OK I know I am going way back.
About 11 years ago I created a replicated database with clusters of three r=
eplicas in each of two remote offices. Users at those offices direct repli=
cated to their local master via high speed LAN. Periodically, the clusters=
sync themselves. Also periodically, the remote clusters replicated with d=
irect replication over a leased line. This worked amazingly well over the =
past 11 years.
Recently, the company has abandoned the two remote sites and has opened a n=
ew location. They have a VPN (Internet based) between it and HQ. They are=
doing direct replication over it. Most of the time it works. However, tw=
o times in 45 days one cluster at HQ has become corrupted.
Is it reasonable to try to do direct replication over a fast, but VPN/inter=