[wplug] NFS Availability Issues -- more info on your network ...

Bryan J Smith b.j.smith at ieee.org
Tue Sep 11 14:20:40 EDT 2007


I guess I didn't make a point clear.

At most, I'd pilot an OpenAFS volume, and see if write caching works.
But that's as far as I'd go, and not put any failover hardware in.

Knowing your users' applications, usage habits, transfer rates, etc...
would probably help me with you a bit more, as I'm "shooting in the dark.”

--  
Bryan J Smith - mailto:b.j.smith at ieee.org  
http://thebs413.blogspot.com  
Sent via BlackBerry from T-Mobile  
    

-----Original Message-----
From: "Bryan J. Smith" <b.j.smith at ieee.org>

Date: Tue, 11 Sep 2007 13:17:29 
To:Michael Semcheski <mhsemcheski at gmail.com>
Cc:General user list <wplug at wplug.org>
Subject: Re: [wplug] NFS Availability Issues -- more info on your network ...


On Tue, 2007-09-11 at 12:41 -0400, Michael Semcheski wrote:
> All defaults on the clients.

My typical recommendation has been (for a local subnet) ...
  vers=3,udp,bg,hard,intr,rsize=8192,wsize=8192

> No, the server has only been online for a month.  And I was at home
> when it happened.

Then I think you have your answer (i.e., failover is overkill ;).

> I'm alright with increasing the complexity.

I don't think you are, honestly.  It really adds some support load that
is really overkill for your configuration.  Again, this isn't web
failover or typical SQL replication.

You need to procure unified storage that is simultaneously targettable.

> But I imagine if we do change things up, it will be to go to AFS,
> which from what I can tell has better availability.  (through
> local caching.)

Yes and no.  Yes, it has caching.  But no, I've never gotten the OpenAFS
implementation to work well for anything but read-only.  But that may
have changed in more recent months.

Again, I think you're really not considering the scale and other details
here.  Heck, and at some point, the "cost" in your time and storage
"hardware," even a NetApp "cheaper" for single subnets with only a dozen
users or two.

Save failover for when you need it.  You don't need it for 5 users.
Don't over-complicate your solution.  Look at what would be best for
your users, at the cost-point _per_user_.  ;)


-- 
Bryan J. Smith         Professional, Technical Annoyance
mailto:b.j.smith at ieee.org   http://thebs413.blogspot.com
--------------------------------------------------------
        Fission Power:  An Inconvenient Solution


_______________________________________________
wplug mailing list
wplug at wplug.org
http://www.wplug.org/mailman/listinfo/wplug



More information about the wplug mailing list