[Veritas-bu] /usr/openv placement
2001-10-08 12:53:13
> I may be wrong here, but NFS was designed originally as
> a connectionless protocol,
Actually it was written as a *stateless* protocol. Every transaction was
(and still is to a certain degree) atomic and shouldn't rely on the
previous transaction. It is perfectly acceptable to have a stateless
protocol be connection oriented. HTTP is a perfect example of this.
> In a NBU environment if that should happen, it
> potentially hang and hang until failure occurs and a
> blocking of your backup window. Ultimately you will
> pay the price of incomplete backups.
>From *nix's perspective loosing an NFS drive is akin to loosing a disk. If
a disk were to fail fails, NBU can't do it's work either. But, in a proper
NFS environ, this would be HA and it going down wouldn't be an issue.
So unless the network transport fails, or a disk dies, there shouldn't be
an issue with timeouts. This is the *exact* same kinds of issues that SANs
based storage has to deal with. If the transport dies, then you're SOL.
> Alternatively, one may use hard mounts (as I know them
> to be called) which will NOT wait when a NFS filesyetem
> should become unavailable, but will terminate the
> connection.
Hard mount behavior is to do the waiting. Soft mounts on the other hand
will return an error and possibly corrupt data. From the Solaris man page
for mount_nfs
soft|hard Return an error if the server does not
respond, or continue the retry request
until the server responds. The default
value is hard.
> written to the mount point and not the NFS filer you
> want it to.
That will *NEVER* happen as the filesystem has been mounted over. In unix,
once something is mounted, it doesn't disappear unless it's told to
explicitly via a umount.
-- Cheers
-- James
|
|
|