Dynamic Path Failover Driver for AIX
FibreQuik Host Bus Adapter
Installation and User Guide
Cambex Corporation
115 Flanders Road
Westborough, MA 01581
Customer support
Document:
Date:
Rev.:
081-468-025
12/07/06
H
Table of Contents
Path Management .................................................................................. 4
Load Balancing...................................................................................... 5
More than two paths............................................................................... 6
Multiple Hosts........................................................................................ 7
AutoRecover......................................................................................... 14
AIX Configuration................................................................................ 15
APPENDIX A............................................................................................. 28
1
Dynamic Path Failover for AIX
Introduction
The Cambex FibreQuik™ Fibre Channel host bus adapters
provide attachment of Fibre Channel devices to PCI or Micro
Channel based servers. They support the SCSI-FCP protocol.
When used with the Cambex DPF (Dynamic Path Failover)
software, two adapters in a system provide greater throughput
and total protection from failure of a fibre channel component.
In addition, the Cambex AutoRecover daemon, cbxdpfar, will
automatically detect when a failed path has been repaired and
make it available as a standby path. For arrays that support
preferred paths, AutoRecover will automatically resume
operation on the repaired path, otherwise, an operator may
manually change paths to do load balancing.
Storage (SAN) and communications (LAN) networks have
resulted in a physical and logical separation between the storage
device (typically a SAN-connected RAID array), the storage
user (typically an Enterprise Server, which we will refer to
below as the server) and the administration station (typically a
remote workstation).
As storage systems have evolved from JBOD (just bunch of
disks) to complex shared RAID systems, the management
software has also grown in complexity. The following sections
describe the different levels of management software.
3
SAN Management
SAN Management is the top layer. It is responsible for
assigning the required access for each server to each storage
device. The unit of storage is the LUN (Logical Unit Number)
which acts the same as a single disk directly attached to the
server. In reality, a LUN is made up of portions of a number of
disks under the direction of a RAID controller in the storage
device. SAN Management typically requires a process running
on each managed server communicating with the manager
running on the administering station.
Storage Management
The Storage Management layer, or RAID manager, is the layer
that communicates with the RAID controllers on the storage
device. This is the layer used to create, modify or delete LUNs.
The Storage Management layer may provide SAN Management
functions, but only for the specific storage device being
managed. Storage management typically runs on a server
attached to the storage device (in-band) but may also run over a
network (out-of-band).
Path Management
Path Management includes a variety of software for managing
the various fibre channel components such as HBAs (Host Bus
Adapters), hubs and switches. It is used to display path status
and allow control over multiple paths between servers and
storage. To avoid having to access a large number of unrelated
software interfaces, integration of the various Path Management
components is usually done using a management tool
4
communicating via a network connection to each managed
device.
The Cambex Dynamic Path Failover driver is a Path
Management component that allows multiple paths to a storage
device and provides automatic failover between available paths.
Cambex Dynamic Path Failover software provides both a
command line interface and access through smit, the standard
integrated management tool for AIX.
SCSI and FCP Drivers
AIX makes two types of device drivers available. SCSI device
drivers support SCSI-2 disk and tape devices. FCP device
drivers support newer SCSI-3 disk and tape devices, which
provides a larger address space for target ID and LUN. When
using the FCP device driver, the adapter will be configured with
a name like fscsi2.
Load Balancing
Selection of which path to use happens in one of three ways.
Static load balancing allows the operator to select which path is
used to access a LUN. This is the default method. Preferred
Paths allow certain RAID arrays to automatically specify a path
choice. See the next section on Preferred Paths for more
information. Finally, turning on Load Balancing will send I/O
on all available paths using round robin scheduling.
5
How Preferred Paths work
The basic idea of multipathing is fairly simple. Access a disk on
one of two paths. If the path fails, move I/O activity to the other
path. We build on that basis by adding more factors into the
configuration and seeing what effect they have.
A path is defined as an end-to-end connection through a SAN,
ignoring any intermediate routing points. A complex SAN may
have multiple connections from an HBA to a storage device, but
this is invisible to the software. The combination of one HBA to
one target is one path, regardless of how it is routed.
Multiple LUNs
The effect of multiple LUNs on a RAID controller is that we
can choose to distribute them so as to balance the I/O across
multiple paths. The choice can be made in one of two ways. If
one RAID controller has preference over the other, we call this
"controller preference". If the selection is made in the host
software we call this "target preference". A third type of
preference is to lock a LUN to a particular HBA, called "HBA
preference".
