Re: Postgres on RAID5 - Mailing list pgsql-performance

From Alexander Kirpa
Subject Re: Postgres on RAID5
Date
Msg-id 42326E49.18978.3A504719@localhost
Whole thread Raw
In response to Postgres on RAID5  (Arshavir Grigorian <ag@m-cam.com>)
List pgsql-performance
Hi Arshavir Grigorian,

0. If possible move to 8.0.1 - bgwriter help you

1. Create RAID1 for redo and place drives on separate
SCSI channel

2. Update postgresql.conf:
shared_buffers = 10000-50000
work_mem             = 100000-300000
maintenance_work_mem = 100000-300000
max_fsm_pages = 1500000
max_fsm_relations = 16000
wal_buffers = 32
checkpoint_segments = 32 # 16MB each !!
checkpoint_timeout = 600
checkpoint_warning = 60
effective_cache_size = 128000
random_page_cost = 3
default_statistics_target = 100
log_min_error_statement = warning
log_min_duration_statement = 1000 # for logging long SQL

3. If possible migrate from RAID5 to RAID10.

4. Add (if need) 2 new drive for OS and use ALL
20x9GB drive for DB storage.

5. Remove CDROM from work configuration and start use
this scsi channel.

Best regards,
 Alexander Kirpa


pgsql-performance by date:

Previous
From: Daniel Schuchardt
Date:
Subject: Re: cpu_tuple_cost
Next
From: "Michael McFarland"
Date:
Subject: Re: adding 'limit' leads to very slow query