Skip to main content

Asset Detail: Metric Reference

This page is the source of truth for the in-app Explain this panels on the asset detail view (/assets/detail?id=...). Each section is written once as a content partial under _explain/assets/detail/ and rendered both here and inside the app's info panel (scripts/build-explain.mjs compiles the registry).

The summary card at the top of an asset's detail page: identity, connection status, a live health/CPU/memory strip, and editable instance settings.

How it's calculated​

  • The header shows the instance name, a Connected/Disconnected badge from the same connection flag as the list, and a subtitle with engine, agent-reported version, and environment.
  • The metric strip renders Health, CPU, and Memory only when the backend has a value for each. Health is tinted green at 80% or above, amber at 60% or above, and red below that — the same thresholds used elsewhere in the product.
  • Expanding the card reveals read-only details (full version, hosting, replication role, Oracle edition, database role, CDB flag, last health check, created/updated timestamps, agent ID) — each row is skipped entirely when the backend has no value, rather than showing a placeholder.
  • Settings — instance name (3–100 characters), environment, and owner — are edited from the same expanded panel. Instance name and environment are required; owner is optional and lists the organization's members, loaded only once the card is expanded.

Reading it​

CPU and memory only appear once the agent has reported a recent system-stats snapshot, so a freshly connected asset can show Health without them yet. Use Edit to correct an instance name or reassign ownership — engine, version, and health are reported by the system and can't be edited here.

Connection & Agent​

How the agent reaches this server, plus a live connectivity check.

How it's calculated​

  • The panel's opening state follows the instance's own signals: no agent ID means first-time setup, an agent ID with an inactive instance means disconnected, and an agent ID with an active instance means connected and collapsed.
  • The summary strip shows Hosting (Self-Hosted or Cloud, plus provider when cloud), Connection (Direct — Bridged is planned but disabled today), and Platform (Linux or Windows). Editing these regenerates the install command shown below them.
  • "Run a connection test" starts an async check and polls its status roughly every 5 seconds for up to about 100 seconds before timing out. A passing test refreshes the instance, its configuration, and the header's Connected badge together, so all three agree without a manual reload.
  • A disconnected instance never collapses this card — the retest prompt and guidance stay visible until a test passes.

Reading it​

If the instance shows Disconnected, re-run the install command before retesting — a stale registration token or an agent process that stopped running are the most common causes. A successful test is what flips the instance back to Connected; simply waiting does not.

Configuration​

Agent version and how often this instance is collected.

How it's calculated​

  • Shown only once an agent has connected; before that, the card says there is no collection configuration to show yet.
  • Agent version and Status (Active/Inactive) come from the agent's own reporting. Last seen is a relative timestamp of the agent's most recent check-in.
  • Collection interval shows the High, Medium, and Low frequency intervals in seconds or minutes — these control how often different classes of metrics are gathered for this instance.
  • Update channel shows the agent's update channel, with "(checks off)" appended when automatic update checks are disabled for it.

Reading it​

An Inactive status with a recent Last seen usually means the agent process stopped cleanly rather than lost connectivity — cross-check against the Connection & Agent section above before assuming a network issue.

Notifications & Tickets​

Maps each database on this server to a Jira project so alerts open tickets in the right place.

How it's calculated​

  • One row per database discovered on this server. If Jira isn't connected at the workspace level, the section shows a "Connect Jira" prompt instead of the mapping list.
  • Each row is a single dropdown: choosing a project creates the mapping, choosing a different one updates it, and choosing "— none —" removes it. Mappings are server-scoped through this card, but stored per database, not per server.
  • "Manage integration" links out to the workspace-level Jira integration page for connecting, disconnecting, or reviewing all projects — connecting Jira itself does not happen from an asset's detail page.

Reading it​

Map only the databases that need alerts filed against a specific Jira project. A database left unmapped simply has no automatic ticket routing configured for it here.