๐Ÿ–ฅ๏ธ

Developers

Self-Hosted Guide

Deploy GlobeMeet on your infrastructure

Step-by-step setup for Ubuntu 20/22 โ€” single-server pilot through HA multi-region clusters. Reference architectures, capacity planning, and runbooks included.

100

Concurrent sessions on a single 8-vCPU node

2,000

Concurrent sessions on a 3-node cluster

Multi-region

Active/active supported

Prom + Grafana

Observability bundled

01

Reference architectures

Single-node (up to 100 concurrent sessions, 4 vCPU + 8 GB RAM), 3-node mediasoup cluster (up to 2,000 concurrent sessions), and multi-region active/active for global distribution. Each architecture comes with a Terraform / Ansible reference.

  • Ubuntu 22.04 baseline images
  • Coturn TURN server config (port 3478, 5349)
  • MinIO + Redis + PostgreSQL stack
  • TLS termination via Caddy or Nginx
  • Prometheus + Grafana dashboards bundled
  • Reference Terraform modules for AWS Mumbai and GCP Mumbai
02

Capacity planning

Each agent's session uses about 1.5 Mbps of media plus negligible signaling. A single mediasoup worker handles ~500 simultaneous producers; the SFU forwards to as many consumers as bandwidth allows. The recorder is the heavier workload โ€” plan for one 4-vCPU recorder per 50 concurrent recordings.

03

Hardening checklist

Disable root SSH, require key-based login, enable UFW, tighten Postgres pg_hba.conf, run MinIO with versioning + object-lock, rotate TLS certs via Caddy auto-issuance, enable audit logging, and run weekly security updates via unattended-upgrades.

04

Backup & DR

Daily Postgres backups to encrypted S3 (or MinIO replication). MinIO bucket replication across regions. Mediasoup is stateless โ€” failover is just rerouting. RTO under 15 minutes for single-region failure with an active passive setup.

SPEC

Technical specifications

OS
Ubuntu 20.04 LTS, Ubuntu 22.04 LTS, RHEL 8/9 (community)
Database
PostgreSQL 14 or 15
Cache
Redis 7
Object store
MinIO RELEASE.2024 or any S3-compatible
TLS
Caddy 2.x or Nginx 1.22+
TURN
Coturn 4.6+
Monitoring
Prometheus + Grafana + Loki bundle included
FAQ

Frequently asked questions

Yes. We provide Helm charts for the API tier; the SFU and recorder run best on dedicated VMs (network/perf-bound) and are deployed via Ansible. A pure-Kubernetes reference is on the roadmap.

The SFU and recorder are Linux-only. The API tier theoretically runs on Windows but is unsupported. All production deployments are on Linux.