Managing a Dual-LAN Client Configuration
ACSLS 5.2 and above supports dual-LAN client configurations, which consist of a primary LAN and a secondary (backup) LAN. If the primary LAN fails, the client switches to the secondary LAN. ACSLS removes all outstanding messages to that client and begins communicating using the secondary LAN. Because ACSLS removes all outstanding messages before switching over to the secondary LAN, a client will not receive a success message even though a request completed successfully.
For example, a client requests ACSLS to eject ten cartridges. ACSLS starts ejecting the cartridges, then communications fail between ACSLS and the primary client LAN. ACSLS removes all outstanding messages to that client and begins communicating through the secondary LAN. ACSLS successfully completes ejecting all ten cartridges, but sends no success message to the client. The client must verify that the request completed successfully. In this example, if the client issues a query volume request against the IDs of the ejected volumes, ACSLS returns a volume not found
error message, which confirms that ACSLS ejected the cartridges.
ACSLS ensures that any transient requests complete successfully if a LAN communications switch over occurs. After switch over, however, the client must cancel any persistent outstanding requests (such as CAP operations) originally submitted on the primary LAN and resubmit these requests through communications from the secondary LAN. Resources (such as, CAPs, locks, drives, and so forth) allocated through primary LAN communications will remain allocated after switchover to the secondary LAN.
The following sections tell how to configure ACSLS for dual-LAN client operations by doing the following:
-
Register the IP addresses of primary and secondary LANs by creating a
csc_ip_switch.dat
file -
Install a second ethernet port for a multi-homed ACSLS server
-
Set the TCP/IP connection timeout interval to decrease the system switchover time to the backup LAN.
Registering the IP Addresses of Primary and Secondary LANs
To register the IP addresses of the primary and secondary LANs with ACSLS, create the file csc_ip_switch.dat
in the $ACSSS_HOME/data/internal/client_config/
directory. The following is an example of a csc_ip_switch.dat
file.
#The following entry is System Zed's primary and secondary LAN IP addresses. 129.80.30.40 129.80.30.50
As shown in the example above, the # sign precedes a comment. The entries consist of a client system's primary LAN IP address in the left column followed by one or more blanks, then a client systems' secondary LAN IP address in the right column. In this example, System Zed's primary LAN IP address is 129.80.30.40
and its secondary LAN IP address is 129.80.30.50.
After you create or update the csc_ip_switch.dat
file, if ACSLS is running, you must stop and restart ACSLS.
If ACSLS successfully reads the csc_ip_switch.dat
file on restart, ACSLS logs the success message 2010 I DUAL PATH OPTION ACTIVATED
in the he ACSLS Event Log. Otherwise, dual-LAN support is not activated.
Installing a Second Ethernet Port for a Multi-Homed ACSLS Server
This section describes the procedure for installing a second ethernet port for a multi-homed ACSLS server. The second ethernet port provides the attachment to the second control path adapter, which controls the backup LAN. This procedure requires the SBUS Buffered Ethernet card. You can order this Ethernet card from Oracle as part number X1053A.
The installation procedure includes:
-
Installing the hardware and rebuilding the kernel
-
Defining a host name for the new ethernet port
-
Creating the
/etc/notrouter
file
To install a second ethernet port, do the following:
- Install the SBus Buffered Ethernet card according to the manufacturer's instructions.
- Reconfigure the system kernel for the new device:
- Power the system on. When it begins to boot, press
[[STOP]]-[[A]]
to enter the PROM monitor. - At the
ok
prompt, boot the server:boot -r
- Power the system on. When it begins to boot, press
- When the boot has completed, login as
root
. - Create a hostname for the second ethernet port:
echo 2nd_host_name > /etc/hostname/hme1
Where
2nd_host_name
is the hostname for the second ethernet port. - Enter the following command:
touch /etc/notrouter
This creates the
/etc/notrouter
file. - Reboot the server:
reboot
This completes the procedure.
Setting the TCP/IP Connection Timeout Interval
The UNIX system variable tcp_ip_abort_cinterval
sets the TCP/IP connection timeout interval between the client and the ACSLS server. Changing the default value (180 seconds) of this variable may decrease the switchover time to the backup LAN. Note, however, that actual switchover time depends on the configuration and type of failure.
For example, if an ACSLS server port fails and multiple clients are communicating through this port, ACSLS serially recovers communications with each client. Multi-client recovery for a failing port therefore takes longer than recovery with a single client communicating with a failing port.
To decrease the switchover time to the backup LAN, do the following:
- Log in as
root
on the ACSLS server. - From the prompt, enter the following command:
/usr/sbin/ndd -set /dev/tpc tcp_ip_abort_cinterval 15000
This command changes the TCP/IP connection timeout interval to 15 seconds (the default is 180 seconds).
HINT: To make this command persistent (across server reboots), add the command to the
/etc/rc2.d/S69inet
file in the “Set configurable parameters" section.