More than two paths
With four paths (for example) we have to sort the paths into an
order in which they are chosen when the primary path fails. This
order should depend on getting the best performance when a
path segment fails. For example, if our performance is limited
by the IOPS that can be handled by each RAID controller, then
we want to balance the LUNs between the RAID controllers,
6
and set up preferences so that the remaining paths to the same
controller are preferred over paths to the alternate controller.
Example 1) Suppose we have two HBAs, a switch, and two
RAID controllers. A total of four cables, and four paths to each
LUN (2 HBAs x 2 RAID controllers). We use the RAID
controller preference to select which controller each LUN is
accessed on. The target preference lets me balance those
accesses across the two HBAs.
Example 2) We have two switches, 2 HBAs and 2 RAID
controllers, a fully redundant system. We add an inter-switch
link for in-band management, but we do not want to use this as
a data path if there are any other paths available. We set the ISL
paths' target preference lower than the direct paths, so they will
be used last.
Example 3) A 2G RAID array is connected to a 2G switch,
which does speed matching to two 1G HBAs in the host. To
avoid saturating either of the 1G paths, We use HBA preference
when balancing the LUNs across the paths.
Multiple Hosts
When we have more than two hosts accessing the same storage,
we find that the limit is usually the number of controller ports.
The HBAs will run below the maximum bandwidth, but the
controllers need to be balanced. Use the controller preferences
and target preferences to balance the loads.
Normally, each host accesses different LUNs through either
controller, so all paths are independent. One special case is
where a RAID controller has LUN groups. All LUNs in a LUN
group are accessed on the same controller. If the different LUNs
7
are accessed by different hosts, we could have the problem of
"thrashing". One host tries to move the LUN group to preferred
controller A while the other host tries to move the group to
preferred controller B. Anti-thrashing code prevents this
problem in the multipathing software.
8
Adapter Specifications
PC2000LC-HPSP PC4000 (Single Port)
PC2000LC-A
PC4002 (Dual Port)
Bus
PCI
PCI-X 2.0
Data rate (each dir.)
Cable length - copper
- multimode optical
- single mode optical
Voltage
100/200 100/200/400 MB/s
30
500
10
70
m
3.3 or 5
7.2
3.3
8.0
10
V
W
oz
Power
Shipping weight
10
Trademarks
Dynamic Path Failover and FibreQuik are trademarks of
Cambex Corporation.
9
Software and Hardware Installation
Driver Installation (command line)
The examples assume that the driver CD is mounted at /mnt and
that you are installing on AIX 5.1. The name and revision level
of the uninstalled driver image may be examined by issuing the
command (as root):
installp -ld <device>
where <device> is the path to the drivers, such as /mnt/AIX4/
for the CD or the name of a driver directory on the hard disk.
Example output would be:
Fileset Name
==============================================
PC1000.driver.obj 2.5.10.25 I b usr
# PC1000 Fibre Channel Adapter Driver (AIX 5L)
Level
I/U Q Content
cbxdpf.driver.obj
5.1.0.23
I b usr
# Cambex Dynamic Path Failover Driver
Before installing a new version of the driver, delete the previous
version of the driver. Note that any file systems must be
unmounted, volume groups exported and SCSI devices must be
removed before removing the driver. Be sure to have a
complete backup of all data before updating the driver. The
command to remove the driver is:
installp -u <name>.driver.obj
where names are cbxdpf and PC1000 or PC4000 as shown
above.
10
The new driver may be installed on the system with the
command:
installp -acd <device> all
where <device> is as described above.
example:
installp -acd /mnt/AIX5.1 all
After installing the driver, see the README file for additional
information and release notes.
/usr/lpp/cbxfc/READ_ME
/usr/lpp/PC4000/READ_ME
/usr/lpp/cbxdpf/READ_ME
11
Driver Installation (smit)
The smit path is
Software Installation and Maintenance
Install and Update Software
Install and Update Software by Package Name
For item " INPUT device / directory" use /mnt/AIX5.1or
the directory where the image files are located. A list should
pop up with the name of the drivers. Select and press enter.
Also make sure that the following drivers are loaded from the
AIX installation media.
devices.fcp.disk.rte
devices.fcp.tape.rte
The recommended maintenance levels for AIX are specified in
the image file prereq section. If not at the correct maintenance
level, the driver installation will fail. The current level may be
found from the command:
instfix -i |grep ML
Installation may fail if conflicting drivers are present for the
same devices. Only one driver may be installed on AIX for a
given device type.
