5 Setting Up an NFS Service

A Network File System (NFS) service includes one or more file systems, Advanced File System (AdvFS) filesets, or Logical Storage Manager (LSM) volumes that a member system exports to clients, making the data highly available. NFS services can also include highly available applications.

An NFS service name is assigned its own Internet address. The member system that runs the service responds to this address. This makes the service autonomous and not dependent on the availability of any particular member system. Clients access the service by including the service name and the exported directory path in their /etc/fstab file. If the service stops on a member system, it fails over to a viable system, and clients only experience a short timeout.

The NFS service name also allows you to use the POLYCENTER NetWorker Save and Restore (NetWorker) to back up the service's storage. NetWorker treats the NFS service as an independent client and stores the storage indexes under the name of the service. This enables you to back up and recover the service's storage independent of the member system running the service. See the NetWorker documentation for information about using NetWorker to back up an NFS service's storage.

To set up an NFS service, you should be familiar with setting up NFS in general, the /etc/exports file, and the /etc/fstab file.

Before you set up your NFS service, both the client and member systems must be running NFS Version 2.0 or Version 3.0 and use the Address Resolution Protocol (ARP). You also must prepare the shared disks that will be used in the service and install any application used in the service.

To fail over an application, in addition to disks, at a minimum, you must create a user-defined start action script that includes the commands to start the application, and create a user-defined stop action script that includes the commands to stop the application. See Chapter 4 for more information about preparing disks, applications, and action scripts for a service.

5.1 NFS Service Requirements

Network Files System (NFS) services have the following requirements:

NFS service names must be included in the local /etc/hosts file on each member system before you set up the NFS service.

You cannot use an NFS service name that is the same as the name of a member system. Service names and member system names must be unique.

When exporting directories from a service, the directory names must be unique within the entire ASE, and unique from any nonshared exports made from the /etc/exports file from any ASE member nodes.

An NFS service name must adhere to the conventions for naming a system, as described in the DIGITAL UNIX Installation Guide.

If you are using a distributed database lookup service such as the Network Information Service (NIS), be sure that the service name information is local to all the member systems. To do this, make all the member systems either master or slave servers, or specify the service name information in the local /etc/hosts file. Be sure your /etc/svc.conf file specifies local as the first entry.

Do not manually mount or dismount disks that are used in an NFS service.

If you set up an NFS service in your ASE, do not use the automount command option /net -hosts on any client system that accesses the service's NFS file systems.
If a client uses the /net/hostname or /net/service_name mount points to NFS-mount a file system from a member system, and TruCluster software relocated the NFS sevice to a different member system, the client may receive "stale file handle" error messages.

Use the automount command -p option to limit the local loopback mounts to primary Internet addresses. When passed this option, automount queries each of the system's configured network interfaces for its primary Internet address. automount uses local loopback mounts for all Internet addresses produced from this query. All Internet alias addresses for the system will be treated as remote addresses and will use NFS mounts.
By default, automount bypasses NFS for all local Internet addresses, including Internet alias addresses. The automount -p option enables Internet alias addresses to use NFS for all automounted file systems, and is necessary for ASE servers to avoid difficulty modifying or stopping services.

5.2 NFS Service Components

When you add a Network File System (NFS) service to an available server environment (ASE), the asemgr utility prompts you for service-specific information, in addition to information that is similar to what you specify with the nfssetup script. See nfssetup(8) for more information.

You can specify the following NFS service information:

Service name-- To enable clients to access an NFS service, the service name is assigned its own Internet address. The service name and Internet address must be included in each member system's local /etc/hosts file before you set up the NFS service. See hosts(4) for more information. In addition, an NFS service name must adhere to the conventions for naming a system, as described in the DIGITAL UNIX Installation Guide.

Automatic Service Placement (ASP) policy--See Chapter 4 for information about the ASP policies.

UNIX file systems, Advanced File System (AdvFS) filesets, or Logical Storage Manager (LSM) volumes--See Chapter 4 for information about setting up disks.

