All of lore.kernel.org
 help / color / mirror / Atom feed
From: Gerrit Hannaert <degerrit@web.de>
To: reiserfs-list@namesys.com
Cc: Juergen Waiblinger <juwa@triplestor.com>
Subject: Re: reiserfs for cluster
Date: Tue, 06 May 2003 23:01:22 +0200	[thread overview]
Message-ID: <3EB822A2.3090005@web.de> (raw)
In-Reply-To: <5.2.0.9.0.20030506160902.03b1da60@pop.puretec.de>

I've tried something like this on a SAN a while ago, using ext3, but
perhaps the results would not be so different using reiserfs. The
situation we were looking at involved flat-file database updates which
would occur only once every day or even less than that.

Obviously the major issue is who has write-access; but things seem to
work if:
- one machine does the writes, after which the other machines will need
to re-mount the filesystem (probably works best if writes affect new
files, and are not appends/modifications...). Otherwise new files will
not show up because this data is cached (I expect someone else can give
some better low-level explanation and possible workarounds)...
- filesystem on read-only machines is mounted with journaling disabled,
and the noatime,ro options (updating access times are writes, too).
Otherwise the filesystem will not be in a consistent state, and will be
checked every time you mount it.

There may be other alternatives, but this is a simple solution if your
'write demands' are not too high. I haven't done much performance or
scalability tests so I don't know if this method is preferrable over NFS
and such (but it's so much cooler!).

Cheers,

- Gerrit


Juergen Waiblinger wrote:

> Hello,
>
> we are running different beowulf clusters and the storage amount
> increases all the time. Until now we have been using Veritas
> Clusterfilesystem, but for different reasons we want to go for an
> other solution. As we are using reiserfs quite happily on the desktops
> we thought about using this for the storagesubsystem attached to the
> masterserver of the cluster.
> there will only be sun and linux machines accessing the files but
> there will be concurrent file access.
> Has anybody tried this or could give some other solutions
> Any information will be appreciated
>
> TIA will summarize to the list
>



      parent reply	other threads:[~2003-05-06 21:01 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2003-05-06 14:13 reiserfs for cluster Juergen Waiblinger
2003-05-06 14:57 ` Ookhoi
2003-05-06 17:09   ` Russell Coker
2003-05-06 21:01 ` Gerrit Hannaert [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=3EB822A2.3090005@web.de \
    --to=degerrit@web.de \
    --cc=juwa@triplestor.com \
    --cc=reiserfs-list@namesys.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.