---
title: "Identify missing sources in an investigation timeline | Proofline Help"
description: "Check the environment portion of the timeline before inferring that nothing changed. It distinguishes earlier observed incidents from sources that could only be read in part."
url: https://proofline.sh/help/investigation-unread-timeline/
---

[Help center](/help/)

# Identify missing sources in an investigation timeline

Check the environment portion of the timeline before inferring that nothing changed. It distinguishes earlier observed incidents from sources that could only be read in part.

## Find unread and partial intervals

In an investigation that a Cloud Monitoring alert, a CloudWatch alarm, a Sentry issue alert or an Azure Monitor alert started, the environment's part also shows earlier incidents of the same alert that opened in the same environment in the 24 hours before the symptom started. For a Cloud Monitoring alert, these are incidents of the same alert policy that another investigation took in. When the policy alerts separately for each resource or metric label value, each of those is its own alert, so only incidents with the same labels as this one count. For a CloudWatch alarm, they are earlier times the same alarm entered the ALARM state, whether that entry started an investigation of its own or joined another. An entry that joined another investigation is counted without an entry on the timeline. For a Sentry issue alert, they are earlier times an alert on the same Sentry issue started an investigation of its own, each at the time of the event that alerted. For an Azure Monitor alert, they are earlier times the same alert rule fired on the same resources and started an investigation of its own, each at the time it fired. When a metric alert rule is split by a dimension, only firings with the same dimension values as this one count. Each earlier incident sits at its opening time, and its **View investigation** link opens the investigation that read it. When Proofline had recorded the incident's close by the time it read the timeline, the entry reads **Earlier incident** and spans the opening to the close, so you can tell whether a change landed while the alert was firing or after it had cleared. Otherwise the entry reads **Earlier incident opened** and shows only the opening: the close is unknown, which does not mean the incident was still open. A close recorded later appears only on a rerun. An earlier Azure Monitor incident's close comes from Azure Monitor's resolved notification for that firing. Sentry sends no close, so every earlier Sentry incident reads **Earlier incident opened**. So does every earlier incident of an Azure Monitor Activity Log alert, Service Health and Resource Health alerts included, because Azure Monitor never resolves one. Proofline places at most 20 earlier incidents, those nearest the symptom start, and only those whose investigation read the alert completely, or for a Sentry issue, read the issue's events completely. It counts the others without an entry. In **Evidence**, **Earlier incidents of this alert** lists the placed incidents by opening time, each with its close when the timeline has one and linked to its investigation, and counts the others, such as **3 more**. A CloudWatch alarm investigation recorded before Proofline kept the alarm's identity shows no earlier incidents.

An entry such as **Configuration changes unread**, **Serving revisions unread** or **Earlier incidents unread** marks a span in which that source returned nothing to Proofline. An entry such as **Configuration changes partly read** marks a span in which the read stopped early: the entries shown from that source were read, and others may be missing. In either span, the absence of an entry establishes nothing about whether something changed. Repeated gaps from one source share one row, which says how many it holds, such as **12 gaps**, spans from the earliest to the latest of them, and lists each recorded reason once. **Serving revisions unread** names the service, its location and its project, and the reason Proofline could not read what it was serving. **Serving commit unknown** means that Proofline knows which revision was serving but not its commit. For a repository that runs on a GKE cluster, the entry names the cluster, its location and its project and spans the whole window, because Proofline does not read what GKE serves.

## Related guides

-   [Find an environment's investigations](/help/environment-investigations/)
-   [Create an environment](/help/create-environment/)
