Lab · SQL Server distributed availability group

Where the log piles up: replay the lab

Four recorded runs of a six-replica distributed availability group. Pick a scenario, press Play, and watch where the transaction log waits between the primary and each secondary.

Lab companion to the newsletter post SQL Server Availability Group Synchronization State as You’ve Never Seen It.

Recorded November 17, 2026 · sampled every 2 seconds · real data from the lab, server names replaced

How to read it

The lab has two availability groups of three replicas each, joined by an asynchronous distributed availability group. node-1 is the global primary, node-4 is the forwarder. Log travels on five paths: P1 and P2 inside the first group, P3 between the groups, P4 and P5 inside the second group. On every path, log that has not yet been applied on the secondary is in one of four places:

The size of each stage is evidence of where the log waits. It does not name the cause by itself: the sentence under the controls adds what the other counters say.

Pick a scenario

What was done

What to watch

run - sample - load -
Loading…

Flow, primary to secondary

senderunsentin flightreceived, not hardenedhardened, not redonereceiver

Backlog per path, by stage

Rate at each point, MB/s

Backlog over the run

Numbers at this sample

Events (click to jump)

About the lab