[Linux-cluster] GFS

[Anne M. Hammond]

Hi Adam,
 
Could you characterize the filesystem sizes you used to test?
In other words, when you say "fsck is too slow for large filesystems",
what size is too large?
 
What size filesystem did you test?  What filesystem did you use?

Thanks in advance,
 
Anne Hammond
University of Colorado at Boulder

>Date: Sat, 23 Oct 2004 10:20:28 -0700
>From: Adam Cassar <adam.cassar at netregistry.com.au>
>User-Agent: Mozilla Thunderbird 0.8 (Windows/20040913)
>X-Accept-Language: en-us, en
>MIME-Version: 1.0
>To: Patrick <nawtyness at gmail.com>, linux clistering <linux-cluster at redhat.com>
>Subject: Re: [Linux-cluster] GFS
>Content-Transfer-Encoding: 7bit
>X-RedHat-Spam-Score: 1.953 *
>X-loop: linux-cluster at redhat.com
>Cc: 
>X-BeenThere: linux-cluster at redhat.com
>X-Mailman-Version: 2.1.5
>List-Id: linux clistering <linux-cluster.redhat.com>
>List-Unsubscribe: <http://www.redhat.com/mailman/listinfo/linux-cluster>, 
<mailto:linux-cluster-request at redhat.com?subject=unsubscribe>
>List-Archive: </archives/linux-cluster>
>List-Post: <mailto:linux-cluster at redhat.com>
>List-Help: <mailto:linux-cluster-request at redhat.com?subject=help>
>List-Subscribe: <http://www.redhat.com/mailman/listinfo/linux-cluster>, 
<mailto:linux-cluster-request at redhat.com?subject=subscribe>
>X-Spam-Checker-Version: SpamAssassin 2.63 (2004-01-11) on lcd.colorado.edu
>X-Spam-Level: *
>X-Spam-Status: No, hits=2.0 required=5.0 tests=DATE_IN_FUTURE_12_24  
autolearn=no version=2.63
>
>I've done a lot of testing with a similar set up as you describe about 1 
>month ago and found the following issues:
>
>1) GFS seems to scale well with raw throughput but not with raw file 
>operations (no matter how we tuned the fast)
>2) fsck times are too slow for large file systems in a production 
>environment
>3) GFS has serious issues recovering from various fail modes
>4) Some fail modes result in file system corruption that can even crash 
>a node when attempting to remount the file system
>
>Other than that it does work and the gnbd gives good performance. Maybe 
>the above issues have been sorted out since I last tested. However it is 
>a complex piece of software that I feel requires a lot of work before it 
>can be used reliably in a production environment.
>
>
>Patrick wrote:
>
>>Hi, 
>>
>>Would GFS be able to handle ( or how would i check ) say 10 or more
>>machines over a san ( IBM FAST T600 ) running a web / mail / i/o
>>dependant system ?
>>
>>Thanks
>>  
>>
>
>--
>Linux-cluster mailing list
>Linux-cluster at redhat.com
>http://www.redhat.com/mailman/listinfo/linux-cluster