Mount point to be exported for each file system, fileset, or volume--Client systems will specify this mount point in their /etc/fstab files to access the service's file systems, filesets, or volumes.

Names of the remote hosts, network groups, or Internet Protocol (IP) addresses to which you want the file systems, filesets, or volumes restricted--See exports(4) for information about remote host access to NFS mount requests.

NFS locking area--If you have more than one writable disk area in a service, you must specify which area to use for the NFS locking area. The NFS locking software maintains some data that must be failed over, so the data must be placed on a writable, shared disk.

Mode (either read/write or read-only) and any mount options other than the default options for each file system, fileset, or volume.

If you also want to fail over an application, you must modify the NFS service and specify the action scripts. See Chapter 4 for information about action scripts. See Chapter 10 for information about modifying services.

5.3 Understanding the Service Exports File

When you add a Network File System (NFS) service to an available server environment (ASE), the TruCluster software edits the /etc/exports.ase file on each member system and includes an entry that specifies the service's exports file. For example:

 # more exports.ase
 
 .INCLUDE /etc/exports.ase.aseba1
 .INCLUDE /etc/exports.ase.aseba2
 #

Service exports file names have the following syntax:

/etc/exports.ase.service

The service variable specifies the service name.

A service exports file contains a list of all the file systems and filesets in the service and their mount points, using a format that is similar to the /etc/exports file. It can include the remote hosts, network groups, or Internet Protocol (IP) addresses to which the service's file systems or filesets are restricted. If none are specified in the file, then all remote hosts can mount the directory. See exports(4) for information about the file format.

Entries in service exports files include a -m option, which specifies the actual mount point for a file system or fileset.

Note

Do not manually edit the /etc/exports.ase.service file to modify services; instead, use the asemgr utility to make modifications.
To delete a file system or fileset from an NFS service, use the asemgr utility to remove its entry from the /etc/exports.ase.service file. When deleting a file system or fileset, the asemgr utility prompts you to invoke an editor, providing the opportunity to delete the entry at this time. If you choose to not run an editor at this time, then you must remember to do so later.

The following example shows an exports file for an NFS service with two file systems:

#
#  ASE exports file for service aseba2 (edit only with asemgr)
#
 
#/dev/rz25c exports (after this line) - DO NOT DELETE THIS LINE
/ase/aseba2 -m=/var/ase/mnt/aseba2/ase/aseba2 -ro=0
 
#/dev/rz26c exports (after this line) - DO NOT DELETE THIS LINE
/ase/aseusr -m=/var/ase/mnt/aseba2/ase/aseusr  testit milan tabby
#

5.4 Adding a Basic NFS Service

To add a Network File System (NFS) service to an available server environment (ASE), choose the "Adding a new service" item from the Service Configuration menu and provide the appropriate information for the service at the prompts. Example 5-1 shows an example of adding a basic NFS service that includes a UNIX file system and a Logical Storage Manager (LSM) volume.

Example 5-1: Adding a Basic NFS Service

# asemgr
 

.
.
.
Adding a service
 
Select the type of service:
 
    1)  NFS service
    2)  Disk service
    3)  User-defined service
    4)  DRD service
    5)  Tape service
 
    q)  Quit without adding a service
    x)  Exit                ?)  Help
 
Enter your choice [1]: 1
 
You are now adding a new NFS service to the ASE.
 
An NFS service consists of an IP host name and disk configuration
that are failed over together.  The disk configuration can include
UFS file systems, AdvFS filesets, and LSM disk groups.
 
                     NFS Service Name
 
The name of an NFS service is a unique IP host name that has been
set up for this service.  This host name must exist in the local
hosts database on all ASE members.
 
Enter the NFS service name: ase3

 
Checking to see if ase3 is a valid host...
 
                     Specifying Disk Information
 
Enter one or more UFS device special files, AdvFS filesets, or LSM
volumes to define the disk storage for this service.
 
    For example:        Device special file:      /dev/rz3c
                        AdvFS fileset:            domain1#set1
                        LSM volume:               /dev/vol/dg1/vol01
 
To end the list, press the Return key at the prompt.
 
