On Thu, Dec 10, 2015 at 12:55 PM, Greg Stark <stark@mit.edu> wrote:
> On Thu, Dec 10, 2015 at 4:47 PM, Robert Haas <robertmhaas@gmail.com> wrote:
>> It's not straightforward, but I don't think that's the reason. What
>> we could do is look at the call sites that use
>> RelationGetNumberOfBlocks() and change some of them to get the
>> information some other way instead. I believe get_relation_info() and
>> initscan() are the primary culprits, accounting for some enormous
>> percentage of the system calls we do on a read-only pgbench workload.
>> Those functions certainly know enough to consult a metapage if we had
>> such a thing.
>
> Would this not run into a chicken and egg problem with recovery?
No.
> Unless you're going to fsync the meta page whenever the file is
> extended you'll have to xlog any updates to it and treat the values in
> memory as authoritative. But when replaying xlog you'll see obsolete
> inconsistent versions on disk and won't have the correct values in
> memory either.
All true, but irrelevant. The places I just mentioned wouldn't get
called during recovery. They're only invoked from queries.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company