Replication tests proposal

[Emmanuel Lecharny <el...@gmail.com>]

Hi guys,

we have some unit tests for replication, but at some point, we would 
like to test replication in a long run scenario.

I would suggest we build a scenario where 3 servers get replicated :
[A] will be the master
[B] will be a slave for [A] and a master for [C]
[C] will be a slave for [B]

[A] -->> [B] -->> [C]

We will use a few clients, injecting data into [A].
Each client will send Add, Delete, Modify, Move, Rename, Move&Rename 
operation
We will use 10 clients, each client will inject Nx5 entries in a row.

Entry 1 will then be deleted
Entry 2 will be modified
Entry 3 will me moved
Entry 4 will be renamed
Entry 5 will be moved and renamed

(we will use a modulo operation to deal with entries, so for instance 
Entry 3 will be the entry which number modulo 5 == 3)

Every hour, we stop injecting modifications for 10 minutes, then we 
compare server [A], [B] and [C] to check that they contain the exact 
same entries.

The test will run for 24 hours.

Do you think it's a decent scenario ?

-- 
Regards,
Cordialement,
Emmanuel Lécharny
www.iktek.com