![[Contents]](TOC.GIF){kind=small}
![[Prev. Chapter]](REW.GIF){kind=small}
![[Next Section]](NEXT.GIF){kind=small}
![[Index]](INDEX.GIF){kind=small}
![[Help]](HELP.GIF){kind=small}
These release notes provide important information about Version 1.5 of the TruCluster software products.
Version 1.5 introduces the features listed in Table 1-1. The table indicates which features are provided by TruCluster Production Server Software (PS), TruCluster MEMORY CHANNEL Software (MC), and TruCluster Available Server Software (AS). The abbreviation "N/A" indicates that the listed feature is not applicable to a particular product.
| Feature | PS | MC | AS | For more information, see: |
| Support for AlphaServer 1200 systems. | Yes | Yes | Yes | Hardware Configuration |
| Support for the HSZ70 RAID controller. | Yes | N/A | Yes | Hardware Configuration |
| Support for tape devices on a shared SCSI bus with server failover for the POLYCENTER NetWorker Save and Restore product. This allows a system administrator to configure a single, highly available backup service for each ASE. | Yes | N/A | Yes | Hardware Configuration, Administration |
| Support for up to 8 member systems in a cluster (up to 4 per ASE). | Yes | Yes
Eight-member MEMORY CHANNEL API clusters have been supported in previous releases of the TruCluster MEMORY CHANNEL Software product. MEMORY CHANNEL API clusters do not contain ASEs |
No
The 4-member restriction still applies to the single ASE within an Available Server configuration. |
Hardware Configuration |
| Support for failover across MEMORY CHANNEL interconnects in a redundant MEMORY CHANNEL configuration when their adapters are installed on different PCI buses on member systems. | Yes | Yes | N/A | Hardware Configuration |
| Support for multiple active logical rails for applications using the MEMORY CHANNEL application programming interfaces (API) library. | Yes
Production Server cluster operation supports only the failover pair mode of operation. Only applications using the MEMORY CHANNEL API library can utilize multiple active logical rails. This mode of operation does not allow for failover between logical rails. |
Yes | N/A | Section 1.7 of these release notes, MEMORY CHANNEL Application Programming Interfaces |
| Other enhancements to the MEMORY CHANNEL API library. | Yes | Yes | N/A | Section 1.7 of these release notes, MEMORY CHANNEL Application Programming Interfaces |
| Support for asynchronous transfer mode (ATM) as a primary, backup, and client network in an Available Server configuration, and as a client network in a Production Server cluster | Yes | N/A | Yes | Hardware Configuration, Administration |
| Introduction of persistent resource locking in the distributed lock manager (DLM) to decrease recovery time for database applications. | Yes | N/A | N/A | dlm_rd_attach(3),
dlm_rd_collect(3),
dlm_rd_detach(3),
dlm_rd_validate(3) |
Extensions to the
asemgr
utility to display service and member information from the ASE database. |
Yes | No | Yes | Administration,
asemgr(8) |
| Ability to update a service's LSM or AdvFS storage configuration, as well as other service information (such as Automatic Service Placement policy and an NFS service's exports file), in the ASE database without interrupting a service's availability to its clients. | Yes | N/A | Yes | Administration |
| Improved performance of DRD remote read operations, as a result of peer-to-peer direct-memory access (DMA), in configurations in which the storage adapters and MEMORY CHANNEL hardware are on the same PCI segment. | Yes | N/A | N/A | Administration,
drd(7),
drd_dma(8) |
| Support for non-ASE member systems in a Production Server cluster. | Yes | N/A | N/A | Software Installation, Administration |
| Extensions to the Cluster Monitor | Yes | N/A | Yes | Administration, Cluster Monitor online help |
| Year 2000 readiness | Yes | Yes | Yes | The DIGITAL UNIX Year 2000 Readiness white paper (in HTML format) on the DIGITAL UNIX V4.0D Documentation, Volume 1 CD-ROM |
You cannot take advantage of Version 1.5 features until the entire cluster has been upgraded to Version 1.5. For example, you cannot add a fifth member or add a non-ASE member to a cluster until the upgrade is complete. Please read the TruCluster Software Products Software Installation manual before installing any of the TruCluster software products.
Depending upon which TruCluster software product you are using, you should read certain sections of this manual, as follows:
Users of the TruCluster Production Server Software should read the entire document.
Users of the TruCluster MEMORY CHANNEL Software should read Section 1.1, Section 1.2, Section 1.6, and Section 1.7.
Users of the TruCluster Available Server Software should read Section 1.1, Section 1.2, Section 1.4, Section 1.5, Section 1.8, Section 1.9, and Section 1.10.
