Documentation
Connector cluster: components, sizing, high availability
Agent and proxy virtual machines behind a load-balanced virtual IP inside your network: the only REAP components that touch devices.
Components
| Component | Role | Notes |
|---|---|---|
| Connector (agent + proxy) | Pulls configurations and polls metrics (CLI, SNMP); receives telemetry (syslog, SNMP traps, gNMI (opens in a new tab) subscriptions) and flow records (NetFlow or IPFIX) where you configure them; brokers credentials from your PAM; forwards to the platform | Deploy at least two per site or domain for availability |
| Virtual IP (VIP) | Single address to which devices publish telemetry, load-balanced across connectors | Lets a connector be replaced without touching device configuration |
| Platform services | Discovery, config tracker, monitoring, topology, reasoning | Run in REAP Cloud, your cloud account or your data center per the deployment model |
Sizing
| Scale | Connectors | vCPU / RAM / disk per connector | Notes |
|---|---|---|---|
| Pilot: 50–250 devices, one or two sites | 2 | e.g., 4 vCPU / 16 GB / 200 GB | One VIP |
| Production: up to N devices per cluster | 2–4 | One cluster per site group or security zone | |
| Telemetry rate | events/s and metrics/s per connector | Size disk for local buffering during outages buffer duration |
High availability
- Two or more connectors behind the VIP; loss of one connector does not interrupt collection.
- Local buffering and replay when the platform is unreachable: whether supported, and for how long.
- Connector health is visible in the console and alertable.
Lifecycle
- Provision VMs per sizing; open the firewall rules on the network requirements page.
- Install the connector package; register it with the platform using a one-time enrollment token enrollment procedure.
- Configure the VIP; point device syslog, SNMP trap and (where used) gNMI (opens in a new tab) dial-out destinations at it.
- Attach the credential source (PAM safe or vault path) and the read-only device account.
- Verify: discovery reaches all in-scope devices; telemetry rate is as expected; health is green.
- Upgrades: rolling, one connector at a time; rollback by reinstalling the previous package.
Engineering to confirm
- Connector VM sizing and supported hypervisors / container hosts
- Buffering duration and disk sizing rule
- Enrollment procedure and token lifetime
- Maximum devices per connector cluster
Last reviewed: October 7, 2026 (draft) · Owner: solutions engineering · Corrections: security@reapnet.ai