I would like to analyze server stats offline, so I attempt to pg_dump my
production database and then pg_restore it into another database. In the
process all stats seem to be reset (they are not completely zeroed). So
in production I have a table with the following stats (from
pg_stat_all_tables as an example):
relid | 25519576
relname | property_contact
seq_scan | 5612
seq_tup_read | 569971320
idx_scan | 4486454
idx_tup_fetch | 180100369
n_tup_ins | 39114
n_tup_upd | 17553
n_tup_del | 21877
After I restore the stats for the same table look like this:
relid | 104017313
relname | property_contact
seq_scan | 9
seq_tup_read | 992493
idx_scan | 0
idx_tup_fetch | 0
n_tup_ins | 110277
n_tup_upd | 0
n_tup_del | 0
These look like stats for table accesses during the restore itself:
11027 is indeed the number of rows in the table, and 992493 / 110277 =
9, which happens to be the number of indexes and FK constraints on the
table.
I do have stats_reset_on_server_start = off on both servers.
Can someone share what exatly happens with stats upon restore? Also is
there anything one can do to keep them intact during a dump/restore?
Apologies if already discussed--I failed to find any references.
TIA,
George