Enter a device special file, an AdvFS fileset, or an LSM volume as
storage for this service (press 'Return' to end): /dev/rz25c
 
Enter the directory pathname(s) to be NFS exported from the storage
area "/dev/rz25c".  Press 'Return' when done.
 
Enter a directory pathname: /ase_dir
Enter a host name, NIS netgroup, or IP address for the 
NFS exports list (press 'Return' for all hosts): [Return]
 
Enter a directory pathname: [Return]
 
              UFS File System Read-Write Access
 
Mount /dev/rz25c file system with read-write or read-only access?
 
    1)  Read-write
    2)  Read-only
 
Enter your choice [1]:[Return]
 
You may enable user and group quotas on this file system by specifying
full path names for the quota files.  If you place the files within
the service's file systems, the quota assignments you make with
edquota will relocate with the service.  Enter "none" to disable
quotas.
 
User quota file [/var/ase/mnt/ase3/ase_dir/quota.user]: [Return]
 
Group quota file [/var/ase/mnt/ase3/ase_dir/quota.group]: [Return]
 
              UFS Mount Options Modification
 
Enter a comma-separated list of any mount options you want to use
for "/dev/rz25c" (in addition to the UFS-specific defaults listed
in the mount.8 reference page).  If none are given, only the
default mount options are used.
 
Enter options (Return for none): noexec
 
Enter a device special file, an AdvFS fileset, or an LSM volume as storage
for this service (press 'Return' to end): /dev/vol/dg3/vol04
 
Enter the directory pathname(s) to be NFS exported from the storage
area "/dev/vol/dg3/vol04".  Press 'Return' when done.
 
Enter a directory pathname: /ase_data
 
Enter a host name, NIS netgroup, or IP address for the 
 
NFS exports list (press 'Return' for all hosts): net_staff
Enter a directory pathname: [Return]
 
The following is a list of device(s) and pubpath(s) for disk group dg3:
 
        DEVICE  PUBPATH
 
        rz20c   /dev/rz20c
 
Is this correct (y/n) [y]: y
 
              UFS File System Read-Write Access
 
Mount /dev/vol/dg3/vol04 file system with read-write or read-only access?
 
    1)  Read-write
    2)  Read-only
 
Enter your choice [1]: 2
 
              UFS Mount Options Modification
 
Enter a comma-separated list of any mount options you want to use
for "/dev/vol/dg3/vol03" (in addition to the UFS-specific defaults listed
in the mount.8 reference page).  If none are given, only the default
mount options are used.
 
Enter options (Return for none): nosuid
 
Enter a device special file, an AdvFS fileset, or an LSM volume as
storage for this service (press 'Return' to end): [Return]
 
NFS needs a disk area that is writable to keep some state information
for NFS locking during ASE operation.  Choose a disk area that is
writable and will not fill up.
 
Select the disk area to use for the NFS locking information:
 
    1)  /dev/rz25c		(UFS)
    2)  /dev/vol/dg3/vol04	(UFS)
    x)  Exit
 
Enter your choice [1]: 1
 
 
        Selecting an Automatic Service Placement (ASP) Policy
 
Select the policy you want ASE to use when choosing a member
to run this service:
 
    b)  Balanced Service Distribution
    f)  Favor Members
    r)  Restrict to Favored Members
 
    x)  Exit to Service Configuration     ?)  Help
 
Enter your choice [b]: b
 
 
       Selecting an Automatic Service Placement (ASP) Policy
 
Do you want ASE to relocate this service to a more highly favored
member if one becomes available while the service is running (y/n/?):y
 
Enter 'y' to add Service 'ase3' (y/n):y
 
Adding service...
Starting service...
Saving the updated database...
Service successfully added...

5.5 Adding an NFS Mail Service

You can use the TruCluster software to set up a mail system and make it highly available. The sendmail program uses the Simple Mail Transfer Protocol (SMTP) to deliver mail messages between users, systems, and networks. You can set up member systems as mail hubs (servers) so that other systems in your mail environment send mail to and through the mail hubs. If a problem occurs in a mail hub, the TruCluster software can fail over the mail to another hub and reroute incoming mail to the new hub.

