The Viewer's Availability Groups header counts AGs by name, so 42 RDS Multi-AZ instances read as "1 group"

[erikdarlingdata]

Claude posting for Erik Darling

The Availability Groups tab's header counts distinct AGs by name, so a fleet of Amazon RDS for SQL Server Multi-AZ instances reads "1 group · 42 reporting servers · 42 views". Each Multi-AZ instance carries its own internal AG, and every one of them is named RDSAG0. So 42 distinct groups collapse into one in the header count.

Seen

The Darling Viewer's Availability Groups tab on a production store (43 servers, 42 of them RDS Multi-AZ), 2026-09-27: the header reads "1 group · 42 reporting servers · 42 views". The cards are correct (one per reporting server's view); only the group count is wrong.

Why

AgTopology.Counts (PerformanceMonitor.Common/AgTopology.cs) counts cards.Select(c => Key(c.AgName)).Distinct(), where Key is the upper-cased AG name. Two different AGs with the same name are one "group" to it. The collectors join on group_id on the monitored server, but don't store it, so the store has no AG identity beyond its name.

Fix options

1. No schema change: count a group as (AG name + its set of replica server names), both of which the cards already carry. Two RDS instances' RDSAG0 have different replica sets, so they count separately. Two monitored replicas of the same real AG share a replica set, so they still count once. That's the case the "views" count exists for.
2. Store the AG's group_id (a collector change + a store migration) and count by it. That's exact, but larger.

Option 1 fixes the header with no migration. Pin it with a fixture of 42 same-named single-AG servers → "42 groups", and two replicas of one AG → "1 group · 2 views".

Related: #4228, #4255 (the AG reads), #4471 (the MCP tool over the same data).

[erikdarlingdata]

Claude posting for Erik Darling.

Fixed by #4481 (2fa4bae). The AG count now identifies an Availability Group by its name together with its replicas. Cards with the same name whose replica sets overlap count as one group, so AGs that share a name but have different replicas (every Amazon RDS for SQL Server Multi-AZ instance's internal RDSAG0) count separately. An AG monitored from both its primary and a secondary still counts once, because a secondary reports only itself. The AG tab header in the Darling Viewer and Lite, and get_ag_health's distinct_ag_count, share the same count. Closing.

[erikdarlingdata]

Claude posting for Erik Darling.

The follow-up is merged as #4495 (087fdc5), store rung V151. The AG collectors now capture each Availability Group's group_id, and the AG count groups by it. So two monitored secondaries of one AG with no monitored primary count as one group, and AGs that share a name but have different group_ids count separately. Rows collected before the upgrade fall back to the name-plus-overlapping-replicas rule from #4481. This covers the Viewer and Lite AG headers and get_ag_health's distinct_ag_count.