William Dunn wrote:
In terms of benefit over a lagging replica Flashback has the benefit of being
transparent to the user
(the user can query over the same database connection, etc), it does not
incur the full cost of having
a replica...
Yes, Flashback (in all ist forms) is something that
Hello Laurenz,
Thanks for your feedback. Actually when I said that the same overhead
occurs in Oracle I was referring to bloat in the UNDO logs, which similar
to PostgreSQL dead tuples has some impact on size/performance and is the
majority of the trade-off considered when deciding to implement
William Dunn wrote:
Just had an idea and could use some feedback. If we start a transaction,
leave it idle, and use
pg_export_snapshot() to get its snapshot_id MVCC will hold all the tuples as
of that transaction's
start and any other transaction can see the state of the database as of that
Hello,
Just had an idea and could use some feedback. If we start a transaction,
leave it idle, and use pg_export_snapshot() to get its snapshot_id MVCC
will hold all the tuples as of that transaction's start and any other
transaction can see the state of the database as of that time using SET