12
Hardware Installation
Power down the system and install the adapter in a free slot.
The PCI cards may be installed in either 32 or 64-bit PCI card
slots, however the 32-bit slots may limit performance at high
data rates. Make sure that any cables do not block airflow to the
card. Observe proper static precautions. Detailed procedures
for opening the case and for suggested adapter placement may
be found in the Operator Guide for the specific server model.
When using copper cable, it is extremely important that the
building wiring present a low impedance ground path to the
frame of the device and the host with which it will
communicate. Noise due to copiers or large machinery sharing
the electrical circuits is a common cause of signal integrity
problems.
13
AutoRecover
cbxdpfar, the AutoRecover daemon, is automatically installed
and enabled when the DPF driver is installed. Control of the
daemon is done using SRC.
stopsrc -s cbxdpfar Stops AutoRecover
startsrc -s cbxdpfar Starts AutoRecover
AutoRecover works by periodically polling the standby paths to
each LUN. If the polls fail, the path is marked as failed. If the
polls succeed, the path is marked as available for use. If path
preferences are set up, the AutoRecover daemon will move I/O
to the preferred paths when they become available.
AutoRecover also deletes paths to hdisks that have been
removed from the system. See the section on SMIT for
changing the default AutoRecover timing parameters.
14
AIX Configuration
During power-up initialization, AIX will find and configure the
new adapter(s) and any supported devices using the driver's
configuration method. Devices are assigned a SCSI ID as they
are discovered starting with 0. SCSI ID is not related to loop
ID, port ID or to SCSI ID assignment on other hosts. It is a
logical value bound to the fibre channel WWN of the device.
Configuration may also be run from the command line:
cfgmgr
Once disks have been configured, the standard AIX utilities
may be used to build volume groups, logical volumes and file
systems. See the AIX System Management Guide for a detailed
explanation.
Find out the names of the installed adapters with the command:
lsdev -Cc adapter
In arbitrated loop, the ID of each device must be unique. For
disk drives, the ID is typically set by switches on the drive
enclosure. The adapter ID is soft assigned and may be
examined after it is participating with the mbtest command. If
the adapter name found by the lsdev command were fscsi1, the
command would be:
/usr/lpp/cbxfc/mbtest fscsi1 id
The requested ID may be changed in the database to any non-
conflicting value in the range 0 to 126 with the command:
chdev -l fscsi1 -a id=6 –P
The ID attribute has no effect in a fabric environment.
15
Persistent Bindings
Persistent Bindings is the term used to indicate the mapping of
SCSI ID to FC Worldwide Name (WWN). By keeping a
persistent binding, the loop ID or port ID of a device may
change and the device will still be properly addressed.
In the Cambex AIX driver, persistent binding is automatic. At
configuration time, all devices that are found are automatically
assigned to a SCSI ID and a persistent binding is created.
Persistent bindings are by portname and are kept in the ODM
database. They are preserved and restored at reboot for all FC
devices.
To display the SCSI ID to WWN mapping for all adapters and
devices, issue the command:
/usr/lpp/cbxfc/fctest
When replacing a failed device on a SAN, we want to keep the
same SCSI ID but change the WWN binding to the
replacement. A perl script is provided for doing this:
/usr/lpp/cbxfc/FRU_replace
16
SMIT Commands
Commands
The adapter management commands available in SMIT are in
the devices menu under Cambex Fibre Channel Adapter:
List all Cambex FC adapters
This provides the name and location of each adapter.
Change / Show Characteristics of a Cambex FC Adapter
Allows the user to display and modify attributes. Attributes
(and default values) include:
exec_throttle (256) number of simultaneous I/O operations.
dma_window (64) PCI DMA memory window in MB.
retry_delay (3) Delay in seconds after bus reset or other error.
logout_delay (7) Time in seconds allowed for ports to be logged
out before failing I/O.
topology_list (LOOP_P2P) Automatically detect topology.
May also be forced to LOOP_ONLY or P2P_ONLY.
frame_size (2048) Maximum FC frame size. Choices are 1024
or 2048.
infinity (300) Maximum SCSI command timeout in seconds.
Set to 0 to disable.
GHz (auto) Auto-detect fibre channel rate in GHz. May also be
forced to 1 or 2. Ignored on 1GHz cards.
