Changelog
No more connection alerts for problems on Artie's side
Artie no longer sends source or destination connection alerts when the real failure is that the Dashboard cannot reach a data plane, and an AWS role that refuses AssumeRole now returns a clear error naming the role.
Artie pings each source and destination on a schedule and alerts you once a ping has kept failing for ten minutes. When the Dashboard itself could not reach a data plane, that failure fell into the same path, and ten minutes later you got a "Source connection error detected" email about your database when your database was fine. Separately, when an AWS IAM trust policy stopped Artie from assuming a role, the request failed with a generic server error that gave no hint about which role or why.
The ping job now recognises a data plane it cannot reach (DNS failure, connection refused, no route) as an Artie-side problem: it records the event internally, leaves your connection alerts exactly as they were, and does not email you. Data-processing alerts for the pipeline still evaluate as before. A related fix stops one pipeline whose destination was deleted from skipping the ping run for the whole company. And when AWS STS denies AssumeRole, Artie returns a configuration error that names the role ARN and tells you to check its trust policy, so the fix is obvious from the message.
This is for anyone who gets Artie's connection alerts, and teams using AWS role assumption for data planes or PrivateLink.
Why this matters
- A connection alert now means your source or destination is the problem
- No on-call wake-ups for outages that are ours to fix
- A denied AWS role tells you which trust policy to correct