Before setting up a mail service, you should understand how the sendmail program works. The sendmail program can receive mail from an SMTP connection or directly from a process; that is, from the mail system or some user interaction. For example:

The sendmail program writes the message to a mail queue area, which is the /var/spool/mqueue directory by default.

After the entire mail message is written to the mail queue area, sendmail tells the sending process that the mail was received, so the sending process is assured that the mail was delivered. If the machine crashes at this point, a copy of mail remains in the mail queue area, so the mail is not lost.

After a secure copy of the message is in the mail queue area, sendmail parses the address and delivers the message according to the instructions in the sendmail configuration file, /var/adm/sendmail/sendmail.cf.

The sendmail program passes the mail to another delivery agent, such as DECnet, UNIX-to-UNIX Copy Program (UUCP) or another SMTP. In addition, local mail is passed to the local mailer (/usr/bin/mail) which, by default, delivers it to the system mailbox, /var/spool/mail/username. If the address is not local, sendmail passes the mail to the mail delivery agent on the remote machine. If sendmail cannot pass the mail (for example, the remote machine is down), the mail remains in the queue area to be processed at a later time.

To set up a highly available mail service with the TruCluster software, the file systems or filesets containing the /var/spool/mail mailbox directory and the /var/spool/mqueue mail queue area must be shared between two or more systems. The /var/spool/mail mailbox directory must be shared so that the mail drop is available for mail delivery and processing on all the member systems that are set up as mail hubs. The /var/spool/mqueue queue area must be shared to ensure that any mail that remains in the queue area can be processed even if the hub that queued the mail is not available. In addition, both areas must have a common floating network connection to which mail can be sent.

You can share the mailbox directory and the queue area by using the TruCluster software to set up an NFS service to share the file systems or filesets and by modifying all the mail hubs' sendmail.cf configuration files.

The following steps describe how to set up two member systems as mail hubs (server1 and server2) using an NFS service named mail_hub:

Set up the disk or disks that will contain the /var/spool/mail and /var/spool/mqueue file systems.

Use the asemgr utility to set up the NFS service mail_hub that will export the /var/spool/mail and /var/spool/mqueue file systems.

Note

NFS locking (the lockd daemon) must be set up and running on the member systems that are mail hubs because locking will be done on both exported spool areas.

Modify the mail_hub service's exports file. Use the asemgr utility to make the /var/spool/mail and /var/spool/mqueue file systems accessible by root.

Set up the sendmail.cf configuration files on the member systems that will be mail hubs, server1 and server2. You must set up the sendmail.cf files to ensure that mail addressed to user@server1, user@server2, or user@mail_hub is delivered locally.

NFS-mount the file systems on the mail hub member systems. On both mail hub member systems, server1 and server2, use the mount command to NFS-mount /var/spool/mail and /var/spool/mqueue from the service (Internet host name) mail_hub and then add this mount point to the /etc/fstab file on each member system.

Optionally, define a Berkeley Internet Name Domain (BIND) mail exchanger (MX) record to point to all mail hubs. You can define an MX record so that if server1 is inaccessible, mail sent to server1 is forwarded to server2 and vice versa.

After you complete these steps, the mail service is ready to use. You can send mail to server1, server2, or mail_hub, and your mail will be delivered to the shared local /var/spool/mail area on the server1 and server2 mail hub member systems.

You can log in to server1, server2, or mail_hub and access your mail. You can also mount /var/spool/mail@mail_hub on another system and access your mail from that system. However, if one of the mail hub member systems goes down, mail sent directly to that mail hub member system will not be delivered until it reboots. You can fix this problem by defining the BIND MX records.

The following sections describe the steps in detail.

5.5.1 Preparing Disks for a Mail Service

To prepare the disks that will contain the shared /var/spool/mail and /var/spool/mqueue file systems, follow the guidelines specified in Chapter 4.