17
sw_fc_class (3) Fibre channel class.
devtype_exclude ( ) List of SCSI device types to ignore during
configuration. For example, specifying "1,5" would exclude
both tape and CD devices.
Attributes only used in loop mode include:
id (0) Default L_ID for adapter. See table of L_ID to ALPA.
login_on_lip (yes) Full login if LIP occurs.
full_duplex (no) Use full duplex loop open, OPNyx.
Generate Error Report
Accesses the AIX error logging facility.
Trace Cambex FC Adapters
Runs the AIX trace facility, looking only for the entries specific
to the adapter driver.
Path Management Commands
The path management commands available in SMIT are in the
devices menu under Cambex Dynamic Path Management:
List all Cambex Dynamic Paths
This provides the information on all paths and devices known to
the DPF driver. Use this when installing to verify that all paths
have been found and configured.
Change / Show Characteristics of a Cambex Dynamic Path
Displays the current path status and allows the user to place
paths online/offline or to change path preferences.
18
The autorecover daemon will automatically move operation to
preferred paths. The preferred path is the combination of path
preference values for HBA, device and path. For some arrays,
target device preference is read automatically from the array at
configuration time. Consult the array user manuals for more
information.
Change / Show Characteristics of Cambex Managed HBA’s
Used to display and change HBA preference for one or more
adapters.
Trace Cambex Dynamic Paths
Runs the trace facility, looking only for the entries specific to
the DPF driver and adapter driver.
Manage AutoRecovery Subsystem
Provides the following functions:
Show Status of AutoRecovery Subsystem
Start AutoRecovery Subsystem
Stop AutoRecovery Subsystem
Change / Show AutoRecovery Subsystem Parameters
Parameters are:
-n
-t
-r
Number of successful inquiries (3 - 9)
Seconds between inquiries (3 - 3599)
Seconds between ODM updates (automatic)
Manage Load-Balancing
Enable or disable dynamic load balancing for the selected disks.
Manage Licensing
Allows the user to enter a new license for the DPF software,
overwriting the existing license, or to display the expiration date
of the current license. To enable a one-time-use, 30-day license
19
for use until the purchased license is received. Issue the
commands:
cd /usr/lpp/cbxdpf
mv license.30day license
cfgmgr
Licenses for OEM versions of the driver are obtained from the
storage vendor who provided the software.
The generic version of DPF purchased from Cambex is licensed
for use only on Cambex cards and on only one LPAR. A
license token is provided with the purchase. Contact Cambex
support by email to obtain a license. You will need to provide
The license tokens for each card
Host machine ID from 'uname –a -L'
The output of /usr/lpp/cbxfc/fcinfo
How to set path preferences
Controller preference is set on the RAID controller using RAID
Management software provided by the storage vendor. HBA
preference and Target preference are set in the host using the
dpfutil Path Management software, or in the smit menus. The
preference value, although internally combined numerically, is a
yes/no value in the user interface. HBA preference outweighs
controller preference. It defaults to yes (all HBAs are preferred)
but can be changed to no with the tool. Target preference ranks
lower than controller preference and defaults to no (no preferred
path) but may be changed to yes with the tool.
Path priority is the combination of the following factors, which
are listed from highest to lowest.
20
• failed or offline status is considered first. A path will not
be used if failed or offline.
• Anti-thrashing prevents paths to the wrong controller
from causing thrashing.
• HBA preference - normally all HBAs have equal
preference to paths. To block an hba from using a path,
do:
Command line:
dpfutil hba_set_wwn cbx0 no (default is yes)
to list HBA preferences:
odmget -q"attribute=hba_pref" CuAt
in smit:
Devices
-> Cambex Dynamic Path Management
-> Change / Show Characteristics of Cambex Managed
HBAs
• Controller preference - this is for multi-controller RAID
arrays where one or more of the controllers reports that
it owns the LUN. Controllers which do not report
preference, are treated equally.
•
Target preference - To make one path higher preference
than other paths to the same LUN, use target
preference.
