Hey, I have a very big toasted table in my db(9.2.5). Autovacuum doesnt gather statistics on it because the analyze_scale/threshold are default and as a result autoanalyze is never run and the statistics are wrong :
select * from pg_stat_all_Tables where relname='pg_toast_13488395'; -[ RECORD 1 ]-----+------------------------------ relid | 13388396 schemaname | pg_toast relname | pg_toast_13488395 seq_scan | 42 seq_tup_read | 71163925 idx_scan | 5374497 idx_tup_fetch | 2530272449 n_tup_ins | 1253680014 n_tup_upd | 0 n_tup_del | 1253658363 n_tup_hot_upd | 0 n_live_tup | 49425717 *wrong* n_dead_tup | 7822920 *wrong* last_vacuum | last_autovacuum | 2019-02-12 20:54:44.247083-05 last_analyze | *last_autoanalyze |* vacuum_count | 0 autovacuum_count | 3747 analyze_count | 0 autoanalyze_count | 0 When I try to set the both the scale_factor / threshold I'm getting the next error : alter table orig_table set (toast.autovacuum_analyze_scale_factor=0); ERROR: unrecognized parameter "autovacuum_analyze_scale_factor" Any idea why ? I didn't find in the release documentations a note for a similar bug.
