I have a couple of questions about how streaming replication works in the presence of a high timeout for query
cancellation:
1. Where is the WAL data stored while waiting for the query cancellation, the primary or the secondary? In case of a
crashduring this interval, could the secondary end up behind the primary due to the WAL data not having been sent over?
2. Can a very high query cancellation value run afoul of wal_keep_segments, or will the primary (assuming it's the one
doingthe storing) keep that data around until the query completes or is cancelled?
Thanks!
--
-- Christophe Pettus
xof@thebuild.com