Re: 10.2 NFS mount points

On 02/01/2006 12:05:12 AM, Kevin Closson wrote:
>
> Mladen,
>
> The idea is to have a central NFS server for
> a bucnh of databases on a bunch of servers being
> backup up with RMAN...this isn't an argument over
> the virtues of DAS vs NAS.

Well, there is an error message, namely ora-27054 and two Metalink notes 338086.1 and 329942.1 which vaguely describe the so called "correct options". The problem appears to be in the maximum read and write sizes for the file system.
There aren't many 10.2 databases in production, especially not on AIX whose owners I find to be more conservative then the most, so he must break the ice. It will take some experimenting and calling both IBM and Oracle support. The gentleman has my sympathy and the advice to use hard drives if he doesn't feel like doing that.
In connection to that, there is a new version of NFS protocol which incorporates some changes:

- allows local caching
- allows async I/O
- has improvements in the locking protocol.

Essentially, things that have been proprietary for vendors like NetApp are now in the new protocol specification. That means that you will not only be able to put RMAN backups on the NFS drive, you will be able to put data files there as well. Red Hat FC3 has NFS4 client and here are the options, which are curiously similar to those recommended by aforementioned Metalink papers:

Mount options for nfs4

       Instead  of  a  textual  option string, parsed by the kernel, the nfs4
       file system expects a binary argument of type struct  nfs4_mount_data.
       The  program  mount  itself  parses  the following options of the form
       'tag=value', and  puts  them  in  the  structure  mentioned:  rsize=n,
       wsize=n,   timeo=n,  retrans=n,  acregmin=n,  acregmax=n,  acdirmin=n,
       acdirmax=n, actimeo=n, retry=n, port=n, proto=n, clientaddr=n,  sec=n.
       The option addr=n is accepted but ignored.  Also the following Boolean
       options, possibly preceded by no are recognized: bg, fg,  soft,  hard,
       intr, cto, ac, For details, see nfs(5).

       Especially useful options include

       rsize=32768,wsize=32768
              This will make your NFS connection faster than with the default
              buffer size of 4096.

       hard   The program accessing a file on a NFS mounted file system  will
              hang when the server crashes. The process cannot be interrupted
              or killed unless you also specify intr.  When the NFS server is
              back online the program will continue undisturbed from where it
              was. This is probably what you want.

This is, of course, taken from the manual page for "mount". Red Hat FC4 also has NFS4 server, but it is compiled with GCC 4.0. Almost nothing else uses GCC 4.x at this point, so I'm waiting for FC5 to upgrade. What I'm really interested in is trying RAC on NFS4.

-- 
Mladen Gogala
http://www.mgogala.com

--
http://www.freelists.org/webpage/oracle-l