DragonFly BSD
DragonFly users List (threaded) for 2013-01
[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]

Re: HAMMER best practices?


From: Dave Hayes <dave@xxxxxxxxxxx>
Date: Thu, 10 Jan 2013 21:42:34 -0800

On 01/10/13 18:38, Justin Sherrill wrote:
> My immediate thought is that without RAID, a disk failure means that if a
> disk fails, you will have to manually take a HAMMER slave PFS and change it
> to a master, which should be possible - but will take time.  RAID will
> (usually, hopefully) handle a single disk failure without interruption.

This is exactly the kind of information I was looking for, thanks!

> The Areca cards are well supported, and I think they've even contributed
> hardware for testing.

Interesting idea.

A quick google search turns up a lot of complaints from various people 
that the 1222 cards stop working 18 months after you start using them. 
The rest of those cards (at least the ones listed on the dragonfly 
hardware compatibility page) are either too expensive or too old.

You can get 1222's real cheap, but I guess that's the market reflecting 
the unreliability of the card? I'm speculating there, in case it wasn't 
clear. :D

The real tale of reliability (which includes DragonFlyBSD) is this:

# zgrep twa0 /var/log/*
...kernel: twa0: <3ware 9000 series Storage Controller> port 
0xb800-0xb8ff mem 0xce800000-0xceffffff,0xdfeffc00-0xdfeffcff irq 9 at 
device 2.0 on pci6

# uname -v
DragonFly 1.8.2-RELEASE #0: Thu May 31 18:51:11 PDT 2007

# uptime
  9:27PM  up 322 days...

This machine has been continually handling email since that kernel build 
date. It's only crashed due to power loss.

As this is not my only 3ware card in production, I'm pretty convinced 
3ware cards are the way to go. I'm still open to other data, but it's 
hard to beat this experience.

> Outside of that, I'd suggest building two machines.  Since you can stream
> to a slave PFS over a network, have a that second machine as a remote
> backup with cheap, large disks with a more generous retention policy - that
> way if complete catastrophe strikes, your data's safe.

Good advice. I'd planned to have two machines anyway. Thanks again.
-- 
Dave Hayes - Consultant - Altadena CA, USA - dave@jetcafe.org
 >>>> *The opinions expressed above are entirely my own* <<<<

There is no greater calamity for a nation or individual
than not finding contentment in one's sufficiency.



[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]