Обсуждение: Blown Index? Corrupted Table?
I've been getting this on and off in my vaccums. Once it happens, the only way to make it go away seems to be a dump/restore. Is this bad, Bad, or BAD? NOTICE: Index idx_rawbebugs_assignedto: Pages 112; Tuples 31161: Deleted 0. CPU 0.01s/0.10u sec. NOTICE: Rel rawbebugs: Pages: 4435 --> 4400; Tuple(s) moved: 0. CPU 0.04s/0.01u sec. NOTICE: FlushRelationBuffers(rawbebugs, 4400): block 287 is referenced (private 0, global 1) FATAL 1: VACUUM (vc_repair_frag): FlushRelationBuffers returned -2 -- Adam Haberlach | A billion hours ago, human life appeared on adam@newsnipple.com | earth. A billion minutes ago, Christianity http://www.newsnipple.com | emerged. A billion Coca-Colas ago was '88 EX500 | yesterday morning. -1996 Coca-Cola Ann. Rpt.
On Fri, Oct 06, 2000 at 12:29:29PM -0700, Adam Haberlach wrote: > I've been getting this on and off in my vaccums. Once it happens, the only way to > make it go away seems to be a dump/restore. Is this bad, Bad, or BAD? > > NOTICE: Index idx_rawbebugs_assignedto: Pages 112; Tuples 31161: Deleted 0. CPU 0.01s/0.10u sec. > NOTICE: Rel rawbebugs: Pages: 4435 --> 4400; Tuple(s) moved: 0. CPU 0.04s/0.01u sec. > NOTICE: FlushRelationBuffers(rawbebugs, 4400): block 287 is referenced (private 0, global 1) > FATAL 1: VACUUM (vc_repair_frag): FlushRelationBuffers returned -2 Ok. It looks like I have one or more Relations pinned in my cache, according to the source code. Do I need to kick all of the connections out of the database in order to vaccum? AFAIK, there isn't anyone accessing at the time (and I tried shutting down httpd while running the vacuum). -- Adam Haberlach | A billion hours ago, human life appeared on adam@newsnipple.com | earth. A billion minutes ago, Christianity http://www.newsnipple.com | emerged. A billion Coca-Colas ago was '88 EX500 | yesterday morning. -1996 Coca-Cola Ann. Rpt.
Adam Haberlach <adam@newsnipple.com> writes: > I've been getting this on and off in my vaccums. Once it happens, the only way to > make it go away seems to be a dump/restore. Is this bad, Bad, or BAD? > NOTICE: FlushRelationBuffers(rawbebugs, 4400): block 287 is referenced (private 0, global 1) > FATAL 1: VACUUM (vc_repair_frag): FlushRelationBuffers returned -2 It's just in the "annoyance" category: to clear the condition you need to stop and restart the postmaster. No dump/reload or anything like that, you just want it to reset shared memory. The most likely cause of this problem is a backend exiting from an open transaction without doing AbortTransaction processing. 7.0.* has a bug that a client disconnect without closing a transaction block will result in that happening. I posted a patch for it on the patches list a few weeks ago, or you could pull the current REL7_0_PATCHES branch (7.0.3-to-be) from the CVS server. regards, tom lane