If you are using one disk partition for both the /var/spool/mqueue and /var/spool/mail directories, create an mqueue directory and a mail directory. You perform these tasks on only one mail hub member system.

The following example sets up a UNIX file system on an entire RZ10 disk and creates two directories:

# newfs /dev/rrz10c
# mount /dev/rz10c /mnt
# mkdir /mnt/mail
# chmod 1777 /mnt/mail
# mkdir /mnt/mqueue
# chmod 755 /mnt/mqueue
# umount /mnt

5.5.2 Using the asemgr Utility to Add a Mail Service

To add a mail service to your available server environment (ASE), run the asemgr utility on one mail hub member system, choose the "Add a new service" item from the Service Configuration menu and provide the information appropriate for your configuration at the prompts.

Example 5-2 shows how to add a service named mail_hub, which consists of the /dev/rz10c file system and the /var/spool/mqueue and /var/spool/mail directories. The example also shows how to restrict access to the service to the server1 and server2 mail hub member systems.

Example 5-2: Adding a Mail Service

# asemgr
 

.
.
.
Adding a service
 
Select the type of service:
 
    1)  NFS service
    2)  Disk service
    3)  User configured service
    4)  DRD service
    5)  Tape service
 
    q)  Quit without adding a service
    x)  Exit                   ?)  Help
 
Enter your choice [1]: 1
 
You are now adding a new NFS service to the ASE.
 
An NFS service consists of an IP host name and disk configuration that
are failed over together.  The disk configuration can include UFS file
systems, AdvFS filesets, and LSM disk groups.
 
                     NFS Service Name
 
The name of an NFS service is a unique IP host name that has been set
up for this service.  This host name must exist in the local hosts
database on all ASE members.
 
Enter the NFS service name: mail_host
 
Checking to see if mail_host is a valid host...
 
 
                   Specifying Disk Information
 
Enter one or more UFS device special files, AdvFS filesets, or LSM
volumes to define the disk storage for this service.
 
    For example:        Device special file:      /dev/rz3c
                        AdvFS fileset:            domain1#set1
                        LSM volume:               /dev/vol/dg1/vol01
 
To end the list, press the Return key at the prompt.
 
Enter a device special file, an AdvFS fileset, or an LSM volume as
 storage for this service (press 'Return' to end): /dev/rz10c
 
Enter the directory pathname(s) to be NFS exported from the storage
area /dev/rz10c.  Press 'Return' when done.
 
Enter a directory pathname: /var/spool/mail
Enter a host name, NIS netgroup, or IP address for the
NFS exports list (press 'Return' for all hosts):server1 server2
 
Enter a directory pathname: /var/spool/mqueue

Enter a host name, NIS netgroup, or IP address for the
NFS exports list (press 'Return' for all hosts):server1 server2
 
Enter a directory pathname:[Return]
 
              UFS File System Read-Write Access
 
Mount /dev/rz10c file system with read-write or read-only access?
 
    1)  Read-write
    2)  Read-only
 
Enter your choice [1]: 1
 
You may enable user and group quotas on this file system by specifying
full path names for the quota files.  If you place the files within
the service's file systems, the quota assignments you make with
edquota will relocate with the service.  Enter "none" to disable
quotas.
 
User quota file [/var/ase/mnt/mail_host/var/mail/quota.user]: none
Group quota file [/var/ase/mnt/mail_host/var/mail/quota.group]: none
 
 
              UFS Mount Options Modification
 
Enter a comma-separated list of any mount options you want to
use for /dev/rz10c (in addition to the UFS-specific defaults
listed in the mount.8 reference page).  If none are given, only
the default mount options are used.
 
Enter options (Return for none): [Return]
 
Enter a device special file, an AdvFS fileset, or an LSM volume as
storage for this service (press 'Return' to end):  [Return]
 
        Selecting an Automatic Service Placement (ASP) Policy
 
Select the policy you want ASE to use when choosing a member
to run this service:
 
    b)  Balanced Service Distribution
    f)  Favor Members
    r)  Restrict to Favored Members
 
    x)  Exit to service config menu      ?)  Help
 
Enter your choice [b]: b
 
        Selecting an Automatic Service Placement (ASP) Policy
 
Do you want ASE to relocate this service if a more highly favored
member becomes available while the service is running (y/n/?):n
 
Enter 'y' to add Service 'mail_host' (y/n): y
Adding service...
Starting service...
Saving the updated database...
Service successfully added...

