Usynced blocks if replication is interrupted during initial sync

[Tim Westbrook]

A follow up to my previous message, we are still seeing unsynced blocks on re-connect, and it happens with just a single secondary


1. Create config on primary with only 1 node (itself) as primary
2. Add another node by modifying the primary config file , and performing an adjust
3. Update the remote node config, up drbd
4. Wait for initial sync to start
5. Disable network interface on primary (drbd driver reports disconnect)
6. Wait for some time
7. Enable network interface on primary (drbd reconnects and reports remote disk is up to date)
8. Run verify, reports thousands of out of sync blocks

Do we need to manually perform an invalidate step every time we bring up an additional node? It seems like drbd should be aware of the unsynced blocks and restart the sync process?

Could this be an issue with the driver? or a misunderstanding on my part about how block syncing is tracked. 

Cheers,
Tim