Brisque TIC
In-Depth NFS Configuration

This document is intended to provide more in-depth information regarding NFS mounts on the Brisque.

The topics covered include:

Mount types (hard or soft)
Number of times to attempt a mount
Interrupting the mount process
Brisque application issues
MountPro vs. Smit
Changing the MountPro NFS Settings
Booting in single-user (stand alone) mode
NFS options
NFS related commands

Mount Types

There are two NFS mount types - hard and soft. Soft mounts should always be used the Brisque.

Soft Mounts

The advantage of a soft mount is that if an NFS server connection fails, after a specified number of attempts ("NUMBER of times to attempt a mount"), the NFS client will give up on the mount and return a software error.  This will allow the application to continue.

The default value for the number of retries is 1000.  The Brisque will continue to "work" after it has tried to mount the server 1000 times.  To reduce this number (and the amount of time the Brisque is "frozen") you must use Smit (or edit /etc/filesystems) and reduce this value for the option "retry" (see below).

Hard Mounts

A hard mount is designed for use over a reliable link (i.e. a connection to an NFS server where the network connection and the server will be permanently up).  When a hard NFS mount to a server fails (i.e. the server or network connection goes down) the client will retry an infinite number of times (i.e. forever).

If the option "intr" (allow interrupts) is set to "Yes", pressing the keyboard interrupt (default "Control-C") can terminate the mount operation.   However, this interrupt cannot be used until you have logged in.  This presents a problem – if a hard mount fails during the boot, you will never get as far as a login prompt and a terminal session so you will not be able to interrupt the mount. In this case the Brisque will sit there forever or until the NFS server becomes available again.  For this reason it is recommended that you use only soft mounts on the Brisque.

Hard mounts on the Brisque will cause un-interruptible application "freezes".   See interrupting the mount process – for a detailed explanation.

Number of Times to Retry a Mount ("retry" Option)

Whenever the Brisque (File Manager or Browser) attempts to access an NFS filesystem whose server has gone down, it will retry the connection according to the value of the option "retry" in the filesystem’s stanza.  As mentioned earlier, if this option is not explicitly set, the default value is 1000.  So, if there are five NFS filesystems from a RiproServer mounted on the Brisque and the RiproServer disappeared, the Brisque will try to re-mount each filesystem 1000 times.  Consequently, the current window will "freeze" for the length of time this takes (probably about 10 minutes).

Note:

The value of "retry" appears to have no effect on the length of time the Brisque will try to connect for at boot time.

The value of "retry" can be set using either Smit, the mknfsmnt or chnfsmount commands or manual editing of filesystem stanzas in /etc/filesystems.

• Smit – set the value of the field "NUMBER of times to attempt mount"
• /etc/filesystems – "retry=#" (where "#" is value required)
• mknfsmnt/chnfsmnt – "-r #" (where "#" is value required)

Examples

To set the number of retries to 5:

SMIT

To do the same using smit (select either "Add a File System for Mounting" or "Change/Show Attributes of an NFS File System". Change the value of the "NUMBER of times to attempt mount" field (shown in bold).

Add a File System for Mounting

Edit /etc/filesystems

Edit /etc/filesystems, add the following string onto the end of the option line (remember to put a comma after the previous option):

retry=5

mknfsmnt/chnfsmnt

Type one of the following commands to either make or modify an NFS mount respectively:

mknfsmnt -f /dataVolumes/NT_APR -d /c/apr -h NT1 -t rw -S -r 5 -A

chnfsmnt -f /dataVolumes/NT_APR -d /c/apr -h NT1 -r 5

The first command will mount "c:\apr" from an NT server called "NT1" on /dataVolumes/NT_APR. The options for this mount are:

-t rw - read-write

-S - soft mount

-A - auto mount on startup

-r 5 - 5 retries

The second command modifies the mount of "c:\apr" from NT1 so that "retry" = 5 (-r 5).

MountPro

When NFS mounts are performed using MountPro, the option "Automatic Mount on System Startup" must be checked (otherwise no stanza is created).

Once the mount is completed, either run smit, edit /etc/filesystems or use the chnfsmnt command to set "retry" to 5

Interrupting the Mount Process ("intr" Option)

The NFS mount option "intr" will determine whether or not an NFS mount can be interrupted from the keyboard.  Unfortunately, it is not possible to interrupt a soft mount; only hard mounts can be interrupted.