5.5.3 Modifying the Mail Service's Exports File

On one mail hub member system, use the asemgr utility to edit the mail_hub service's exports file and make the /var/spool/mail and /var/spool/mqueue directories accessible by root on all the mail hub member systems. You must add the -root=0 option to the entries for the /var/spool/mqueue and /var/spool/mail directories in the /etc/exports.ase.mail_hub file.

To edit the mail_hub service's ASE exports file, follow these steps:

Invoke the asemgr utility and choose the "Modify a service" menu item from the Service Configuration menu.

Choose the name of the service you want to modify. In this example, choose mail_hub.

Choose the "General service information" menu item when prompted for what you want to modify.

Choose the disk area that contains the mail areas to modify. In this example, choose /dev/rz10c.

Choose the "Modify the NFS exports list" menu item. The asemgr utility invokes an editor (as defined by the EDITOR system variable) so you can edit the mail service's ASE exports file.

Edit the exports file to include the -root=0 option. For example, the file should look like the following:

#
#  ASE exports file for service mail_hub
#
 
/dev/rz10c exports (after this line) - DO NOT DELETE THIS LINE
/var/spool/mqueue -root=0 server1 server2
/var/spool/mail -root=0 server1 server2

Exit the asemgr utility. The mail_hub service's exports file, /etc/exports.ase.mail_hub, is updated on all the mail hub member systems.

5.5.4 Setting Up the sendmail.cf Configuration File

On each mail hub member system, you must set up the sendmail.cf configuration file to handle mail sent directly to the mail hub member systems and to the mail service name as local mail. For example, if server1 and server2 are mail hub member systems for the NFS mail service mail_hub, you must set up the sendmail.cf configuration file on both mail hub member systems to ensure that mail sent to server1, server2, and mail_hub is handled as local mail.

Because server1, server2, and mail_hub share the /var/spool/mail area, mail sent to any of the three addresses is delivered to the shared local /var/spool/mail area.

You can use several methods to configure sendmail to do this:

Use the mailsetup command

Modify the server.m4 file and then reconfigure sendmail

Directly modify the sendmail.cf file

You must configure the sendmail.cf file on all the mail hub member systems. To do this, invoke the mailsetup command and choose the option to perform an advanced mail setup. Add server1, server2, and mail_host to the NICKNAMES FOR THIS MACHINE section. Example 5-3 shows how to use the mailsetup program.

Example 5-3: Using mailsetup to Configure the sendmail.cf File

# mailsetup


.
.
.
NICKNAMES FOR THIS MACHINE
 
Are there any other names that are used to send mail to this
machine?  For instance, if you have changed this host's name
(or plan to in the near future), a nickname allows sendmail to
recognize both names, "pearly" and the nickname, as synonyms
for this machine.
 
Another good use for nicknames occurs when a host receives mail
from multiple different networks.  A host's name may not be the
same on all of the different networks.  Again, nicknames allows
sendmail to recognize these different names as synonyms for this
host.
 
Do you wish to enter nicknames for this machine (y/[n])? y
 
The following have been defined for the nicknames for server1 class:
 
add to list, delete from list, or continue on (a/d/c)? a
 
Enter additions to class (space or <cr> separated) - end list with a <cr>
 
? server1 server2 mail_hub
? [Return]
 
The following have been defined for the nicknames for server1
class:
 
    server1 server2 mail_hub
 
add to list, delete from list, or continue on (a/d/c)? c
 

.
.
.

If you have already set up mail using the mailsetup program, follow these steps to manually configure the /var/adm/sendmail.cf file:

Change your directory to /var/adm/sendmail.