Command line:
dpfutil target_set_wwn cbx10 yes (default is no)
to list target preference:
odmget -q"attribute=tgt_pref" CuAt
in smit:
Devices
-> Cambex Dynamic Path Management
-> Change / Show Characteristics of Cambex
Dynamic Paths
-> Change / Show Physical Path Routing
Preferences
21
Troubleshooting
Use the errpt command to identify the cause of failure. The
lsdev command will provide the location of the component. For
example, if errpt identifies a loose cable on device hdisk1, lsdev
will give an address similar to:
hdisk1 Available 00-02-01
Cambex SCSI...
and lsattr -El hdisk1 will show:
clr_q
no
Device CLEARS it’s
Queue on error
location
lun_id
node_name
pvid
Location Label
Logical Unit Num ID
0x20000080e511b744 FC Node Name
0x0
none
yes
Physical Volume id
Use QERR bit
q_err
q_type
queue_depth 16
reassign_to 120
rw_timeout
scsi_id
simple
Queuing TYPE
Queue DEPTH
REASSIGN time out
READ/WRITE timeout
SCSI ID
30
0x0
start_timeout 60
ww_name
START UNIT timeout
0x23000080e511b744 FC World Wide Name
Which indicates that the adapter is in card location 00-02, and the
drive is at SCSI ID 0, LUN 0.
22
Checklist for driver and HBA installation
The following commands may be used to isolate problems
during installation of the fibre channel driver and card. Type
the command shown in bold and verify the output.
lslpp -l PC1000.driver.obj
must give:
Fileset
Level
State
Path: /usr/lib/objrepos PC1000.driver.obj
01.05.00.00 (or greater)
Committed
Description Cambex Fibre Channel Adapter Driver
lsdev -Cc adapter should print a line for each card:
fscsi<x>
<x> is a unique number for each SCSI card
Available <SS-SS> <S> is the location in the machine where
the card is installed
Check that the number of SCSI cards and their slots are correct
If 'Available'then card and driver are installed.
If 'Defined'or not listed then card was not found.
lscfg -vl fscsi4 should print more information for a card:
DEVICE
LOCATION
DESCRIPTION
Cambex Fibre
fscsi4
00-03
Channel I/O Controller
Manufacturer................CAMBEX
Machine Type and Model......MC1000F
ROS Level and ID............V2.01.04
Serial Number...............
2354
Displayable Message.........Cambex Fibre ...
Network Address.............500502E020000932
Device Specific.(Z0)........Build:TMNX
cfgmgr -v >/tmp/cfg.log 2>&1
Should identify the card and configure it.
23
If cfg.log contains the ID 7710002x and no driver is found,
then the adapter driver is not installed.
During boot, the number displayed when configuring the
adapter is hex 0x911 (FC adapter).
/usr/lpp/cbxfc/mbtest fscsi1
This test code may be used to get fibre channel status from the
adapter or from attached devices. Type the command as above
for a list of subcommands.
/usr/lpp/cbxfc/mbtest fscsi1 fab
This command will list the persistent bindings between SCSI ID
and WWN on adapter scsi1. It will also list the loop ID and
port ID (fabric address) which can change dynamically as
devices are connected or disconnected.
/usr/lpp/cbxfc/mbtest fscsi1 getp
This command will show the fibre channel ports which are
currently logged in to adapter scsi1.
/usr/lpp/cbxfc/sctest fscsi1
This test code may be used to get IOCINFO information from
the adapter or to see the results of SCSI Inquiry or
TestUnitReady commands to SCSI devices attached to the
adapter. Type the command as above for a list of
subcommands.
/usr/lpp/cbxdpf/dpfutil listall
This will list the failover devices and paths. The active path
will be shown on the left. Status is shown by letter codes:
“F’ indicates a failed path.
“L” indicates dynamic load balancing is active.
“O” indicates offline paths.
24
The message no failover devices available indicates that the
attached storage is not yet recognized by the DPF driver. This
could mean that a platform kit needs to be installed for your
storage or that the storage is not yet configured.
/usr/lpp/cbxfc/runtrace
This should be run to start a trace if Cambex support requests a
trace of the card failure. To stop the trace and generate a trace
report, type:
trcstop; trcrpt > /tmp/trc.log
Questions or problems related to Cambex fibre channel adapters
25
Error Messages
Various errors and notifications will be placed in the AIX error
log. These can be read using the errpt command, or automatic
notification may be set up.
DPF_FAILOVER
This indicates that the active path to a device has failed, and that
DPF successfully transferred I/O to an alternate path.
DPF_STDBYDN
This indicates that a path other than the active path has failed to
a device.
DPF_UNRECOV
This indicates that the active path to a device has failed, and that
DPF was not successful in transferring I/O to an alternate path.
DPF_OFFLINE
DPF_ONLINE
These messages indicate that a path has been manually set to the
offline or online state.