The default interrupt character is Control-C but it is not possible to use this during the initial boot phase (when automatic mounts are performed).  In cases where auto-mounts are failing at boot time, there are only 3 solutions:

• Get the NFS server (or network) up and running again
• Allow the mount to timeout (be backgrounded) – soft mounts only
• Boot in single user mode (stand alone) and remove the mount – saves a cold start

Additionally, the interrupt character (Control-C) will have no effect within CreoScitex applications (the interrupt character is an attribute of stdin – i.e. the terminal). For example, it is not possible to interrupt a mount in the following circumstances.

• When the File Manager displays the status "Reading…" whilst trying to refresh /dataVolumes/
• When the browser is opened in applications such as HF Setup, Assembly, ExportPS, etc.

Therefore, the only time an interruptible hard mount is of any practical use is when running the mount command from a UNIX prompt. If the mount fails, pressing Control-C will abort the mount and return the user to a UNIX prompt.

Brisque Application Issues

The main application problems caused by NFS on the Brisque occur when a connection to an NFS mount goes down.

For example, if Brisque1 has mounted the filesystems Brisque2.1.0 and Brisque2.2.0 from Brisque2 and Brisque2 is subsequently shutdown or loses its network connection, the two main problem areas will be:

• FileManager
• Browser

FileManager

Whenever a new instance of FileManager is opened, the default directory will be /dataVolumes/.  This will cause the operating system to "examine" the attributes of all directories in /dataVolumes/ (usually all JFS and NFS filesystem mount points).  The same situation will occur when the user double clicks on the "dataVolumes" folder in the icon path.

Whenever the FileManager’s current directory is /dataVolumes/ and an NFS filesystem is unavailable, the NFS client will attempt to remount it and the status "Reading…" will be displayed in the FileManager’s status bar.

Browser

In most Brisque applications that utilize the Browser service (shown), the default location is set to /dataVolumes/.  This will also cause the NFS client to attempt a remount of the NFS filesystems on servers that are down.

Brisque Browser

"Freeze" Duration

When NFS connections are lost and /dataVolumes is accessed, the FileManager will display "Reading…" and the Browser will display a blank window until one of the following conditions is met:

• The NFS server reappears and the filesystems are remounted
• Soft mounts – the number of retries for each filesystem has been attempted
• Hard mounts – the application will be "frozen" forever

To reduce this "freeze" duration, the "retry" value should be changed to a low number (e.g. 3) for a LAN connection. (If an NFS server is on the other side of a very slow connection (e.g. WAN link) a larger number will be required otherwise the filesystem will erroneously be shown disconnected.)

MountPro vs. Smit

There are some significant differences between smit and MountPro

Smit

• Smit allows many more options to be set
• Smit will work with non-Scitex servers

MountPro

• MountPro has a nicer user interface but will only work under the following conditions:
  • The mount point of the disks on the remote station (or symbolic links to these) must exist under /dataVolumes/
  • The name of the mount point on the remote station must start with the hostname of the remote station (E.g. if the hostname is "mark" then disks mount points called "market", "Mark", "Mark1", "mark1.1.0" will all appear in the MountPro and Send volume list.
  • The UNIX version must support the same command-options and return the same command-output format as AIX.
• MountPro will only create a stanza in /etc/filesystems if the option "Automatic Mount on System Startup" is checked.
• The following options are used when NFS mounts are carried out using MountPro
  • Soft mount
  • All other values are default (e.g. retries = 1000)

These requirements, practically, limit MountPro to CreoScitex stations.

The NFS mount created by MountPro is identical to a mount (with the same options) created using "smit".

Changing the MountPro Settings

MountPro makes NFS mounting of other stations very easy. Unfortunately, the default NFS mount settings make the feature unusable in many multi-Brisque customer sites where the Brisques are constantly restarted.

• Do NOT pass this information arbitrarily to customers.
• Do NOT use this unless you are 100% certain of what you are doing.
• Only use this where it is really necessary AND where there is a GOOD systems administrator who understands this, who knows it is UNSUPPORTED and may not function in future versions.

Settings

MountPro uses the file /scitex/version/config/sciMountPref.ans to determine which options to use with themknfsmnt command.  By modifying this file (correctly) it is possible to enter your own settings for NFS mounts (even when no stanza is created). Before modifying this file make a backup:

cp sciMountPref.ans sciMountpref.ans.orig

The file default file contents are shown below:

#Scitex
#AnswerFile v1
#File Creator: DialogResourceManager
! DM BEGIN
!
!     Default values for the mount command used by libSciMount.
!      (The parameters are used by the '/usr/etc/mknfsmnt' command.
!       To add/change the flag see the 'mknfsmnt' info.)
!
.*SciMountDefaultFlags.value: -S -f
.*SciMountAutoRestartMountFlags.value: -B -A
.*SciMountNoAutoRestartMountFlags.value: -N
! DM END

There are three lines of interest here. These are the three lines preceding "! DM END"

To modify the default options change the first line (SciMountDefaultFlags).  Make sure that you enter new options before "-f" otherwise the mount will fail.

