Service Detail

The Service detail page shows a single Service within an App, including resource allocation, scaling configuration, environment variables, GPU telemetry, live logs, events, and cost data.

Navigating to the Service Detail Page

From the Console sidebar, select an App under a Rack. On the App page, click any Service name in the Services list to open its detail view.

The URL follows the pattern: /<org>/<rack>/<app>/services/<service-name>.

Header

The header displays:

• Service name and current replica count (e.g., 3 / 5 for 3 running out of 5 max)
• Agent badge: Shown when the Service runs as a DaemonSet (one replica per node).
• Domain links: Clickable HTTPS links for each domain assigned to the Service.
• Resource cards: CPU (millicores), Memory (MB), and GPU (count and vendor) at a glance.
• Refresh button: Forces a re-fetch of all Service data.

Panels

A tab bar below the header switches between panels. The active panel is reflected in the URL query string (?section=logs, ?section=scaling, etc.) so links can deep-link to a specific panel. V2 Racks show only the Overview panel.

Overview

The default panel. Four cards:

• Identity: App name, Service name, domains, current Release ID.
• Configuration: CPU, memory, GPU allocation, min/max replicas, cold-start indicator.
• Recent activity: Last 3 events (deploy, restart, scale, override toggles).
• Health: App status badge, replica count, autoscale enabled/disabled, agent type.

Logs

Streams live log output from the Service. Controls:

• Pause / Resume: Buffers incoming lines while paused and flushes them on resume.
• Clear: Clears the log terminal.
• Filter: Regex filter applied to the displayed lines.
• Wrap toggle: Enables or disables line wrapping in the terminal.

The log stream reconnects automatically (up to 3 retries with exponential backoff). A manual Retry button appears if all retries are exhausted.

Events

A table of events scoped to this Service, including deploys, restarts, scale changes, and autoscale override actions. Each row shows the timestamp, event summary, actor, and status. A link at the bottom navigates to the full App events view.

Scaling

Displays the Service scaling configuration and provides controls to change it.

Bounds card shows min replicas, max replicas, and current replica count. A cold-start badge appears when the Service can scale to zero.

Autoscale settings shows the current autoscale state:

Badge	Meaning
Override active	Console-driven autoscaler is managing this Service
From convox.yml	Autoscale triggers are declared in the manifest
Not configured	No autoscaler is configured

A trigger table displays thresholds for CPU, memory, GPU utilization, and inference queue depth. When an override is active, click the pencil icon on a threshold cell to edit the value inline.

Actions card provides two scale modes:

• Fixed count: Set a single replica count.
• Range (min-max): Set a min and max for autoscaling.

Both modes include a confirmation dialog for high-risk changes (scaling to 0 or large jumps). A Restart button performs a rolling restart of all replicas.

Enable / Override / Disable autoscale buttons manage Console-driven autoscale overrides. See Autoscale Triggers for details.

Environment

Displays per-Service environment variable overrides defined in convox.yml under the services.<name>.environment block. Values matching sensitive key patterns (TOKEN, KEY, SECRET, PASSWORD) are masked by default with a reveal toggle.

This panel requires Rack version 3.24.6 or later.

Cost

Shows month-to-date spend for this Service. Requires cost tracking to be enabled on the Rack.

• Headline card: Total spend in USD with pricing table label.
• Breakdown table: Per instance type and capacity type (on-demand vs. spot), showing active replicas and spend.

A link at the bottom navigates to the full App cost breakdown.

GPU

Displays GPU telemetry for Services with GPU reservations. Requires GPU observability to be enabled on the Rack.

• Summary cards: GPU utilization, memory used/total, vendor and count.
• Extended counters: Tensor active, SM active, DRAM active, FP16/FP32 active, power draw.
• Utilization chart: Time-series GPU utilization with a configurable display window (5m, 30m, 1h, 24h).
• Per-process table: Snapshot of each running GPU Process with utilization and memory stats.
• Grafana deep link: When a Grafana URL is configured on the Rack, a button links to the per-Service dashboard.

Test Model

Available when the Service has a domain (including internal .local domains for private Services). Opens an interactive playground for sending requests to the Service endpoint.

See Also

• Autoscale Triggers for managing Console-driven autoscaling
• Cost Tracking for enabling per-App spend tracking
• Scaling for convox.yml scale configuration