Thread: pg_buffercache description
Hi, I found that the manual page for pg_buffercache module explains "usagecount" as "Page LRU count". http://www.postgresql.org/docs/devel/static/pgbuffercache.html However, AFAIK, "usagecount" is not a LRU count nowadays, because LRU is no longer used for the buffer management algorithm. I think it should be described as "Clock-sweep count" or something like that. And I'd like to add some additional explanation for the "forknumber" as below: "0 for heap file, 1 for FreeSpaceMap, and 2 for VisibilityMap". Any comments? Regards, -- Satoshi Nagayasu <snaga@uptime.jp> Uptime Technologies, LLC. http://www.uptime.jp
On Sat, Nov 24, 2012 at 11:57:51AM +0900, Satoshi Nagayasu wrote: > Hi, > > I found that the manual page for pg_buffercache module explains > "usagecount" as "Page LRU count". > > http://www.postgresql.org/docs/devel/static/pgbuffercache.html > > However, AFAIK, "usagecount" is not a LRU count nowadays, because > LRU is no longer used for the buffer management algorithm. > > I think it should be described as "Clock-sweep count" or something > like that. > > And I'd like to add some additional explanation for the "forknumber" > as below: > > "0 for heap file, 1 for FreeSpaceMap, and 2 for VisibilityMap". > > Any comments? Good suggestion. I ended up referencing the C file that contains the "forknumber" definitions, rather than recreating the list in the docs. I applied the attached patch to head and 9.3. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + It's impossible for everything to be true. +
Attachment
On Wed, Jul 3, 2013 at 02:23:11PM -0400, Bruce Momjian wrote: > On Sat, Nov 24, 2012 at 11:57:51AM +0900, Satoshi Nagayasu wrote: > > Hi, > > > > I found that the manual page for pg_buffercache module explains > > "usagecount" as "Page LRU count". > > > > http://www.postgresql.org/docs/devel/static/pgbuffercache.html > > > > However, AFAIK, "usagecount" is not a LRU count nowadays, because > > LRU is no longer used for the buffer management algorithm. > > > > I think it should be described as "Clock-sweep count" or something > > like that. > > > > And I'd like to add some additional explanation for the "forknumber" > > as below: > > > > "0 for heap file, 1 for FreeSpaceMap, and 2 for VisibilityMap". > > > > Any comments? > > Good suggestion. I ended up referencing the C file that contains the > "forknumber" definitions, rather than recreating the list in the docs. > I applied the attached patch to head and 9.3. I ended up applying it to 9.2 as well. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + It's impossible for everything to be true. +
Thanks, Bruce! 2013/07/04 3:26, Bruce Momjian wrote: > On Wed, Jul 3, 2013 at 02:23:11PM -0400, Bruce Momjian wrote: >> On Sat, Nov 24, 2012 at 11:57:51AM +0900, Satoshi Nagayasu wrote: >>> Hi, >>> >>> I found that the manual page for pg_buffercache module explains >>> "usagecount" as "Page LRU count". >>> >>> http://www.postgresql.org/docs/devel/static/pgbuffercache.html >>> >>> However, AFAIK, "usagecount" is not a LRU count nowadays, because >>> LRU is no longer used for the buffer management algorithm. >>> >>> I think it should be described as "Clock-sweep count" or something >>> like that. >>> >>> And I'd like to add some additional explanation for the "forknumber" >>> as below: >>> >>> "0 for heap file, 1 for FreeSpaceMap, and 2 for VisibilityMap". >>> >>> Any comments? >> >> Good suggestion. I ended up referencing the C file that contains the >> "forknumber" definitions, rather than recreating the list in the docs. >> I applied the attached patch to head and 9.3. > > I ended up applying it to 9.2 as well. > -- Satoshi Nagayasu <snaga@uptime.jp> Uptime Technologies, LLC. http://www.uptime.jp