|
PROBLEMS ACCESSING THE HISTORIC EVENT LOG FILE IOMGR.LOG
In extremely rare circumstances, entries in the Historic Event Log
file "iomgr.log" may become unreadable, perhaps due to client and
system software revisions being out of sync. Should this happen, the
situation will be automatically detected and the existing file will
be renamed to "iomgrlog.old". Then, a new log file "iomgr.log" will
be created and subsequent events entered to that new file.
Furthermore, an event will be posted in the new log, stating "Server
Log database cannot be read. Database Reinitialized."
If only a specific record in the log is found to be in error, the
log will continue to report all valid records. The entry for that
specific log event, however, will say "No data available for this
entry."
USE OF CI/O WITH WINDOWS/NT SERVERS
A conflict has been found between ARCServeIT v6.6 Workgroup Edition
for Windows NT and CI/O Management Software v4.0x. When the
ARCServeIT Message Engine is started before CI/O's Portmapper
service, then the following message is observed:
"Unable to register(TEV_TEV_VER, tcp)"
At the time of release, a fix is still under investigation. As a
work-around, the system administrator should manually start the
"CIOArrayManager RPC Event" service first. This causes the
"NobleNet Portmappe" and "CIOArrayManager RPC Command" services to
start. Finally, the "ARCServIT Message Engine" service can be
started.
USE OF CI/O WITH NETWARE SERVERS
When installing CI/O Management Software on a NetWare server, it is
important to carefully follow the instructions provided in the CI/O
User's Guide.
If the remote client tends to disconnect from a NetWare 3.x or 4.x
server, the TIRPC components installed by CI/O on the server may have
been overwritten by an older version. This can occur when various
other applications are installed on the server. If so, reinstall
TIRPC from the CI/O distribution media.
If using an Ultra-2 SCSI host adapter with the FMS v3.01 (or later)
drivers for NetWare, be sure that the NWASPI.CDM driver is also
loaded. Otherwise, CI/O cannot manage the adapter and connected
devices. This file is automatically installed by CI/O during Setup.
If installation of both an Ultra-2 and non-Ultra-2 host adapter is
required on a NetWare 4.11 server, then adjustment of the driver
loading order is necessary. The proper order is:
load aspitran.dsk
load aic7870.dsk
load nwaspi.cdm
load aic78u2.ham
The following message may be observed from the NetWare console:
"IOMGR.NLM - 1 Process did not relinquish control frequently"
This message may be ignored. At the time of this release, Novell
is planning a NetWare 5.0 Support Pack update to avoid the message.
Support Pack fixes are already available for NetWare 4.11. As a
temporary workaround, insert the following as the first line in
your STARTUP.NCF file:
set display relinquish control alerts = off
Note that certain other applications may ensure that this parameter
is turned on. If so, turn it off by additionally inserting the
line as the last entry in your AUTOEXEC.NCF file.
Installing the Novell NetWare components of CI/O to a server with a
very large SYS volume (>2 Gbytes) may result in a low disk space error
message. This can be ignored.
USE OF CI/O WITH UNIXWARE SERVERS
UnixWare will accept changes in the configuration only after a
reboot. Hence, when multiple arrays are created, the system may
require multiple reboots to insure that all new arrays are
registered by the system and accessible to the user. Similarly,
if an array goes offline due to multiple drive failures, then
following reactivation, the system must be rebooted to bring the
array back on-line.
When using CI/O v4.0 with UnixWare v7.0, it is required that the
server boot ID be set to 0. Other values cannot be supported.
The following warning message may be ignored, when running the CI/O
start script on UnixWare v7.0.0:
dynamic linker: ./tcf_server: warning: copy relocation size
mismatch for symbol svc_fdtest
The message will not appear on UnixWare v7.0.1.
The "Make first Virtual Device" command is not supported for use with
UnixWare.
USE OF CI/O WITH ARRAY1000 CONTROLLERS
Previously-used drives should be freshly formatted or initialized
before being connected to the Array1000 controller. Otherwise, the
Array1000 controller may not properly reconcile the fragments left
over from previously created arrays.
The "Make first Virtual Device" command, used to modify the boot order
of arrays, will only operate upon the last array in the boot-order
list. It will change the identifier of the array with the highest-
numbered virtual device identifier to zero, thus making it the lowest-numbered identifier. The operation will be ineffective on arrays with
intermediate identifiers. By repetitive use of the command, however,
the boot sequence can be ordered as desired. This command is not
supported for use with UnixWare.
