playground
Replication, Lag & Failover
Pick a replication mode, set the lag, then kill the primary. Watch what the trade-off you chose actually costs — in write latency while healthy, and in lost writes when it isn't.
replication guidestep 1/3
→ On async mode, set lag to ~2000ms and writes to ~1000/s, then kill the primary.
waiting for you to try it…
write p50 (healthy)
5ms
failover window
~8s
writes lost on failover
1,000
▣
primary
async · 2000ms behind
▣
replica
✓ healthy — writes acknowledged in 5ms
The takeaway:replication mode is a contract about which failure you accept — async risks the lag window's writes, sync chains your latency to the slowest replica, and nothing removes the failover window itself.
Read the replication lesson →