Guides: deploy & operate

How to view live logs in the admin GUI

This guide shows you how to tail and filter the proxy's operational logs from the admin UI — no SSH, no grep on stdout.

Where it lives#

Open Settings → Observability → System logs (/_/admin/diagnostics/logs). The view shows the proxy's operational log stream — security, rate-limit, S3-error, replication, and lifecycle lines — captured at INFO and above.

This is admin-session-gated: you must be signed in to the admin GUI.

Tail logs live#

Toggle Follow to stream new log lines as they happen (over server-sent events). Leave it off to inspect a static snapshot of the recent backlog and refresh on demand.

Filter#

Three filters narrow both the backlog and the live tail, server-side:

  • Level — Error, Warn+, Info+, Debug+. (Lines below the capture floor never enter the ring; see below.)
  • Target — substring match on the log target (Rust module), e.g. auth or replication.
  • Search — free-text over the message and structured fields, e.g. a bucket name or client IP.

Click a row to expand its structured fields.

Reproduce-and-watch#

To debug a specific request, turn Follow on, set the level and a target or search term, then trigger the request. The matching line appears as it's logged. For per-request trace detail, widen the capture floor with DGP_LOG_RING_LEVEL=debug (see below) and restart.

What it is — and isn't#

The viewer reads an in-memory, per-instance, bounded ring. It is a triage convenience, not a log store:

  • DGP_LOG_RING_SIZE (default 2000) sets the ring capacity.
  • DGP_LOG_RING_LEVEL (default info) sets the minimum severity captured, independent of DGP_LOG_LEVEL.

For retention, search, and aggregation across instances, point a log shipper at the proxy's stdout. Set DGP_LOG_FORMAT=json for one JSON object per line, which is jq-greppable and ingests cleanly into Loki, Quickwit, or an ELK stack.