The IncidentRelay project has been launched, developing an open system for organizing on-call shifts, routing alerts, and managing incidents, which operates on a self-hosted server. The project is aimed at SRE, DevOps, and infrastructure teams who need a locally deployable alternative to SaaS services for on-call management, escalation policy implementation, and incident response. The project code is written in Python and licensed under the MIT License.
IncidentRelay accepts events from monitoring systems, matches them with routing rules, and delivers notifications to the responsible on-call personnel or teams. The system includes on-call schedules, rotations, shift overrides, incident acknowledgment, incident resolution, reminders, escalations, and silences to suppress known or scheduled alerts.
It supports the reception of events from Prometheus Alertmanager, Zabbix, and arbitrary webhooks. Notification channels include Mattermost, Telegram, email, webhook, and voice providers. In Mattermost and Telegram, notifications can contain actions for acknowledgment and issue resolution, allowing incident handling without switching to a separate interface.
IncidentRelay features a model for access separation by groups and teams. This allows visibility of schedules, routes, notification channels, and alerts to be restricted among different teams. An HTTP API is available for automation, and bearer tokens and route tokens are used for integrations.
The project can be used as an intermediate layer between monitoring systems and notification channels: Alertmanager or Zabbix sends an event to IncidentRelay, after which the system identifies the team, the current on-call person, applies routing rules, and sends a notification to the appropriate channel. For unacknowledged incidents, repeated reminders and escalation to the next rotation participant can be executed.
Source: opennet.ru
