Database health
All your databases at a glance, checked every minute by the collector.
Updated just nowISTHow this page works
FleetDatabasesReportsUsage
6need attention
of 9 databases · 1 being worked on · 5 with no owner yet
3 Critical3 Degraded3 Healthy
Open longest
2 h 37 m
Billing-EU failing invoice calls since 08:05
No owner
5
Select “I’m on it” so others know it’s handled
Fixable here
4
have a suggested fix you can review and run
Open issues
Worst first · one row per problem, not per databaseAll · 6Mine · 0No owner · 5What changed: the “Needs attention” cards become issue rows. Each row says what is wrong in plain words, names the culprit (query, procedure, session or table), shows who owns it, and offers actions right there. Selecting a row opens the drill-down without leaving the page.Needs: an issue record per alert (owner, acknowledged-at) and a culprit captured when the alert opens. Today the backend stores neither.
Recent events
24hPlain-language events. Today’s feed reads “Blocked sessions 1 count (≥ 1) again (time 6); joined the alert that was still open”. The same event is rewritten below.
All databases
9 databases · sorted worst first| Database | Status | Score | Open issue | Owner | Engine |
|---|
Fix workspace
Review the suggested fix with the agents, run it with guard rails, and watch whether the problem goes away.
Notifications
Alerts from your databases, issues people shared with you, and results of fixes.
What pops up
Your choice, per typeCritical alerts on my databasesStays on screen until you open or dismiss it
WarningsMost are noise at first. Start quiet.
Someone shares an issue with meIncludes their message
A fix I ran finishesResolved, not resolved, or failed
E-mail copyNeeds e-mail set up by an admin
Toast rules that keep people from muting everything: group repeats (the same alert coming back 6 times is one toast, not 6); at most 3 on screen; only critical alerts stay until dismissed; everything else fades after 6 s and waits in the inbox.Needs: a notifications table (recipient, kind, issue link, read-at) and a push channel to the browser (SSE or WebSocket). Neither exists today; polling every 30 s is an acceptable first step.