Roust_m (5/31/2016)

---

TheSQLGuru (5/30/2016)

---

Replicas being offline will NOT cause a transaction on the primary to fail. They will simply queue up on the primary and catch-up when the secondary that was offline become available again (assuming just a transient disconnect/outage where the secondary is still in it's pre-outage state SQL Server wise). Do note that your transaction log will be unable to flush committed transactions from VLFs until the secondary(s) are caught up though.

Please note that doing failover cluster instances and AGs is all kinds of persnickety.

I thought the definition of a synchronous replica is the one, that will fail the transaction on the primary replica if it can't commit on the secondary. Asynchronous replica will continue to queue up until the link is available again.

They are not failover instances, they can't fail over from one node to another. Instead there are 4 independent instances, each having either a primary or secondary replica. The failover will be happening on the availability group level. It is my understanding that AG need windows cluster to operate.