Replicated-ZEO promises great ways to deal with a storage server being destroyed; It should transparently avoid avoiding downtime, and prevent data loss even if a storage server is destroyed. Unfortunately, Replicated ZEO is not yet released (as of July 2001).

Until replicated-ZEO is released, ReplicatedFileStorage provides an easy way of preventing data loss, by ensuring that data is safely synced onto multiple disks before allowing a ZODB transaction to commit.

ReplicatedFileStorage has three main advantages over other OS-level solutions to this problem (such as network RAID):

• Minimal administrative hassle.
• Easily add and remove replicas.
• Flexibility in configuring how many replicas are necessary.

If you are reading this after replicated-ZEO has been released then you probably should be using that instead.

This product has been tested so far on Linux and Windows. On Linux it works great, but on Windows it never proved reliable. I assume someone with more experience of async sockets on windows should be able to fix it up; please let me know if you do..

Current Status

This product "works", and it has saved me from a couple of disasters, but I am no longer using it internally so please consider this product unsupported. I would be happy for anyone else to take up maintenance of ReplicatedFileStorage... any takers?