C
ChUIMonster
Guest
If replication is involved then "it depends". The db that my 468 million row example comes from is almost exactly 1TB. I can back that up in about 90 minutes and restore it in a bit less than an hour. So that still fits pretty easily into a 10 hour outrage window. As it happens I do not generally have 10 hours to work with. So in reality I would probably add it inactive, leave replication running, and build it online with a combination of idxfix and idxactivate. See this: pugchallenge.org/.../318_IndexMaintenance.pdf OTOH, as I recall, you are on an old klunky HPUX server with some sort of horrifically awful SAN (which is sort of from the department of redundant statements department). There's an argument to be made that you could get migrated to Linux before that index will be built.
Continue reading...
Continue reading...