Edit the server1.m4 file and add server1, server2, and mail_hub to the definition of _MyNicknames:

dnl -- Other names for me - aliases of my machine
define(_MyNicknames,    {server1 server2 mail_hub})dnl

Use the make command to update the server1.cf file:

# make -f Makefile.cf.server1
 
# mv sendmail.cf sendmail.cf.sav

# cp server1.cf sendmail.cf

Restart the sendmail program:
```
# /sbin/init.d/sendmail restart
```

To directly edit the sendmail.cf file, follow these steps:

Change to the /var/adm/sendmail directory.

Edit the sendmail.cf file and add the following line:
```
Cw server1 server2 mail_hub
```

Restart the sendmail program:
```
# /sbin/init.d/sendmail restart
```

5.5.5 Mounting the Disks

After you complete the preliminary steps, you can use the mail service. The final step is to mount the /var/spool/mqueue and /var/spool/mail directories on the server1 and server2 mail hub member systems. Perform the following steps on both mail hub member systems:

Disable the sendmail program:
```
# /sbin/init.d/sendmail stop
```

If your mail hub member systems are active servers, you must save the old /var/spool/mqueue and /var/spool/mail areas so you do not lose any mail or queue files:
```
# cd /var/spool
 
# mv mqueue mqueue.old
 
# mv mail mail.old
```
You can move the old mail files to the new /var/spool/mail area after it is set up. You can run any queue files later by using the following command:
```
# sendmail -q -oQ/var/spool/mqueue.old
```

Re-create the directories:
```
# mkdir mqueue
 
# mkdir mail
```

Mount the mail service spool areas:

# mount mail_hub:/var/spool/mqueue /var/spool/mqueue
# mount mail_hub:/var/spool/mail /var/spool/mail

Start the sendmail program:
```
# /sbin/init.d/sendmail start
```

Add the loopback mounts to the /etc/fstab file so they will mount on the next reboot. The lines in the /etc/fstab file should resemble the following:
```
/var/spool/mqueue@mail_hub /var/spool/mqueue nfs rw,fg 0 0
/var/spool/mail@mail_hub /var/spool/mail nfs rw,fg 0 0
```
You must specify the fg option to ensure that /var/spool/mqueue is NFS-mounted before the sendmail program starts. Do not put the mount command into the background to retry the mount if the original mount fails, because the sendmail program could start before the mqueue area is mounted. This situation causes problems because sendmail tries to use the mount point for the mqueue area instead of the mounted file system.

5.5.6 Defining a BIND Mail Exchange Record

You can define a BIND mail exchanger (MX) record in a database file, such as the /etc/namedb/hosts.db file, on the primary BIND server to point to all your mail hubs. The sendmail program uses the BIND MX record to define a list of mail machines that can receive mail sent to a specific address. See the DIGITAL UNIX Network Administration manual for detailed information about BIND MX records.

The sendmail program delivers the mail to the machine with the lowest specified preference, if possible. If that machine is not available, it tries the machine with the next lowest preference, and so on.

You can specify both mail hub member systems as the mail exchange for each system. The following example shows the mail exchange resource records:

;name		ttl	class	type	preference	deliver-to
 
server1.foo.com	IN	MX	1	server1.foo.com
 
			IN	MX	100	server2.foo.com
 
server2.foo.com	IN	MX	1	server2.foo.com
 
			IN	MX	100	server1.foo.com
 
mail_server.foo.com	IN	MX	100	server1.foo.com
 
			IN	MX	100	server2.foo.com

In this example, all mail going to server1 goes to server1 if it is available, because it has a preference of 1. If server1 is unavailable, then the mail goes to server2, which delivers the mail to the shared /var/spool/mail area. Using this configuration, mail continues to be delivered to either mail hub member system internal address as long as at least one mail hub member system is available.

5.6 Accessing NFS Services from Client Systems

To access a Network File System (NFS) service from a client system, you must edit two system files:

You must include the NFS service name and Internet address in a client system's local /etc/hosts file.

You must edit the /etc/fstab file on a client system and include the NFS service name and its exported mount points. See Section 5.3 for information on ASE exports files for services.