|
Troubleshooting Guide
Finding error messages
Error messages are generally written to the MANAGER.log file in
the Windows NT (winnt) directory. This can be viewed using the 'view
MANAGER.log' option under the Program group (or Start menu) for
MANAGER/DIAMOND. The MANAGER.log file is deleted each time
MANAGER/DIAMOND is started.
Some startup error messages are written to the NT 'System' Event
Log.
Finally, there is a file called LOGFILE stored in the installation
directory - this file is never deleted, and lists all SCSI errors
noted by the MANAGER/DIAMOND software.
Troubleshooting Plasmon NT Products
MANAGER/DIAMOND cannot find SCSI devices
Initial checklist
Each device on the SCSI bus should be on a unique SCSI ID. The
SCSI card uses a SCSI id as well, usually SCSI id 7. Action:
ensure each device on the SCSI bus.
Ensure the SCSI bus is terminated only at the ends of the SCSI
bus.
MANAGER/DIAMOND does not find SCSI devices, but they
are listed in the NT SCSI control panel
Load the Windows NT Control panel, and select the 'Devices' section.
Ensure that all the device drivers for SCSI cards are configured
to load at BOOT time, and that the MGRSCSI driver is also configured
to load at BOOT time. Action: change the Control Panel settings
and reboot.
MANAGER/DIAMOND finds the SCSI devices but cannot start up
"Database check or license violation" message
If you are using a demo license, see if it has expired. These are
generally generated with a one-month duration. Action: obtain
new license key.
Check to see if you have more disks in a jukebox than your license
permits. If so, then obtain a larger license or remove disks from
the jukebox. Action: check license_data parameter in MANAGER.CFG
and, if necessary, obtain a new license key.
It is possible that a corrupted database file can cause this message.
To determine if this is the case, try renaming the JUKEBOX.DB file
to some other name, and restarting MANAGER/DIAMOND. If MANAGER/DIAMOND
then starts successfully it can be assumed that the database was
corrupted. Action: in this case, delete old database file, and
then rescan the database using the 'rescan all' option from JCP.
Failed to set drive count on device MGRPFS0. Error code = 2
Look in the NT Control panel under 'Devices'. If there are devices
which are of the form "X:" (where X is any capital
letter) then ensure that that they are configured for manual startup.
Action: change the Control panel settings and reboot.
Error 1450 in the MANAGER.log file
The system has been configured to use too much kernel memory. Action:
Reduce the size of num_buffers in MANAGER.CFG to a lower value,
and reboot.
"Can't find a jukebox" when using HP -st or
-fx range of optical disk jukeboxes
This jukebox needs to be configured with 'Dual Picker = No'. This
configuration option can be found on the front-panel menu of the
HP jukeboxes. Action: refer to HP configuration manual for advice
on setting this option
Startup problems
Can't find configuration file MANAGER.CFG
MANAGER/DIAMOND requires this file in order to start. Action:
manually recreate/restore MANAGER.CFG, or re-install MANAGER/DIAMOND,
thus creating a default MANAGER.CFG file
MANAGER/DIAMOND crashes at startup
Check the date on your Windows NT system. MANAGER/DIAMOND
will currently work correctly with any date prior to 18th
January 2038. Action: To determine if this is the case, use the
Date/Time applet in Control Panel (or double-click the clock on
the taskbar if you are using Windows NT 4.x) - Note - data is
stored on Plasmon and ISO-13346 format disks in a representation
that can handle all dates up to and including New Years Eve, in
the year 9999.
Compaq Insight MANAGER crashes when MANAGER/DIAMOND starts
Compaq Insight MANAGER is a service running on some Compaq systems.
The program gathers statistics about the use of resources on the
computer. One of these routines attempts to calculate the percentage
of free space on a drive letter. At the root of a jukebox drive
letter, MANAGER/DIAMOND typically reports 0 bytes free at root,
and 0 bytes total. Compaq Insight MANAGER attempts to calculate
the proportion of free space still available (free/total) and crashes
due to a divide by zero exception.
This parameter forces a lower bound for the 'total space' reported
on any MANAGER/DIAMOND filesystem. Typically setting 'total_blocks_min
= 1' is sufficient to prevent Compaq Insight Manger crashing.
