On Wed, Feb 28, 2007 at 02:29:52PM -0800, Joshua D. Drake wrote: > > > I don't know, but in my opinion, I don't see anything bad in requiring > > dropping the data if the quota is full. That's what usually occurs in > > the case of normal filesystem quota... If you don't have a space there, > > you cannot edit files, copy them etc... > > And that solution should be definitely better than the filesystem quota > > for the PostgreSQL user for example. > > The bad point is not that we would rollback the transaction. The bad > point is what happens when you need to rollback a transaction and in > your scenario it is quite plausible that a large rollback could occur, > more than once, causing the requirement of something like a vacuum full > to clean things up.
ISTM that if the transaction is that big it's likely going to be extending the heap, which means you'd get space back on a plain vacuum. As for things like CLUSTER, and REINDEX it would probably be useful to make an exception, since we know that those operations are intended to shrink the size of a relation. I also think there's a lot to be said for a soft limit. -- Jim Nasby [EMAIL PROTECTED] EnterpriseDB http://enterprisedb.com 512.569.9461 (cell) ---------------------------(end of broadcast)--------------------------- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match