Connection Health

functional

Every saved connection's observed health - badge, ages, last error, and a measured test latency.

Connection Health puts every saved connection in one overlay so you can see, at a glance, which ones are reachable. Each row shows a colored healthy / failed / untested badge, the profile name, an engine and routing subline (Direct or Via Relay), how long ago it last connected and was last tested, a redacted summary of its last error, and a client-measured test latency. Rows sort needs-attention first: failed, then untested, then healthy.

Opening it

Right-click any server in the explorer and choose Tools › Connection Health..., or run Connection Health from the command palette. It always lists every saved connection, whichever server you opened it from.

Reading a row

PartWhat it tells you
DotThe live state, the same dot the explorer draws: green connected, red unreachable, gray not connected. A faded green or red is the last known state from a previous session, not something checked since launch.
BadgeThe result of the last test or connection attempt, kept between launches: healthy, failed, or untested.
SublineEngine · Direct or Via Relay · tested 5m ago · connected 2d ago. Ages read just now, minutes, hours, or days, and never when it has not happened.
ErrorThe last failure in red, folded to one line and cut at 120 characters. The stored summary has password=-style values stripped before it is saved.
LatencyHow long the test you just ran took from this machine, start to finish: <1 ms, milliseconds, or seconds. A dash until you test in this session.

The line above the list counts the saved connections and how many are connected, unreachable, and not connected right now.

Testing

  • Per-row Test - re-check a single connection on demand; the row shows "Testing…" until the result lands.
  • Test All - sweep every connection sequentially and watch the badges settle, with a running tally of passed and failed. Profiles that need an interactive sign-in (Entra browser or device code, AWS or GCP IAM) are skipped and counted, so the sweep never pops sign-in windows. Their own Test buttons are disabled here too; test those with the Test button in the connection editor.
  • Refresh - re-reads the saved connections and clears the latencies measured so far. It is disabled while a test is running.
  • Redacted errors - enough of the last failure to act on, without leaking secrets into the row.
  • Real routes - Via-Relay connections are tested over their actual relay route, not a shortcut, so the badge means what it says.
SQLly
SQLly Connection Health overlaySQLly Connection Health overlay
Connection Health: a row per saved connection with status badge, last-connected and last-tested ages, last error, and measured latency.

Diagnosing a slow or failing connection

  1. Test it on its own. Press the row's Test. A fresh result replaces the stored badge, error, and test age.
  2. Read the error line. The saved SQL account has no password in the system keyring means this machine's keychain has no secret for the connection, common right after copying your setup to a new machine. Edit the connection and enter the password again, or import a connection file exported with credentials. If the test you just ran fails with a dash where the latency should be, it never got as far as dialing the server.
  3. Compare latencies. The figure covers the whole test, reaching the server, signing in, and checking the connection, so it is not a network ping. Test All gives you comparable numbers across servers; one that is much slower than its neighbors on the same network points at that server, its route, or its sign-in.
  4. Check the route. A Via Relay connection depends on its relay session; if that is not open, the test fails and says so. See Reach your database from anywhere.
  5. Look at the server itself. Once it connects, the Server Dashboard and Activity Monitor show whether it is busy or blocked.

Results persist to the same connection store the Server Explorer reads, so the badges here and the tree agree afterward. Hover a server in the explorer to see the same last-known state, and click a red dot on a server that failed to load to try it again.