DPF_LIC_WARNING
DPF_LIC_EXPIRE
These messages indicate that a temporary license for the DPF
software is either close to or has expired. Once the DPF license
has expired, you will need to uninstall the DPF driver in order
to regain access to the hdisks.
26
Upgrading the Driver
When upgrading to a newer version of the driver there are two
methods:
1) Use rmdev to remove all dependent devices, remove the
existing driver(s), then install the new driver(s) and run cfgmgr
to recreate the devices. This method allows a new driver to be
installed without rebooting the server. File systems which are
unmounted before the process may be mounted again
afeterwards without affecting the data.
2) Install the new driver with devices in place. The server must
be rebooted afterwards so that the driver on disk will match the
driver in memory.
When upgrading AIX to a new major version, remove the
drivers beforehand and install the new versions after completing
the AIX upgrade.
27
Copper HSSDC Cable
Duplex SC Optical GBIC
29
Small form factor (SFF) optical LC cable
30
PC4000 LED Meanings
Y G A
Y
G
A
Off
On
Flash
Off
Off
Off
On
Flash
Off
On/Flash Off
Off
Off
On
Flash
No Power
Power Applied
Initializing *
On/Flash 1G (Flash indicates active)
2G (Flash indicates active)
4G (Flash indicates active)
On/Flash Off
∗
Flashing in unison indicates initialized, while
alternate flashing indicates a firmware error.
Y G A
32
Declaration of conformity
We Cambex Corporation of Westborough, Massachusetts, USA, declare
under our sole responsibility, that the products PC2000LC-HPSP,
PC2000LC-A, and PC4000 are in conformity with the following standards:
EN 55022:1994/A1:1995/A2:1997 Class A ITE emissions requirements
EN 50082-1:1992 EMC residential, commercial and light industrial generic
immunity standard
Warning
This is a Class A product. In a domestic environment this product may cause
radio interference in which case the user may be required to take adequate
measures.
33
Safety Notices
FCC Warning
This equipment has been tested and found to comply with the limits for a
class A computing device pursuant to Part 15 of FCC rules which are
designed to provide reasonable protection against interference when
operating in a commercial environment. This equipment generates, uses and
can radiate radio frequency energy and if not installed and used in
accordance with the instruction manual may cause interference to radio
communications. Operation of this equipment in a residential area is likely
to cause interference in which case the users will be required to take
whatever measures may be required to correct the interference at their own
expense. Changes or modifications of this device could affect the user's
authority to operate this equipment.
Safety Notice
Class 1 Laser Product
Versions of this product supplied with an optical transceiver meet the safety
requirements of a class 1 laser as specified by the US Food and Drug
Administration Center of Drug and Radiological Health (CDRH) 21 CFR
1040.10 & 1040.11 and by IEC 825-1. The U.S. Department of Health and
Human Services (DHHS) does not consider Class 1 lasers to be hazardous.
Customer supplied laser GBICs must meet the same requirements. If in
doubt, contact Cambex customer support before operating the product.
The International Electrotechnical Commission (IEC) 825 Laser Safety
Standard requires labeling in English, German, Finnish, and French stating
that the product uses Class 1 lasers. Because it is impractical to label the
transceiver, the following label is provided in this manual.
CLASS 1 LASER PRODUCT
LASER KLASSE 1
LUOKAN 1 LASERLAITE
APPAREIL A LASER DE CLASSE 1
TO IEC 825 (1984) + CENELEC HD 482 S1
34
Dynamic Path Failover License Agreement
Cambex Dynamic Path Failover (the Software) is a licensed product. The
user must consent to the provisions of this license agreement in order to use
the Software. Use of the Software beyond the number of purchased licenses
is prohibited.
The user may not cause or permit copying, modification, rental, sublicense,
lease, transfer, reverse engineering, or disassembly or decompilation of the
Software, or creation of derivative works based on the Software.
This license for the Software remains in effect until terminated by either
party upon written notice. Upon termination of the license for the Software,
the user must discontinue use of and destroy or delete any copies of the
Software.
Except as limited by applicable law, Cambex’s total liability under this
agreement is limited to the purchase price of the Software. Dynamic Path
Failover may not be used to provide life-critical protection such as operation
of medical equipment or control of nuclear facilities. The Software is
intended to provide improved availability of data but makes no claim for
greater reliability or protection of data, and Cambex may not be held liable
for data loss or consequential damages.
Any provision of this agreement found unenforceable will not affect other
provisions. This agreement is governed by laws of the Commonwealth of
Massachusetts.
35
|
