Monitoring & metrics

Documentation Index

Fetch the complete documentation index at: https://docs.rafftechnologies.com/llms.txt

Use this file to discover all available pages before exploring further.

_{Updated May 8, 2026} A running VM emits a continuous stream of operational signals — CPU, memory, network, disk, and aggregate data transfer. Raff collects these from the hypervisor, not from inside the guest OS, so there’s no agent to install and metrics work the same way regardless of distro, kernel, or whether the OS is even healthy. Hypervisor-level metrics treat the VM as a black box: they tell you how loaded the VM is, but not why. Pair them with in-VM tools (top, htop, your APM) for the full picture.

​

What’s tracked

All five share a single time-range selector and update on a regular interval.

​

Where to find it

Open the VM detail page either way:

• Click the row anywhere in the Instances list, or
• Open the row’s ⋮ menu and pick View Details

The detail page opens on the Graphs tab by default. The time-range selector at the top of the tab applies to every chart at once: 30m / 1h / 2h / 6h / 24h / 7d / 1m / All. Pick a wider range for trend analysis, narrower for live troubleshooting. The Refresh button forces an immediate re-fetch; otherwise charts auto-refresh.

​

Reading each graph

​

Total Data Transfer

Cumulative outbound bytes from the VM to the public internet for the selected window. Outbound bandwidth is unlimited and free on every plan — this graph is for visibility, not billing. Useful for:

• Catching unexpected egress (compromised host, runaway loop, data leak)
• Sizing CDN offload for static assets
• Capacity-planning across regions

Inbound traffic isn’t counted here; private VPC traffic between your own VMs is also excluded.

​

CPU Usage

Average CPU utilization (%) across all vCPUs over time. The dashed red line at 80% is a soft threshold — sustained CPU above it means the VM is becoming a bottleneck. What the shape tells you:

• Single spikes — normal: a deploy, a backup, a cron job
• Sustained 80%+ — undersized; resize up, optimize, or split workload
• Saw-tooth — periodic load: a scheduler running every N minutes
• Stuck at 100% — a runaway process; SSH in and check top

​

Memory Usage

Used and Free memory in MB, summing to the VM’s total RAM. What to watch for:

• Used trending up over hours — memory leak; check long-running services
• Sharp drop in Used — OOM kill, service restart, or VM reboot
• Used near total — paging or swap pressure; resize up

Note: “Used” includes the OS page cache, which Linux happily fills. Treat it as a load indicator, not a panic signal.

​

Network Traffic

RX (receive) and TX (transmit) per network interface, in KB/s. Each interface is numbered (0, 1, …). Typically 0 is the public NIC and 1 is the VPC NIC, but labelling depends on attachment order — verify on the Network tab. Use it to:

• Confirm an app is actually receiving traffic (RX > 0 on the right interface)
• Catch traffic on the wrong NIC (e.g. internal calls hitting the public IP)
• Detect DDoS, scraping, or bandwidth abuse (sudden sustained RX spike)
• Verify private-network setups (RX/TX on the VPC interface)

​

Disk Throughput

Read and Write in KB/s on the boot disk (attached volumes have their own metrics). What to watch for:

• Sustained writes — background jobs, log churn, or a database flushing
• Sustained reads — cache misses; consider more RAM or a read replica
• Periodic spikes — backups, cron jobs, snapshot windows

​

VM Information side panel

The left side panel on the detail page is a live summary of the VM — useful even when you’re not looking at graphs. It collapses with the chevron when you want more room for charts. Click an arrow on a section header to jump to the related management page (Resources → Resize, Storage → Volumes, etc.).

​

What hypervisor metrics can’t tell you

Because the data comes from outside the guest, the dashboard can’t see: For production observability, run an in-VM agent and ship to your own monitoring stack (Prometheus + Grafana, Datadog, New Relic, etc.). Raff’s graphs are best used as a first-glance signal: is the VM alive, is it loaded, is it transferring data — then dig deeper with in-VM tools.

​

Retention and refresh

• Time range — up to All (since VM creation). Older windows are aggregated; recent windows show finer-grained samples.
• Refresh — charts update on a regular interval. The Refresh button forces an immediate refetch.
• Export — not available in the dashboard today; for long-term retention or alerting, scrape from inside the VM or use an external monitoring stack.

​

Related