Example

To set the default number of retries to 3, enter "-r 3" between "-S" and "-f":

.*SciMountDefaultFlags.value: -S -r 3 -f

This will generate the following mknfsmnt command:

mknfsmnt -m nfs -t rw -N -S -r 3 -f /dataVolumes/B2.1.0 -d /dataVolumes/B2.1.0 -h B2

It should be clear from the above command that if any options are entered after "-f" the command will fail because the remote path must immediately follow the "-f" switch (e.g. -f /dataVolumes/br2.1.0)

Remember, use this at your discretion and be very selective about where and when it is used.

Booting In Single-User (stand-alone) Mode

When the Brisque will not boot due to problems with NFS auto-mounts and the last resort is to remove the auto-mount, the Brisque needs to be booted in stand-alone mode. To do this you will need the AIX CD-ROM. The procedure described below is for AIX v4.2.1

1. Insert the AIX Version 4.2.1 CD-ROM into the CD drive.
2. Power ON the Brisque, the system will perform the POST (Power On Self Test).
3. During the POST, four progress indicators appear as icons across the bottom of the screen.
4. Press [F5] during the POST process, before the fifth indicator appears boot from CD-ROM.
5. After a few minutes the boot process begins and the software installation menus will appear.
6. Press the [F1] key and press Enter to use main display as the system console.
7. Press [1] and press Enter to use the English language during the installation.
8. Press [3] and Enter to "Start Maintenance Mode For System Recovery".
9. Press [1] and Enter to "Access a Root Volume Group".
10. Press [0] and Enter to "Continue".
11. Press [1] and Enter to "Select hdisk0".
12. Press [1] and Enter to "Access this Volume Group & Start a Shell".
13. A message saying "Importing Volume Groups…" will appear on the screen for a few minutes>
14. When the UNIX prompt appears type the following command: export TERM=vt100<Enter>
15. Either use "vi" or "smit" to remove or change the options pertaining to the problematic mounts.
16. Type "reboot" to reboot the Brisque.

NFS Options

Below is the full syntax of the mknfsmnt command, which shows all available options and defaults for AIX v4.2.1. (Not all of these are available in smit, e.g. NFS version, Transport protocol, etc.)

Syntax

/usr/sbin/mknfsmnt -f PathName -d RemoteDirectory -h RemoteHost
[-t {rw|ro}] [-m MountTypeName] [-w {fg|bg}] [-X|-x] [-S|-H] [-Y|-y]
[-Z|-z] [-e|-E] [-a|-A] [-s|-n] [-I|-B|-N] [-r TimesToRetry]
[-b ReadBufferSize] [-c WriteBufferSize] [-o TimeOut] [-P PortNumber]
[-u AcRegMin] [-U AcRegMax] [-v AcDirMin] [-V AcDirMax] [-T AcTimeOut]
[-p NumBiods] [-K {any|2|3}] [-k {any|tcp|udp}]

Description

The mknfsmnt command constructs an entry that will be appended to the /etc/filesystems file, thus making a file system available for mounting.  If the mount is to be permanent, this entry will remain.  If the mount is temporary, the flags will be used directly for the mount command. If the mount is soft, the system returns an error if the server does not respond.  If the mount is hard, the client continues trying until the server responds.  The hard mount is the default.

Flags

Example

To add the mount of a remote directory, enter:

mknfsmnt -f /dataVolumes/ripro1.1.0 -d /dataVolumes/ripro1.1.0 -h ripro1

In this example, the mknfsmnt command mounts the remote directory /dataVolumes/ripro1.1.0 that resides on ripro1 on the local /dataVolumes/ripro1.1.0 directory

Files

/etc/filesystems         Lists the remote file systems to be mounted during the system restart.

Related Information

The chnfsmnt command, mount command, rmnfsmnt command.

How to Mount a NFS File System Explicitly in AIX Version 4.3 System Management Guide:   Communications and Networks.

Network File System (NFS) Overview for System Management in AIX Version 4.3 System Management Guide: Communications and Networks.

List of NFS Commands

Below are a list of commands related to NFS configuration.