Skip to content
M↓ View as Markdown ↗

Data Source Health Monitoring

Actian AI Analyst checks your data source connections once a day. If a connection fails or its queries become very slow, it alerts the people who maintain the Workspaces that use that connection.

Availability

  • Who gets alerts: members with the Edit capability in a Workspace that the affected connection is attached to.
  • Who can see a connection's health: members with the Edit capability in a Workspace the connection is attached to.
  • Deployment: Actian SaaS and self-managed deployments. Slack and Microsoft Teams channel alerts are not available on air-gapped self-managed deployments.

What gets checked, and when

AI Analyst checks every warehouse connection that is attached to at least one Workspace. By default the check runs at 05:30 UTC.

  • A connection attached to several Workspaces is checked only once.
  • A connection that isn't attached to any Workspace isn't checked, so it never produces an alert. To learn how to attach a connection to a Workspace, see Connections and Workspaces.
  • Uploaded files, such as CSV files, are not checked.

Each daily run makes two checks: a connection check and a latency check.

Connection failure alerts

AI Analyst tries to connect to each data source. If the connection can't be made within 30 seconds, the check fails.

  • An alert is sent when a connection starts failing, not on every daily check while it stays down. You get one alert per outage.
  • After the connection recovers, a later failure sends a new alert.
  • The alert links straight to the affected connection in Studio.

High latency alerts

AI Analyst also looks at how long your queries took on each data source over the past 7 days. If the average duration of successful queries is more than 30 seconds, it sends a high latency alert.

  • AI Analyst alerts only after at least 5 successful queries in the past 7 days, to avoid false alarms on sources you rarely use.
  • Failed queries don't count toward the average or the 5-query minimum.
  • The average covers every query your organization ran on that connection, in all Workspaces.
  • After a high latency alert, AI Analyst doesn't send another latency alert for that data source until it has gone 14 days without a slow result. A source that stays slow every day is alerted once, not every 14 days.

Where alerts are delivered

Health alerts go only to the Workspaces that the affected connection is attached to. Each alert can be delivered in up to three ways, as soon as the check finds the problem:

Where Who sees it How to control it
Email Each member with the Edit capability in one of those Workspaces Each member turns their own emails on or off in Settings → Notifications.
Steward inbox Everyone who can open the Steward inbox in those Workspaces An alert item is always added. Health alerts are sent right away and are not part of the daily Steward digest.
Slack or Microsoft Teams channel Everyone in the channel Set per Workspace in Settings → Administration → Notifications. If two Workspaces post to the same channel, the channel gets one message.

Organization Admins who don't have Edit in an attached Workspace don't receive the email.

Manage your email alerts

Every member has a personal Settings → Notifications page. If you have Studio access, the Email notifications section shows the Connection failure alerts and High latency alerts toggles. Both are on by default.

  1. In Studio, go to Settings → Notifications.
  2. Under Email notifications, turn Connection failure alerts and/or High latency alerts on or off.
  3. Click Save in the save bar.

Your choices apply only to you. You receive the emails only if you have Edit in a Workspace that uses the affected connection.

Post alerts to a Slack or Teams channel

You can also post health alerts to a Slack or Microsoft Teams channel. Channel alerts are set per Workspace and apply to everyone in it.

Availability

  • Who can do this: members with the Edit capability in the active Workspace.
  • Deployment: needs a connected Slack workspace or Microsoft Teams tenant. Not available on air-gapped self-managed deployments.
  1. Go to Settings → Administration → Notifications.
  2. Under Channel notifications, find Connection failure alerts or High latency alerts.
  3. Choose the channel to post to. Choose Not posted to turn channel alerts off for that alert type.
  4. Click Save in the save bar.
  5. Optional: click Send test to post a test message to the saved channel.

If no Slack workspace or Teams tenant is connected yet, the page says Connect a Slack or Microsoft Teams workspace to send channel notifications. Click Go to Connections to add one.

Check a connection's health

To see how a connection has been doing, open it in Connections. The Overview tab has a Connection health card:

  • Status: the result of the latest connection check, OK or Failed.
  • Last checked: when the latest connection check ran.
  • Connect latency: how long the latest connection took to open.
  • A row of colored dots for the last 20 checks, oldest first: green for OK, red for Failed, and amber for Slow (a latency check that found slow queries). Hover over a dot to see the check type, result, date and latency.
  • A summary line, for example 18 of 20 recent checks OK, back to followed by the date of the oldest check shown.

AI Analyst also records a successful check when you create a connection, save new credentials that connect, or click Re-test connection and the test succeeds. A failed test doesn't record a check. Before any check exists, the card says No probe yet. Re-test the connection to record one.

For query performance, open the Query stats tab. It shows Total queries, Error rate and Avg query latency, with charts for the time range you choose (7d, 30d or 3M). Query stats cover only the Workspace you're working in, while the high latency alert averages queries from all Workspaces, so the two numbers can differ.

Fix a failing connection

  1. Open the alert link, or find the connection under Connections.
  2. Check the Connection health card to see when the problem started.
  3. If your warehouse password or key changed, click Edit credentials and save the new values.
  4. Click Re-test connection. A successful test records an OK check right away.

If the warehouse is reachable from your network but not from AI Analyst, check your firewall rules. See Network access.