Books on writing distributed systems

[Peter Corlett]

On Thu, Mar 27, 2008 at 01:37:02PM +0000, Paul LeoNerd Evans wrote:
[...]
> Would this be the replication that, for all databases run by a daemon,
> just writes the text of SQL queries into some journal file, that another
> process reads and streams across a network, to a third process on the
> other side, that replays them into its copy of the data with the mere hope
> that the same query will apply, and all will be well? That one that breaks
> for us about once a week... That replication?

It merely breaks for you? You're getting off light. The usual failure mode
is silent corruption which is only noticed about a year later when the
master blows up and all the backup tapes have been recycled.