Hello Postgres community.
In a few words I'm figuring out this stats from the background writer where
you can see buffers_clean is 0 and buffers_backend = 44849371, I would like
it to be the other way around, that's more efficient, right?
-[ RECORD 1 ]---------+------------------------------
checkpoints_timed | 97504
checkpoints_req | 4
checkpoint_write_time | 21171407908
checkpoint_sync_time | 684641
buffers_checkpoint | 273934008
buffers_clean | 0
maxwritten_clean | 0
buffers_backend | 44849371
buffers_backend_fsync | 0
buffers_alloc | 44744713
stats_reset | 2019-09-10 08:42:34.490899-07
Using default settings
postgres=# select name,setting,unit from pg_settings where name like
'%bgwriter%'
;
name | setting | unit
-------------------------+---------+------
bgwriter_delay | 200 | ms
bgwriter_flush_after | 64 | 8kB
bgwriter_lru_maxpages | 100 |
bgwriter_lru_multiplier | 2 |
shared_buffers = '6GB'
and the VM has a RAM of 24GB
The working set is about 160MB, I know, shared buffers are oversized for
this DB, and I know that the bgwriter is supposed to do work when the
working set doesn't fit into shared_buffers, but backends cleaning their
buffers produces undesirable sync waits, right ?
Any advice ?
Thank you.
--
El genio es 1% inspiración y 99% transpiración.
Thomas Alva Edison
