A customer of mine has tried to snapshot a large number of
tables (roughly 120), to many nodes (12), using a tiny
refresh interval for most of them (30 minutes). It seems
that something has gone wrong with the master tables's log
entries, and they've usually got(ten) a REFRESH FAST
CANNOT BE USED on at least one table a day. They've tried
to protect the logs by creating fake snapshots, one for
each table, having a large refresh interval. The log
entries would take longer to get deleted, they've thought,
thus avoiding unwanted complete refreshes, but things
hasn't worked that fine.
Any clues? suggestions? They've been thinking about
writing snapshot refresh procedures themselves...
--
Francisco C Piragibe de Almeida
e-mail: piragibe_at_esquadro.com.br