HP CIFS Server Administrator Guide Version A.02.04.04 (5070-6710, October 2011)

4. For any UNIX users used to authenticate CIFS clients, check that they have the same name,
user ID number, primary group and password on both of the nodes.
This is required for any users used to authenticate to either Samba server in the Active-Active
configuration. This means that any user name used on both Samba servers must have the same
user ID, primary group ID, and password on both cluster nodes. If this isn't the case, you
cannot use Samba as an Active-Active server for this MC/ServiceGuard cluster.
5. Check that the RUN_SAMBA and RUN_WINBIND parameters in the /etc/rc.config.d/samba
file is set to 0 on both nodes.
Configure a Highly Available HP CIFS Server
Introduction
Before configuring the MC/Serviceguard packages, it is important to understand how HP CIFS
Server is able to support active-active configurations.
The HP CIFS Server permits multiple instances of its NetBIOS and SMB master demons.
Each CIFS Server has its own smb.conf file to define its behavior. The NetBIOS name and IP address
that the client connects to is used to decide which smb.conf file is used for the connection. This
multiple CIFS master demon configuration allows HP CIFS to run multiple MC/ServiceGuard
packages simultaneously.
When a failover occurs, MC/ServiceGuard transfers the IP address from the failing cluster node
to another node. When MC/ServiceGuard moves the package from the failing cluster node to the
other node, it activates the appropriate CIFS Server on a remaining node. With the IP address
switched, all the traffic that was going to the failed node now goes to the other active node. The
key is to have a CIFS Server configured to look and act just like the CIFS Server that was running
on the original node.
Load balancing between systems while all systems are up can be achieved by having the CIFS
shares accessible only through certain CIFS Server names (NetBIOS names). Keep this in mind
when you associate the CIFS shares and directories with logical volumes during server configuration.
Note that each cluster node needs to know all the UNIX users that connect to the samba servers
(packages). This means that the /etc/passwd file may need to be updated. For NIS installations,
you can generate new maps using theypmake or similar tool if there are new passwd or group
files. For LDAP installations, you may generate LDAP data for new accounts using the migration
tools provided by the LDAP-UX Integration product. These tools are found in
/opt/ldapux/migrate and documented in the LDAP-UX Client Services Administrator's Guide
available at http://docs.hp.com/hpux/internet.
Instructions
The following instructions are for one of the MC/ServiceGuard package. You will have to go
through these steps for each CIFS server package (one for each node). You will then need to copy
all the files to all nodes in your cluster.
When complete, each HP-UX system will have a package using the unique name for each node
in the cluster, though only the package corresponding to itself will be active until a failover occurs.
For example, if you have a three node cluster, you will have three packages on each of the three
HP-UX systems.
There will be three cluster directories:
1. /etc/cmcluster/samba/pkg1
2. /etc/cmcluster/samba/pkg2
3. /etc/cmcluster/samba/pkg3
There will be three configuration files:
144 Configuring HA HP CIFS