Re: Concurrency bug with vacuum full (cluster) and toast

2019-04-03 Thread Michael Paquier
On Wed, Apr 03, 2019 at 11:26:20AM -0400, Stephen Frost wrote: > If you're referring to the v12 open items list, then, no, I wouldn't > think it would be as it's not a new issue (unless I've misunderstood). > Only regressions from prior versions are appropriate for the v12 open > items list, long-s

Re: Concurrency bug with vacuum full (cluster) and toast

2019-04-03 Thread Stephen Frost
Greetings, * Justin Pryzby (pry...@telsasoft.com) wrote: > On Fri, Mar 22, 2019 at 02:27:07PM -0400, Robert Haas wrote: > > On Tue, Mar 19, 2019 at 1:37 PM Alexander Korotkov > > wrote: > > > Thank you for pointing, but none of the threads you pointed describe > > > this exact problem. Now I see

Re: Concurrency bug with vacuum full (cluster) and toast

2019-04-03 Thread Justin Pryzby
Hi, On Fri, Mar 22, 2019 at 02:27:07PM -0400, Robert Haas wrote: > On Tue, Mar 19, 2019 at 1:37 PM Alexander Korotkov > wrote: > > Thank you for pointing, but none of the threads you pointed describe > > this exact problem. Now I see this bug have a set of cute siblings :) > > Yeah. I really t

Re: Concurrency bug with vacuum full (cluster) and toast

2019-03-22 Thread Robert Haas
On Tue, Mar 19, 2019 at 1:37 PM Alexander Korotkov wrote: > Thank you for pointing, but none of the threads you pointed describe > this exact problem. Now I see this bug have a set of cute siblings :) Yeah. I really thought this precise issue -- the interlocking between the VACUUM of the main t

Re: Concurrency bug with vacuum full (cluster) and toast

2019-03-19 Thread Alexander Korotkov
On Tue, Mar 19, 2019 at 6:48 PM Robert Haas wrote: > On Mon, Mar 18, 2019 at 12:53 PM Alexander Korotkov > wrote: > > I've discovered bug, when vacuum full fails with error, because it > > couldn't find toast chunks deleted by itself. That happens because > > cluster_rel() sets OldestXmin, but t

Re: Concurrency bug with vacuum full (cluster) and toast

2019-03-19 Thread Robert Haas
On Mon, Mar 18, 2019 at 12:53 PM Alexander Korotkov wrote: > I've discovered bug, when vacuum full fails with error, because it > couldn't find toast chunks deleted by itself. That happens because > cluster_rel() sets OldestXmin, but toast accesses gets snapshot later > and independently. That c

Concurrency bug with vacuum full (cluster) and toast

2019-03-18 Thread Alexander Korotkov
Hi all, I've discovered bug, when vacuum full fails with error, because it couldn't find toast chunks deleted by itself. That happens because cluster_rel() sets OldestXmin, but toast accesses gets snapshot later and independently. That causes heap_page_prune_opt() to clean chunks, which rebuild_