[prev in list] [next in list] [prev in thread] [next in thread] 

List:       netbsd-users
Subject:    Re: setting up raid5
From:       Steven Bellovin <smb () cs ! columbia ! edu>
Date:       2010-05-31 14:16:57
Message-ID: B204A75E-3743-4270-816E-CBCF4549A19C () cs ! columbia ! edu
[Download RAW message or body]


On May 31, 2010, at 10:11 32AM, Manuel Bouyer wrote:

> On Mon, May 31, 2010 at 10:06:39AM -0400, Steven Bellovin wrote:
>>>> 
>>>> I'm not 100% sure, but I think you're limited to 128 sectors per
>>>> stripe unit (64kB or MAXPHYS) here.  Maybe try the "32 1 1 5" as
>>>> per the example first, and bump higher when the rest is working.
>>> 
>>> Yes, this is likely the problem. raidframe will schedule a transfer
>>> larger than MAXPHYS, and the underlying driver will bail out.
>> 
>> Thanks, I'll try that, though it leaves unanswered why the first two
>> drives had no problem -- I was even able to create a file system on
>> them.
> 
> I can immagine it blows up when trying to write the first full
> parity stripe. Once the drive is failed it doens't try to do it again with
> other drives.
> 

OK.

It's running now with 64 sectors/stripe, and seems much happier.  Thanks.
(It's probably worth a documentation fix to the man page to mention that
limit, and to the code to check.  I'll also send-pr for documentation changes
to gpt and dkctl to make them *much* clearer in a few crucial spots.)


		--Steve Bellovin, http://www.cs.columbia.edu/~smb





[prev in list] [next in list] [prev in thread] [next in thread]