![[Prev. Section]](PREV.GIF){kind=small}
You must set up the TruCluster hardware configuration before you install any TruCluster software product. This section lists the hardware supported by (and, in some cases, required by) Version 1.5 of each TruCluster software product. To obtain more detailed information on the role any hardware component plays in a cluster or available server environment (ASE), as well as configuration assistance, see the TruCluster Software Products Hardware Configuration manual. That manual also provides procedures for verifying the firmware revision level of individual hardware components.
Table 1-2 lists the systems supported by the TruCluster Version 1.5 software products, indicating which are supported by TruCluster Production Server Software (PS), TruCluster MEMORY CHANNEL Software (MC), and TruCluster Available Server Software (AS). See the pertinent Software Product Description (SPD) for a detailed list of supported models in each system family. All systems must be running DIGITAL UNIX Version 4.0D.
Table 1-2 also lists the minimum Alpha System Reference Manual (SRM) firmware version required for each system. For systems running DIGITAL UNIX Version 4.0D, firmware is distributed on the Alpha Systems Firmware Update Version 5.0 CD-ROM.
| System | Minimum SRM Firmware for Version 4.0D | PS | MC | AS |
| DEC 3000 | V7.0 | No | No | Yes |
| DEC 7000 and DEC 10000 | V1.7 | No | No | Yes |
| AlphaServer 300 | V6.6 | No | No | Yes |
| AlphaServer 400 | V6.6 | No | No | Yes |
| AlphaServer 800 | V5.0 | Yes | Yes | Yes |
| AlphaServer 1000 | V5.0 | No | No | Yes |
| AlphaServer 1000A | V5.0 | Yes | Yes | Yes |
| AlphaServer 1200 | V5.0 | Yes | Yes | Yes |
| AlphaServer 2000 | V5.0 | Yes | Yes | Yes |
| AlphaServer 2100 | V5.0 | Yes | Yes | Yes |
| AlphaServer 2100A | V5.0 | Yes | Yes | Yes |
| AlphaServer 4000 | V5.0 | Yes | Yes | Yes |
| AlphaServer 4000A | V5.0 | Yes | Yes | Yes |
| AlphaServer 4100 | V5.0 | Yes | Yes | Yes |
| AlphaServer 8200 and 8400 | V5.0 | Yes | Yes | Yes |
Table 1-3 describes the MEMORY CHANNEL hardware supported by TruCluster Production Server Software Version 1.5 and TruCluster MEMORY CHANNEL Software Version 1.5. MEMORY CHANNEL hardware is not supported by TruCluster Available Server Software.
| Hardware | Description | MEMORY CHANNEL Model Number |
| MEMORY CHANNEL adapter | Adapter used for a cluster interconnect connection. | CCMAA-AA or CCMAA-BA |
| MEMORY CHANNEL hub | PC-class enclosure that is populated with line cards and used to connect MEMORY CHANNEL adapters. A hub is required if you have more than two member systems. | CCMHA-AA |
| Line card | Line card for a MEMORY CHANNEL hub. | CCMLA-AA |
| Copper link cable | Cabling for adapter-to-hub and adapter-to-adapter (virtual hub) connections. | BC12N-10 (10 meters) |
Table 1-4 indicates which SCSI-2 adapters are supported by TruCluster Production Server Software Version 1.5 (PS) and TruCluster Available Server Software Version 1.5 (AS).
| SCSI-2 Adapter | Hardware Revision | Firmware Revision | PS | AS |
| KZPSA PCI-to-Fast-Wide Differential [Footnote 1] | F01 | A11 or higher | Yes | Yes |
| KZTSA TURBOchannel-to-Fast-Wide Differential | F01 | A09 or higher | No | Yes |
| KZMSA XMI | F03 | V5.6 or higher | No | Yes |
| PMAZC TURBOchannel Dual-Channel | N/A | V1.8 or higher | No | Yes |
Table 1-5 lists the RAID controllers supported by TruCluster Production Server Software Version 1.5 (PS) and TruCluster Available Server Software Version 1.5 (AS). It indicates which Hierarchical Storage Operating Firmware (HSOF) revisions are required for each RAID controller.
