Skip to main content
Migration Notice
We're migrating documentation from the old portal into this one. Some things may look a little different or out of place in the meantime — we know, and we're working to get it right. If something's unclear or doesn't look right, let us know.

Agent Dashboard — Ingress (Rx Mbps)

What this chart shows​

Ingress (Rx Mbps) — megabits per second of data arriving at the agent host across all physical network interfaces, sampled from /proc/net/dev on every heartbeat (~30 s + 0–10 s jitter).

During an active copy job, Ingress primarily reflects:

  • S3 GET traffic — object data being read from the source bucket before being uploaded to the vault target
  • S3 PUT acknowledgements — TCP ACK and response traffic from the target S3 endpoint
  • Console polling — job-dispatch payloads and configuration updates received from the console

For a direct source-to-target copy, Ingress and Egress are closely correlated: data is read from the source (Ingress) and written to the target (Egress). A significant gap between the two may indicate multipart overhead, retries, or protocol-level overhead.

Data collection — how bandwidth is sampled​

The agent reads raw byte counters from /proc/net/dev and sums across all physical interfaces. Virtual and container interfaces are excluded:

Excluded: lo, docker*, br-*, veth*, virbr*, tunl*, sit*, dummy*

Formula applied between two consecutive readings:

Rx Mbps = rx_bytes_delta × 8 / 1_000_000 / elapsed_seconds

Results are stored in ClickHouse and displayed here rounded to 3 decimal places.

This metric requires /proc/net/dev and is reported on Linux hosts.

First heartbeat after agent start​

The first sample after an agent starts (or restarts) always reports 0 Mbps. The sampler seeds a baseline on the first call and needs two readings to compute a delta. The second heartbeat (~30–40 seconds later) produces the first real value.

Fleet Total line​

When more than one agent has data in the selected time window, a Fleet Total dashed line is drawn over the per-agent lines. It shows the combined Ingress of the entire agent fleet at each timestamp — a measure of total source-read throughput across all agents.

The Fleet Total is computed client-side by summing all agent series per timestamp. It is not stored in ClickHouse.

Interpreting the chart​

  • Egress ≈ Ingress — healthy copy job; data is being read from source and written to target at roughly equal rates.
  • Ingress > Egress significantly — the source is delivering data faster than the target is accepting it. The agent may be buffering; check Heap Used for memory pressure.
  • Ingress < Egress significantly — the target is receiving more than expected; possible retries are inflating Egress. Check copy job logs for S3 error responses.
  • Agent shows 0 — either the agent just started (first heartbeat), is running on a non-Linux OS, or has no active copy jobs.

Polling interval​

~30 seconds + 0–10 seconds jitter per heartbeat. Historical data is retained for 90 days in ClickHouse.