Thread: Does pg_reset_stats() influence autovacuum?
Hello, Does resetting of stats through pg_reset_stats() influence the way autovacuum determines whether autovacuum_vacuum_scale_factor and autovacuum_analyze_scale_factor settings are reached? In other words, are the n_live_tup and n_dead_tup values kept somewhere internally, apart pg_stat_all_tables? -- Kouber Saparev
On Thu, Aug 23, 2012 at 10:59:56AM +0300, Kouber Saparev wrote: > Hello, > > Does resetting of stats through pg_reset_stats() influence the way > autovacuum determines whether autovacuum_vacuum_scale_factor and > autovacuum_analyze_scale_factor settings are reached? > > In other words, are the n_live_tup and n_dead_tup values kept somewhere > internally, apart pg_stat_all_tables? No. pg_reset_stats() will affect autovacuum, usually in a bad way. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + It's impossible for everything to be true. +
On 08/24/2012 05:47 AM, Bruce Momjian wrote: > On Thu, Aug 23, 2012 at 10:59:56AM +0300, Kouber Saparev wrote: >> Hello, >> >> Does resetting of stats through pg_reset_stats() influence the way >> autovacuum determines whether autovacuum_vacuum_scale_factor and >> autovacuum_analyze_scale_factor settings are reached? >> >> In other words, are the n_live_tup and n_dead_tup values kept somewhere >> internally, apart pg_stat_all_tables? > > No. pg_reset_stats() will affect autovacuum, usually in a bad way. > So, I guess the only safe moment to reset stats is right after manual vacuum analyze. Thanks. -- Kouber Saparev