn the diagnosis steps responders actually took.
Root-cause reasoning
Postmortems give the answer a diagnosis should reach.
Evaluation from real outages
Replay a timeline and score an agent’s next step.

Typical preparation requirements

Agreed with the supplier before any work begins. Typical requirements include:

  • Details that would expose security weaknesses removed or generalized
  • Customer names in impact statements pseudonymized
  • Secrets and internal hostnames redacted

Every dataset has a documented owner and confirmed licensing rights. See data governance on sourcex.si.

What makes a strong package

  • Postmortems linked to incidents and tickets
  • Log or metric snapshots kept with the timeline
  • Blameless postmortems with detailed reasoning

Who typically holds it

  • SaaS companies
  • Fintech companies
  • E-commerce platforms
  • Infrastructure companies

Know a company like this?

Introduce the company to SourceX. If its data deal closes, you can earn up to $100,000 in referral fees, paid after the buyer accepts the data and SourceX receives payment.

Refer a company

Questions

Are raw logs included?

Where they were kept and can be cleaned. Packages can include the log excerpts referenced in the timeline rather than full log stores.

Need this data for a model?

Describe what you need: domain, volume, history, format and licensing terms. SourceX looks for companies that hold it and can license it.