Re: POSTGRES_FSM_RELATIONS CRITICAL: DB control fsm relations used:79569 of 80000 (99%)

Поиск
Список
Период
Сортировка
От Adrian Klaver
Тема Re: POSTGRES_FSM_RELATIONS CRITICAL: DB control fsm relations used:79569 of 80000 (99%)
Дата
Msg-id 804b26fe-269e-82d7-01ef-86da933a6566@aklaver.com
обсуждение исходный текст
Ответ на POSTGRES_FSM_RELATIONS CRITICAL: DB control fsm relations used: 79569of 80000 (99%)  (Julie Nishimura <juliezain@hotmail.com>)
Ответы Re: POSTGRES_FSM_RELATIONS CRITICAL: DB control fsm relations used:79569 of 80000 (99%)  (Julie Nishimura <juliezain@hotmail.com>)
Список pgsql-general
On 5/23/19 11:57 PM, Julie Nishimura wrote:
> Hello,
> We have an issue with fsm_relations utilization reaching 99%, I was able 

How are you arriving at the above percentage?

How many tables/indexes do you have in the database(s)?

> to vacuum a handful of tables, but it wasn't enough to make a noticeable 
> difference. I think at this point we will need to increase the number of 
> fsm_relations from 80,000 to 100,000 which will require a restart. 
> Because there aren't any more dead rows to delete. I confirmed this by 
> connecting to each db and running the following query:
> SELECT relname, n_live_tup, n_dead_tup from pg_stat_user_tables order by 
> n_dead_tup desc
> daily_mail_shared_state_cddt_3588-
> relname | n_live_tup | n_dead_tup
> ----------------+------------+------------
> article_errors | 0 | 0
> article_names | 3375193 | 0
> indexdefs | 0 | 0
> tabledefs | 0 | 0
> 
> Above output, shows n_dead_tup is zeroed out, this makes me believe that 
> we need to increase the number of fsm relations to a number between 90k 
> and 100k.But I might be wrong, need your advice.
> 
> PostgreSQL 8.2.15 (Greenplum Database 4.3.8.1 build 1)
> 
> Thanks!


-- 
Adrian Klaver
adrian.klaver@aklaver.com



В списке pgsql-general по дате отправления:

Предыдущее
От: Luca Ferrari
Дата:
Сообщение: Re: pg_stat_user_tables.n_tup_ins empty for partitioned table
Следующее
От: Jeremy Finzel
Дата:
Сообщение: Questions about btree_gin vs btree_gist for low cardinality columns