Long startup delays
HP Jukeboxes take a long time to start
When MANAGER/DIAMOND starts, it issues a 'rezero unit' command
to any attached jukeboxes. HP jukeboxes will do a diagnostic self-test
when this command is received , and this can take several minutes.
If you wish to bypass this test, then use the rezero_jukebox
parameter in the MANAGER.CFG file:
This parameter can take the value 0 or 1.
0: Do not send a 'rezero unit' SCSI command to the jukebox at
startup.
1: (Default) Do send a 'rezero unit' SCSI command to the jukebox
at startup.
This should normally be left at the default value of 1, but if
there is a reason why MANAGER/DIAMOND is being stopped and started
(testing manager.cfg options for example), then it can be useful
to set it to 0. However, it is recommended that when in production
use, this parameter should be returned to the default.
Panasonic LFJ50 and LFJ100 jukeboxes take a long time to initialise
after JUKEBOX.DB is deleted
These units take up to 5 minutes to scan their slots and sense
for disks in the PD drives.
Files or directories disappear after a disc is re-scanned or
re-imported
File or directory was intended to be empty
If the file has zero length and has never had any content or attributes
changed then this is normal behaviour. MANAGER/DIAMOND
does not flush a file to the destination media until it has content
or changed attributes. Action: manually re-create the required
zero-length files.
Similarly, if a directory was created, but no files or directories
ever created in that directory, then the directory will disappear
after a disc is re-scanned or re-imported. Action: manually recreate
the required directories.
Cannot format a disk
Platter failed to format in stand-alone drive, with "error
22: Invalid Argument"
The volume name should be no longer than 32 characters. Action:
use a shorter volume label.
WORM disk fails to format, with error 81
The disk is already formatted in another WORM disk format. If a
WORM disk is initially formatted in one way, it cannot be reformatted
later in a different format. For example, it is not possible to
reformat a Corel WORM disk as a Plasmon disk.
Cannot write data to a disk
Check that MANAGER/DIAMOND supports writing data
in the chosen data format
The ISO-9660 and Corel DOS formats can only be read by MANAGER/DIAMOND.
Action: Format new disks in a format that MANAGER/DIAMOND
supports writing data in, for example the 'Plasmon' format.
Check that disk is not write protected
Write protected disks would be listed with the phrase [WriteProtect]
or [WriteFailed].
[WriteProtect] indicates that the disk has been physically protected
by engaging the write protect tab on the media cartridge, or that
the disk is inherently read-only. Action: export the disk, disengage
the write protect tab, then re-import the disk.
[WriteFailed] indicates that MANAGER/DIAMOND found
that the disk is not writeable due to a previous write fail. Action:
run the 'repair' utility against the disk and retry.
[ReadOnlyFileSystem] indicates that the file system only allows
data to be read - see above.
Check that there is enough free space on the disk
MANAGER/DIAMOND reserves the last 4MB of space on
disk so that it cannot be written - this is controlled by the reserved_kb_per_side
parameter. Action: check free space available on disk using JCP
utility
Problems with dates and times
File timestamps are incorrect by one hour
If it is summertime, then check the date settings on your Windows
NT system . Diamond's MANAGER/DIAMOND does not adjust correctly
for daylight savings time if the 'Automatically adjust clock for
daylight saving changes' mode is set. Action: To determine if
this is the case, use the Date/Time applet in Control Panel (or
double-click the clock on the taskbar if you are using Windows NT
4.x) Select the 'Time Zone' section, and ensure the 'Automatically
adjust clock for daylight saving changes' is not set: If
an existing file is altered on a jukebox or optical drive that MANAGER/DIAMOND
is controlling, the timestamp may not be updated immediately. The
correct timestamp is flushed to media, but there is an 'attribute
cache' in MANAGER/DIAMOND which is not immediately updated.
This is a limitation of the current shipping versions of MANAGER/DIAMOND
SCSI errors in LOGFILE
General advice
We advise the use of Adaptec SCSI adapters where possible. The
Adaptec 154x and 294x cards are particularly commonly used with
optical devices. We generally advise against using motherboard
SCSI controllers to control optical drives, and particularly optical
jukeboxes.
