Correct use of NFS

[Wim Van den Wyngaert]

TCP 5.3. On system A I have a disk DISK with directory DIR. I want to NFS mount it on system B as "/export". System B may not know of DISK and DIR, only of /export.

What do I need to map/export/mount ?

(tried it for an hour and went crazy of the error messages, also crazy because operator.log is deferred write)

Wim

[Uwe Zessin]

No immediate idea on NFS, sorry, but an idea how to work around OPCOM's buffering:

if you send the '$REPLY/LOG' command, it looks like all buffers are flushed to the current OPERATOR.LOG before a new file is opened. You can then access the contents, e.g. with:

$ type/tail sys$manager:operator.log;-1

[Joseph Huber_1]

tcpip map /disk DISK:
tcpip add export /disk/dir

On the client side: mount point is /disk/dir.
Unix fstab: node:/disk/dir:/mnt/vmsdir:rw:0:0

[Joseph Huber_1]

Sorry, in my previous answer, the Unix style pathes /disk, /disk/dir must be enclosed in string quotes ("").

And: for root to access the VMS export, there must be a NFS proxy for /gid=/uid=0.
At least I don't know off hand how to specify non-root mounts.

[Michael Yu_3]

Hi Wim,

If system B is not to know about DISK and DIR, you may have to use the container file.

On system A:-

TCPIP> add proxy local_user /host=systemB /GID=gid /UID=uid/NFS=incoming
TCPIP> create container DISK:[DIR]/user=local_user/host=systemB
TCPIP> map "/disk" DISK:
TCPIP> map "/export" DISK:[DIR]
TCPIP> add export "/export"/host=systemB


On system B:-

TCPIP> add proxy local_user /host=systemA /GID=gid /UID=uid/NFS=outgoing
TCPIP> mount dnfs0: /host=systemA /path="/export"

Thanks and regards.

Michael

[Wim Van den Wyngaert]

Joseph : System B may not know of DISK and DIR, only of /export

Michael : not possible without a container ?

It's VMS to VMS.

Wim

[Wim Van den Wyngaert]

Michael,

Create container wants to create the directory itself while it already exists.
I want to present an EXISTING directory !

Wim

[Joseph Huber_1]

Wim, You are completely free to name the map as You like. But You can't map a subdirectory unless it is a container, that's what You don't want, You want to share the directory between systems A and B.
So, if You don't want to export the whole disk, there is no way around exporting the path to the particular directory.
Why can't the client system mount /export/dir ?
It is just the path in the mount command (or in the /etc/fstab entry).
If the other system is Unix, then You can choose any local mount point (the Unix directory to mount the export TO).
So the mount would be
A:/export/dir:/mnt/export
and the access on Unix to VMSs disk:[dir] would be /mnt/export .

[Joseph Huber_1]

Wim, now I see You have both sides VMS, so my Unix examples are irrelavant.

Would You explain, what "B knows only of /export, not DISK and DIR" really means ?

The NFS client MOUNT command has to know the exported path, wether it is /export or /export/dir. The tcpip mount command still can choose which local directory it mounts it on:
TCPIP MOUNT DNFSn: results in DNFSn:[000000] being the exported directory. The users never see what is behind /export/dir. If You want to hide [000000] then You can choose any "mount point" like
TCPIP MOUNT DNFSn:[systemA]

Or is Your problem with existing system-management (or whoever does the TCPIP MOUNT on B), which now mounts /export, and can't be asked to change ?

[Wim Van den Wyngaert]

Joseph,

I wanted clients to use a logical when mounting. This way I can move the files to other disks / directories without having to modify all 100 clients.

Wim

[Joseph Huber_1]

But then the solution is easy:

the clients mount command needs once and for all fix the mount path (or the logical name resulting in) to say "/export/dir".
There is to my knowledge no way around specifying a subdirectory: in Unix speech "/export" is a device path only, and TCPIP mount "speaks" Unix in this regard.

Then only the server side need to change when the disk/directory becomes different:

Map /export to the new disk.
Create a directory alias (or hardlink on ODS5 disks) [DIR] pointing to the new directory,
then "/export/dir" need not be changed at all, and all the clients can mount without changing a single character in the pathname.

[Wim Van den Wyngaert]

Ok J but I avoid hardlinks. I was searching for a pure tcpip thing (e.g. via mapping).

Wim

[Jan van den Ende]

Wim,

if I understand correctly, you want to be able to just use:

SOMEDIR:

as a directory spec?

In that case you CAN
$ DEFINE/SYSTEM/EXEC SOMEDIR DNFS:[topdir.subdir]
Then you just
We have been using it for some time (although it has been quite some time ago).

Proost.

Have one on me.

jpe