Re: Verifying Consistency across dbs

[Tsz Wo Sze <sz...@gmail.com>]

Hi Asad,

I guess there are many ways to do the detection.  One way is to compare the
checksum of the current Raft log (equivalently, the checksum of the current
state) as below:
- For every log index i, compute checksum_i on each server.  Note
that checksum_i is the checksum for the entire log from index 0 to index i.
- Each server sends (i, checksum_i) to the other servers.
- When a server receives (i, checksum_i) from another server.  Compare the
checksum with its local value.

We may incorporate the checksum in appendEntries -- the leader includes the
checksum_i to the log entry with index i.  Then, the followers compare the
received checksum_i with their local calculation.

Hope this helps!
Tsz-Wo


On Sat, Jul 30, 2022 at 6:55 AM Asad Awadia <as...@gmail.com> wrote:

> Hello,
>
> Not a ratis specific question but let's say I am using raft to replicate
> KV across 3 nodes. Simple in memory hash map.
>
> In theory they should all be at the same state eventually but how does one
> verify that assumption? Is my DB consistent across replication?
>
> How to detect nodes not in sync? And if they are in sync? While live
> traffic may be getti g served?
>
> Regards,
> Asad
>
>