[Linux-cluster] Using GFS on Compact Flash

[Lin Shen (lshen)]

> -----Original Message-----
> From: linux-cluster-bounces at redhat.com 
> [mailto:linux-cluster-bounces at redhat.com] On Behalf Of Robert Peterson
> Sent: Tuesday, March 13, 2007 11:16 AM
> To: linux clustering
> Subject: Re: [Linux-cluster] Using GFS on Compact Flash
> 
> Lin Shen (lshen) wrote:
> > Does it make sense to use  GFS (local or cluster mode) on 
> Compact Flash?
> > Will it greatly reduce the life expectancy of the Compact Flash 
> > compared to using a local file system? The rational behind this is 
> > that GFS will issue way more writes to the disk for its internal 
> > operations ( such as dlm locking etc).
> > 
> > Lin   
> Hi Lin,
> 
> Here are my thoughts:
> 
> I'm not aware of anyone using compact flash with GFS.
> GFS has no wear-leveling, so life expectancy might be an issue.
> The same goes for most file systems except those specifically 
> written for CF, like jffs2.
> 
> The file and directory data isn't a concern: Linux page cache 
> should manage the data buffers normally.  The cluster locks 
> and glocks are also not a concern, since everything is 
> managed in memory.  However, the GFS Resource Groups (RGs) 
> (Not to be confused with rgmanager's resource groups) are a 
> bigger concern.  Part of the RGs have bitmaps of blocks to 
> indicate which blocks are allocated and that may be rewritten 
> many times as blocks are allocated and released.  However, I 
> haven't studied how often these are actually written back to 
> the media.
> 
> Another concern are the journal areas, which are being 
> written over and over.
> They aren't as bad as the RGs, though, because the journals 
> have lots of space, and therefore it's not always the same 
> block getting written over and over.  The journals may be 
> forced to disk more though, because if a node crashes for 
> whatever reason, the other nodes need access to the journals 
> to replay the node's data to ensure file system integrity.
> 
> Regards,
> 
> Bob Peterson
> Red Hat Cluster Suite

Hi Bob,

This is very good info. The CF we use has builtin wear-levelling, wonder
if this will make running GFS on top of it more resonable. Of course, if
GFS is like FAT in the way that file system structure is frequently
re-written in-place and cannot be reallocated or moved after wear
failure, then the builtin wear-levelling is not good enough. 

Since we'll need to run a file system on the CF, and most likely one of
either GFS, Reiser or FAT, we'd really like to compare the extra disk
writes introduced by the various file systems. I know it's hard to get
the exact numbers, but do you think GFS is far more worse in this
regard. 

BTW, is there also a big difference when GFS is running in local and
clustered mode?

Last, does GFS issue any disk writes when it's idle, meaning no one is
using the file system? Do the cluster and fencing stuff also introduce
disk writes?

Lin     


> 
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster
>