| RAID Controller | Firmware Revision | PS | AS |
| HSZ10-Ax | 0306 | No | Yes [Footnote 2] |
| SWXRA-Z1 (HSZ20) | HSOF V3.0 | Yes | Yes |
| HSZ40-Ax | HSOF V2.0 or V2.5 | Yes | Yes |
| HSZ40-Bx | HSOF V2.5 or higher | Yes | Yes |
| HSZ40-Cx | HSOF V2.7 or higher | Yes | Yes |
| HSZ50-Ax | HSOF V5.0 or higher | Yes | Yes |
| HSZ70 [Footnote 3] | HSOF V7.0 | Yes | Yes |
The following hardware revisions of SCSI signal converters are required:
DWZZA-AA--Revision E01 or later
DWZZA-VA--Revision F01 or later
DWZZB-AA--Revision A01 or later
DWZZB-VW--Revision A01 or later
Table 1-6 lists the disk devices supported by TruCluster Production Server Software Version 1.5 and TruCluster Available Server Software Version 1.5 on a shared SCSI bus. It indicates the firmware levels that are known to work for each disk at the time of this manual's publication.
| Disk | Data Path | Firmware Revisions |
| RZ26-VA | Narrow | 392 or 392A |
| RZ26L-VA | Narrow | 442D |
| RZ26L-VW | Wide | 442E |
| RZ26N-VA/RZ26N-VW | Narrow/Wide | 0466, 0568, 0616, or 1103 |
| RZ28-VA | Narrow | 442C or 442D |
| RZ28-VW | Wide | 442E |
| RZ28B-VA | Narrow | 0006 |
| RZ28D-VA/RZ28D-VW | Narrow/Wide | 0008 |
| RZ28L-VA/RZ28L-VW | Narrow/Wide | LYJ0 |
| RZ28M-VA/RZ28M-VW | Narrow/Wide | 0466, 0568, 0616, or 1103 |
| RZ29B-VA/RZ29B-VW | Narrow/Wide | 0014 or 0016 |
| RZ29L-VA/RZ29L-VW | Narrow/Wide | LYJ0 |
| RZ40-VA/RZ40-VW | Narrow/Wide | 0305 |
| RZ40L-VA/RZ40L-VW | Narrow/Wide | LYJ0 |
| RZ1BB-VW | Wide | LYJ0, 0844, or 0845 |
| RZ1CB-VW | Wide | LYJ0, 0847 |
| RZ1DB-VW | Wide | LYJ0, 0307 |
Table 1-7 lists the tape devices supported by TruCluster Production Server Software Version 1.5 and TruCluster Available Server Software Version 1.5 on a shared SCSI bus. It indicates the firmware levels that are known to work for each tape at the time of this manual's publication.
| Tape Device | Firmware Revisions |
| TZ885 | V52 [Footnote 4] |
| TZ88 | V52 |
| TZ89 | V31 |
You can find a list of patches for reported problems that have been corrected in this release on the DIGITAL UNIX Operating System Associated Products Volume 2 CD-ROM, in the following file:
<mountpoint>//TCR/doc/txt/TCR_V1.5_Subsumed_Patches.TXT
The following notes apply to the use of network adapters in TruCluster Production Server clusters or TruCluster Available Server configurations:
TruCluster Available Server Software and TruCluster Production Server Software do not support the operation of the DE500-XA and DE500-AA PCI/Ethernet network adapters in full-duplex mode, either at 10 Mbits/sec or 100 Mbits/sec. Both adapters are supported in half-duplex mode.
Occasional
system failures (in the form of
pcierror
panics) have been
seen when local area transport (LAT) functionality is enabled on AlphaServer
8200 and 8400 systems that have DE435 PCI/Ethernet network adapters that are
not connected to a network.
You can use any of the following methods to resolve
this problem:
Disable LAT functionality on the system by using the
latsetup
command.
Disable LAT functionality at the DE435 PCI/Ethernet network
adapter by using the
latcp
command with the
-E
flag.
Attach each DE435 PCI/Ethernet network adapter on the system to a network.
The following note applies to the use of Network File System (NFS) services in an ASE:
Version 1.5 of TruCluster Production Server Software and TruCluster
Available Server Software do not support NFS mounts over TCP.
If a client
has established an NFS mount to an ASE service over TCP (for example,
mount -t nfs -proto tcp), a server failover may cause the client
to hang.
You can use the
nfsconfig
utility on each cluster
member system to set the number of server TCP daemons to zero, which forces
the NFS service not to accept any NFS connection over TCP.
The following notes apply to the cluster MEMORY CHANNEL subsystem in TruCluster Production Server Software and TruCluster MEMORY CHANNEL Software configurations only:
In prior versions of the TruCluster
Production Server Software and the TruCluster MEMORY CHANNEL Software, MEMORY CHANNEL
controller numbers were determined solely by probe order (that is, the first
controller probed would become
mchan0, the second
mchan1, and so on).
