From: | Pavel Suderevsky <psuderevsky(at)gmail(dot)com> |
---|---|
To: | Postgres General <pgsql-general(at)postgresql(dot)org> |
Subject: | perc 9 series fastpath feature for database data |
Date: | 2016-08-04 19:37:46 |
Message-ID: | CAEBTBzs4LHMSz0nuSpq7XoiEuUNzf3jijvtmHc_tyYAYSb527g@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
Hi,
I need advice regarding RAID Controller configuration for PostgreSQL data.
Known common best practices:
- RAID cache at least 512MB
- BBU is must have
- cache mode = write back
- disk write cache mode = disabled for hdd and low level ssd and disk write
cache mode = enabled for ssd enterprise level with backup condenser
- barrier=0
- vm.dirty_background_bytes=67108864 and vm.dirty_bytes=536870912 for RAID
with 512MB cache.
But now I want to ask about special configuration for Dell PowerEdge RAID
Controller (PERC) 9 series with 2Gb write cache due to Fastpath feature
listed in Dell's documentation:
> FastPathis a feature that improves application performance by delivering
> high I/O per second (IOPs) for
> the Solid State Drives (SSD).
To enable FastPathon a virtual disk the Dell PowerEdge RAID Controller
> (PERC) 9 series cache policies
> need to be set to Write-Throughand No Read Ahead. This enables FastPathto
> use the proper data path
> through the controller based on command (read/write), IO size, and RAID
> type.
Using dell enterprise level SSD's in RAID 10 for PostgreSQL data is
supposed.
It seems that with this raid controller and ssd's the best practice would
be slightly corrected:
- cache mode = write through
- disk write cache mode = disabled
- barrier=0
- vm.dirty_background_bytes=16777216 and vm.dirty_bytes=134217728
Wouldn't this configuration be dangerous in case of power-loss disaster and
especially isn't it dangerous to use barrier=0 with Write Through raid
cache mode?
Thanks in advance.
From | Date | Subject | |
---|---|---|---|
Next Message | Patrick B | 2016-08-04 19:55:07 | Re: Question about wal files / pg_xlogs |
Previous Message | John R Pierce | 2016-08-04 18:17:26 | Re: Tunning Server 9.1. |