If one particular drive in a jukebox is giving a large number of
SCSI problems, it can be temporarily disabled using the 'drive_failed'
parameter in the MANAGER.CFG file:
It is possible for one drive in a jukebox to fail but for the rest
of the system to continue working normally. In this case, the system
administrator can inform MANAGER/DIAMOND that one drive has failed
by using this option. The parameter taken by the option is the drive
number, with 0 representing the first drive in the jukebox, 1 the
second drive etc. If more than one jukebox is attached to the system,
the numbering continues in sequence. Note that it is possible to
have more than one drive_failed line in the configuration file should
more than one drive have failed. The system will not function if
all drives have failed!
An introduction to SCSI error codes
In general, there are two classes of SCSI error messages:
1. Error messages generated by a SCSI target. With MANAGER/DIAMOND,
these are generally in the range 500 to 540. These are generally
caused by cabling and termination issues. Action: check
cabling and termination. The exception is error 528 which is
often a file system corruption problem, in this case action:
run repair on disk exhibiting this error code.
2. error messages generated by host adapter
cards or driver software. These are always in 3 parts. A 'key',
a 'code' and a qualifier.
The 'key' is a broad indication of the type of problem. This is
often enough to give an indication of the remedial action that needs
to be taken. The most common sense keys, and general actions to
undertake if they occur are as follows:
|
Sense Key
|
Action
|
|
0
|
No error
|
|
1
|
Recovered error - no action
|
|
2
|
Not ready.
|
|
3
|
Media error. Action: Check log over time to see if this
relates to a specific drive or disk. If it is a specific drive,
consider cleaning or routine maintenance on that drive. If
it is a disk, ensure that the data is backed up, and write
new data on some other disk. Consider visually inspecting
the media surface to look for physical contamination such
as dust or debris.
|
|
4
|
Hardware Error. If this is an optical drive, this may be
a drive error or a media error.
Action: consult the SCSI manual for the device in order
to determine the problem in more detail.
|
|
5
|
Illegal request. This is generally a software related issue.
This can be triggered by devices not supporting a specific
SCSI command, or by a request to read or write a sector past
the end of the disk surface.
|
|
6
|
Unit attention: This is usually sent by a device to notify
a device driver of a change - for example a new disk inserted
into a drive , or a magazine door opened on a jukebox. Action:
these can generally be ignored.
|
Seagate/Arcada BACKUPEXEC for Windows reports corrupt files.
A typical configuration exhibiting this problem would be:
+ Server running Diamond & a jukebox (particularly 1 drive
jbs)
+ Client running BackupExec + tape drive/jukebox
Symptoms are that BackupExec starts to report 'could not read file
- corrupt', and times out, when it progresses on to files on the
2nd disc. While this is happening the jukebox activity
light will indicate heavy activity, and there may be periodic disc
swaps.
The reason for this is that BackupExec issues a 'set attributes'
operation (attributes are flags like the archive bit, and the last
read time) after reading each file - even when it is set to run
in 'Copy Only Backup' mode. These write operations are queued up
in MANAGER/DIAMOND behind all the (data) reads, so when BackupExec
tries to read the files on the next disc MANAGER/DIAMOND still has
to flush all the attribute writes to disc before switching discs
- hence BackupExec reports timeouts and corruption (because it couldn't
read the file in time). This problem has so far only been
seen on single drive jukebox installations. Where more than 1 drive
is available reading can continue from a different drive while the
attributes are written in parallel - however the problem may still
occur eventually.
Action: On the Diamond machine, give the user account which
BackupExec logs in under READ-ONLY access to the jukebox share.
Do this in the Permissions dialog of the SHARING tab in the Jukebox
properties dialog - NOT in the Security tab.
Web GUI Troubleshooting
When I interact the GUI behaves strangely
Check that you have not logged in twice using different browsers:
this is not recommended, since it will not be clear which browser
is to be the destination for any interaction.
There is no java menu system after I log in
Check that your browser is configured to accept Java applets /
use Javascript
'Document contains no data' has appeared when using the web
gui
Check that you have waited long enough for a page to fully load:
this can take several seconds
I cannot log in
Check that the Plasmon security server is running in NT services.
Also, check that your browser has been configured to accept cookies
Empty, useless browser windows appear when I use IE4 with the
web gui
Remove the 'Browse in a new process' option from Advanced options,
if it has been set
|