Starting in Version 1.5 of these products, MEMORY CHANNEL
controller numbers are derived from the number listed in the system configuration
file (in
If you need the older behavior, edit the system configuration file so that MEMORY CHANNEL controller entries are positioned in probe order, and rebuild the kernel. For example:
controller mchan1 at pci0 slot 6 controller mchan0 at pci0 slot 7
Avoid disconnecting a MEMORY CHANNEL cable from a member system in a running cluster. Disconnecting the cable breaks the cluster interconnect, separating the system from the rest of the cluster. However, if a MEMORY CHANNEL cable is accidentally disconnected from a member system that uses redundant MEMORY CHANNEL interconnects, each member system (including the one from which the cable was disconnected) attempts to fail over to the secondary MEMORY CHANNEL interconnect to reestablish connection with the lost member system.
The failover process is initiated only if all member systems have a secondary MEMORY CHANNEL interface configured and available. When the member systems complete the failover, reconnect the MEMORY CHANNEL cable to the MEMORY CHANNEL adapter of the system from which it was disconnected.
A member system could display a panic message, similar to the following, if it experiences a MEMORY CHANNEL receive/transmit error that causes the single active MEMORY CHANNEL interconnect to fail:
rmerror_int: Error_count = 2 unit = 0
Err_reg = 0xffffffffa0000001 Node = 1
panic (cpu 0): rmerror_int: fatal error and no alternate mc to failover
When you see this panic message, ensure that the MEMORY CHANNEL cabling is fully connected and reboot the member system.
When a cluster node deliberately crashes another node, the crashing node displays a machine check message that includes output similar to the following:
Machine Check SYSTEM Fatal Abort
.
.
.
MC SHUTDOWN
MEMORY CHANNEL SoftC for board #0: interrupt count = 242
240 notifications
2 state changes
0 dropped acks
0 error Interrupts (13 processed)
NO ERRORS LOGGED
.
.
.
A full description of the machine check is displayed only if
the
mchan_debug
attribute is set to 1 in the
mchan
subsystem section of the
/etc/sysconfigtab
file.
The following new features of the MEMORY CHANNEL application programming interface library (API) are available in Version 1.5 of the TruCluster Production Server Software and TruCluster MEMORY CHANNEL Software products:
Multiple active rails
The MEMORY CHANNEL API now supports multiple active MEMORY CHANNEL logical
rails.
A user may now allocate, read from, or write to MEMORY CHANNEL regions on
up to four logical rails.
Two new API functions have been introduced to support
this functionality:
imc_ckerrcnt_mr(3)
and
imc_rderrcnt_mr(3).
Use
these functions instead of
imc_ckerrcnt(3)
and
imc_rderrcnt(3)
in new or recompiled code.
See the TruCluster Production Server Software
MEMORY CHANNEL Application Programming Interfaces
manual and the
appropriate reference pages for a detailed discussion of this functionality.
Point-to-point attach
Unless a different transmission mode is explicitly specified, all MEMORY CHANNEL transactions are broadcast. Starting in the Version 1.5 product, you can attach for transmit to a particular MEMORY CHANNEL region in point-to-point mode. As a result, all writes to the region are to the specified host only. This should generate increased bandwidth in future MEMORY CHANNEL implementations. Use this in situations in which processes on only one host must read from a particular region. See the TruCluster Production Server Software MEMORY CHANNEL Application Programming Interfaces manual for additional information on this feature.
Cluster change notification
The new
imc_wait_cluster_event(3)
function allows a user to watch for certain events in the MC API cluster,
such as nodes leaving or joining the cluster or new logical rails coming
on line.
See the TruCluster Production Server Software
MEMORY CHANNEL Application Programming Interfaces
manual and the
imc_wait_cluster_event(3)
reference page for further information on this feature.
Optimization for specifying addresses to which MEMORY CHANNEL receive regions are to be attached
The
IMC_USE_ADDR
flag to the
imc_asattach(3)
function allows you to specify the address at
which to attach MEMORY CHANNEL receive regions when that address represents a hole
in the current process's address space.
See the
imc_asattach(3)
reference
page for more information.
Faster spinlocks
The
imc_lkacquire(3)
and
imc_lkrelease(3)
functions have been
improved to provide a faster lock implementation.
New get cluster information function
The new
imc_getclusterinfo(3)
function allows a user to retrieve certain information about
a MEMORY CHANNEL API cluster: the number of members and their names, the number
of logical rails, and a bitmask of active logical rails.
