Capacity Scaling Runbook

Use this runbook when a capacity warning fires or when weekly baseline reports show sustained growth.

Triage

1. Open the Grafana dashboard Akira Capacity Sizing.
2. Confirm the alert is sustained over the configured for window.
3. Compare current load with docs/architecture/capacity-planning.md.
4. Check whether the same node has CPU, RAM, disk, and application-level growth.
5. Record the decision in the weekly capacity report under Recommendations.

Scale actions

Signal	First response	Scale action
CPU >70% on Kamailio	Check CPS, retransmits, and routing reloads	Resize sip nodes cx33 to cx43
RTP sessions >800	Check active dialogs and media distribution	Add rtp-03 or resize existing rtp nodes
Postgres connections >80%	Check pgbouncer pools and backend pool settings	Tune pools before resizing DB
Root disk >80%	Check Prometheus, Loki, logs, and Timescale size	Expand volume or reduce retention
CDR growth >50GB hot	Check compression policy and retention plan	Review Timescale compression/offload

Weekly baseline

Run manually from the management node if the timer is not enabled:

/usr/local/bin/akira-capacity-baseline

The report is written to /opt/akira/reports by default when deployed through Ansible, or reports/ when run from the repository checkout.