If all disk drives connected to the Array1000 controller's channels
are members of a single array, and that array goes off-line, then the
CI/O display will change. This occurs because the array driver unloads
from the system. The CI/O Management Software will no longer display
any array controller, channels, or array drives. If all members of an
array are missing, the array and its members are not displayed after a
reboot or Rescan operation.
In the rare circumstance that an array created on an Array1000
controller has multiple device failures and goes off-line, a critical
event will be posted to notify the user. In certain situations, a
second event may also be received, stating that the array is critical.
The second event should be ignored.
Reactivation of an off-line array may only be performed in certain
situations. Specifically, the array must have had at least one member
recognized by CI/O since the last system reboot. If the command fails,
CI/O will post an error message. In this case, a system reboot is
required before the array may be reactivated.
If a RAID 0/1 array experiences multiple device failures, and
reconstruct initiates from multiple defined spares, then an incorrect
event may be observed. The event will state that an array member is
down, but will reference the device ID of the spare.
Upon Reconstruct of a critical RAID 0/1 array using a Pool Spare, the
new member may be marked as "down" (red arrow through the icon) until
reconstruct completes.
USE OF CI/O WITH DMI AND SNMP
If attempts are unsuccessful to make the ciodmi service start
automatically, review the syntax of the "Depends On" entries in the
Windows NT Registry. Enter a carriage return after each dependent
entry.
When upgrading CI/O to a newer version, make sure to update the
corresponding DMI MIF or SNMP MIB to the same level. Otherwise,
displayed values may be incorrect. To update the MIF on Windows NT,
stop the CIODMI service, open a console window, change to the CI/O
installation directory (where CIODMI.EXE and CIO.MIF should be
located) and type:
ciodmi r
After the MIF is reloaded, type control-C to exit, and restart the
CIODMI service. To update the MIF on NetWare, instead of "load
ciodmi" type "load ciodmi r".
With early versions of DMI Service Layers and browsers used with
Novell NetWare, intermittent loss of event notification has been
observed when using the SPX/IPX protocol. The TCP/IP protocol has
not exhibited this behavior.
After unloading ciodmi for NetWare, a message may be observed
indicating that not all resources were released. This appears to
be a problem in the DMI Service Layer for Novell NetWare and is not
related to CI/O.
The DMI 2.0 SDK for NetWare has been found to install an older version
of TIRPC, potentially overwriting the version installed by CI/O. See
comments in section USE OF CI/O WITH NETWARE SERVERS
RESTORING COOPERATION BETWEEN CI/O AND INTEL LANDESK V2.5
Customers using certain versions of Intel LANDesk as their work group
management solution may experience problems when simultaneously using
Adaptec CI/O Management Software. The situation arises because some
versions of Intel LANDesk include an older version of the CI/O
software for use in monitoring SCSI adapters via the Desktop
Management Interface (DMI.) The LANDesk product loads the SCSI
monitoring software by default. CI/O Management Software, however,
only allows a single copy of CI/O to run on the system at one time.
The work around consists of disabling either the SCSI monitoring
software or the management software. The user must decide, therefore,
which capability is most important for their system configuration and
use. In general, if the system includes a disk array controller
product from Adaptec, we advise that the SCSI monitoring functionality
of LANDesk be disabled.
Software Installation Process:
- Install Windows NT and the updated Windows NT Server Pack, per
instructions for the Adaptec controller being used. If Windows NT
is to be installed upon a "bootable" disk array, the user should
refer to instructions on the use of Adaptec ArrayConfig software.
- Install CI/O Management Software, per the Adaptec instructions.
- After rebooting the system, go to the Services dialog box within
the Windows NT Control Panel. For each of the following, change
the entry from Automatic to Manual, and then stop each service:
CIO Array Management Service
CIOArrayManager RPC Command
CIOArrayManager RPC Event
- Install Intel LANDesk, per the appropriate instructions, then
reboot the system.
To Use Adaptec CI/O Management Software:
- Go to the Services dialog box in the Windows NT Control Panel.
- Stop the CIORAIDManagement Service.
- For each of the following, start each service:
CIO Array Management Service
CIOArrayManager RPC Command
CIOArrayManager RPC Event
To Use the Native SCSI Monitoring Feature within Intel LANDesk:
- Go to the Services dialog box in the Windows NT Control Panel.
For each of the following, change the entry from Automatic to
Manual, and then stop each service:
CIO Array Management Service
CIOArrayManager RPC Command
CIOArrayManager RPC Event
- Start the SCSIDMI service and CIORAIDManagement Service.
|