[feat] High Availability with Replicas and Failover Sync

[ajuijas]

Implementing Quorum-Based Replication

This implementation may increase latency for both reads and writes. Since the client will write and read from nodes asynchronously, the increase in latency should not be significant. However, a test plan to evaluate these latency impacts is required.

How writes should work:

• The client decides a value m (where m < n).
• A write is considered successful when at least m nodes acknowledge it.
• The write then continues asynchronously until all n nodes store the value.

How reads should work:

• The client decides a value p (where p < n).
• A read fails if p nodes return conflicting values.

Handling Failover:
If a node goes down and later comes back online, it should sync itself with the latest data from other nodes.

[ajuijas]

The above mentioned mechanism is handled from the client. So let's create a client to use my db.

[ajuijas]

Client package in go ✅

[ajuijas]

#23

Sync db while different responses for same operation