This function supersedes
the
imc_gethosts(3)
function.
All new code should use the newer function
instead of
imc_gethosts(3).
See the TruCluster Production Server Software
MEMORY CHANNEL Application Programming Interfaces
manual and the
imc_getclusterinfo(3)
reference page for additional details.
Efficient copy function
The new
imc_bcopy(3)
function
allows the user to copy efficiently to MEMORY CHANNEL transmit regions.
See the
imc_bcopy(3)
reference page for additional information.
API initialization function
All programs should invoke the new
imc_api_init(3)
function
before calling any other MEMORY CHANNEL API library functions.
Existing programs
that do not call this function will continue to work.
See the TruCluster Production
Server Software
MEMORY CHANNEL Application Programming Interfaces
manual and the
imc_api_init(3)
reference
page for more information.
Monitoring the MEMORY CHANNEL API with the
imcs(1)
command
The new
imcs
command lets the user to see
details about MEMORY CHANNEL regions allocated through the MEMORY CHANNEL API, such
as their keys, size, and owner.
The command also displays information about MEMORY CHANNEL
spinlocks.
See the
imcs(1)
reference page for details.
Online examples
Every example published in the TruCluster
Production Server Software
MEMORY CHANNEL Application Programming Interfaces
manual is supplied on line
in the kit at
/usr/examples/cluster/mc_*.
One additional
example, of the use of the
imc_bcopy
function, is provided
at this location.
IMC_VERSION
constant
Starting in TruCluster MEMORY CHANNEL Software Version 1.5 and TruCluster
Production Server Software Version 1.5, the
imc.h
file
defines the
IMC_VERSION
symbol to the current version of
the MEMORY CHANNEL API library.
For Version 1.5, its value is 0x150.
When you use the
dxlsm
or
dtadvfs
utility in a TruCluster Production Server cluster or a TruCluster
Available Server configuration, the utility issues an informational message
on startup and upon any configuration change.
These informational messages
remind you to run the
asemgr
utility to update the ASE
configuration database to reflect changes in storage associated with ASE services.
These informational messages are optional; you can suppress them by specifying
the
-N
command-line flag when invoking the
dxlsm
or
dtadvfs
utility.
You can use the
ASEROUTING
configuration variable
only with the old
gated
(ogated).
(ogated
is the default selection in the
netsetup
script.) If you use the
ASEROUTING
configuration variable
when the new
gated
is running on ASE members, all service
operations will fail and error messages are entered in the
daemon.log
file.
This restriction existed, but was undocumented, in previous releases of TruCluster Production Server Software and TruCluster Available Server Software. In previous releases, service operations would not fail and no errors were logged.
You must ensure that the settings of the
/etc/sysconfigtab
attributes listed in
Table 1-8
are the same on
the new system as on all member systems.
| Kernel Attribute | Subsystem | TruCluster Software Configuration | Description |
dochecksum |
mcnet |
Production Server cluster only | Enables or disables TCP/IP checksums.
By
default, checksums are on (dochecksum=1).
Turning checksums
off results in a slight increase in performance of TCP/IP over MEMORY CHANNEL.
See the TruCluster Software Products
Administration
manual for more information. |
rx_mapping_enabled |
mcnet |
Production Server cluster only | Enables or disables copy avoidance.
By default,
copy avoidance is off (rx_mapping_enabled=0).
Turning copy
avoidance on results in a slight increase in performance of TCP/IP over MEMORY CHANNEL.
See the TruCluster Software Products
Administration
manual for more information. |
rm_rail_style |
rm |
Production Server cluster only | Configures the reliability style of the MEMORY CHANNEL interconects on a cluster member. See the TruCluster Software Products Administration manual and the TruCluster Production Server Software MEMORY CHANNEL Application Programming Interfaces manual for more information. |
enable_extended_uids |
proc |
Production Server cluster or Available Server configuration | Enables or disables extended UIDs in the base operating system. See the DIGITAL UNIX Release Notes for more information. |
Failure to match the current configuration can cause one or more systems to panic when you attempt to add a new system to the cluster. Thus, if you want to change any of these attributes in a running cluster, you must perform the following steps on each member system:
Edit the appropriate
/etc/sysconfigtab
attribute on each member.
See Appendix B of the TruCluster Software
Products
Administration
manual for instructions
for doing so.
Halt each member system.
Reboot all member systems.
If you must change these attributes and do not intend to perform a rolling upgrade from Version 1.4A to Version 1.5, consider changing them during a simultaneous upgrade procedure. See the TruCluster Software Products Software Installation manual for a description of the upgrade procedures.