Replication Lag

What it is

Replication lag is the time it takes for a change to move from the primary to a replica.

During that window, two reads can legitimately see different states.

When it matters

This matters when the application writes in one place and reads from another right after.

It is especially sensitive in:

• profile edits
• balances
• inventory
• permissions

Common mistake

The classic mistake is treating every stale read as a write failure.

Sometimes the write worked perfectly.

What has not caught up yet is replication.

Short example

A user saves a new address.

The write goes to the primary.

Right after that, the UI reads from a replica and still shows the old value for a few seconds.

Why it helps

Understanding replication lag helps you separate:

• a real persistence bug
• a normal convergence delay

It also helps product and architecture decide where to read from the primary and where delay is acceptable.

Replication lag is not mysterious. It is the cost of not demanding immediate consistency from every read.