The installation program for Windows NT copies files to a directory on your hard drive, installs iXOS-JUKEMAN services and sets up registry entries.
Before you begin installation, log on as administrator or a super user.
Start by setting up an installation directory, for example \ixos
. Copy the jukeman.exe
file into the directory and extract it by typing:
C:\ixos>jukeman.exe
The Jukeman installation program starts. In the setup dialog, specify where you want to install the iXOS-JUKEMAN user
programs. The default is c:\jukeman
.
Then select the portions of the software you want to install:
cdglow
recording software.
You will probably want to install all these items.
Click OK to start installation. The installer creates an iXOS-JUKEMAN folder containing Jukeman Administrator and Jukeman CD-Writer icons. Icons from previous versions of iXOS-JUKEMAN are not removed automatically.
When installation is complete, reboot your system.
Be sure all jukeboxes and drives are turned on. Create an empty directory for installation, using any new path. You need about 2 MB of disk space for installation.
Copy the iXOS-JUKEMAN software into the installation directory using the iXOS-JUKEMAN CD or distribution files from our web site, www.jukeman.com.
If you have an iXOS-JUKEMAN CD, then you can find the software packages for certain UNIX platforms and example configurations in a subdirectory called 'unix' or 'os'.
If you download the software via WWW, then please refer to the information on the Web pages. Important files, which should be downloaded, are the example configurations ('examples.tar.Z') and the compressed tar file for the appropriate UNIX platform.
Copy the compressed tar files into the installation directory. (For example 'solaris.tz' or 'os_solaris.tar.Z')
Extract the file corresponding to your operating system. For example, if your operating system is Solaris, enter:
# zcat < solaris.tz | tar xvf -
Now you have all the files except license files, device descriptions, and configuration. If you are upgrading an old installation, copy the
old configuration files and the volumes
database into the new installation directory. For first installation it
is recommended to extract template configuration files. If necessary, you can obtain such templates (except
volumes
) by entering:
# zcat < examples.tz | tar xvf -
The volumes
file is generated automatically by cdnfsd
. Delete the .tar
or .tz files by typing:
# rm *.tar.Z *.tz
Next, install the SCSI driver and create standard iXOS device names by typing:
# jmsetup
Since the cdnfsd
program runs only under root id
and the performance of
cdglow
is much better if it runs under root id
, change the owner to
root
and the permissions to suid
:
# chown root cdnfsd cdglow
# chmod u+s cdnfsd cdglow
For HP-UX 10.* change the permissions for the cdadm
command:
# chown root cdadm
# chmod u+s cdadm
This section explains how SCSI devices are represented by the iXOS-JUKEMAN generic SCSI driver.
A computer can have several SCSI controllers for multiple SCSI buses. Each SCSI bus has 8 IDs, named 0-7. Normally, each device on a SCSI bus requires a SCSI ID. Jukeboxes often need one ID per drive and one for the changer. Usually, the SCSI controller occupies ID 7 and 0-6 are free for devices. SGI computers occupy ID 0, which means you can connect devices with IDs 1-7.
WARNING: When you connect devices to the bus, be sure the new devices use IDs that are not occupied by existing devices or the controller. If you use an occupied ID, you can damage your hardware.
Each SCSI ID can be split into 8 Logical Units (LUNs). The advantage of LUNs is that a jukebox uses only one SCSI ID, so you can attach more devices to a single bus. Addressing LUNs is explained later in this manual.
In UNIX and NT, each device is represented by a device name, which is a path in the root file system tree.
NT hides the devices, but iXOS-JUKEMAN makes the names visible. For example, SCSI ID x on bus y of controller z is
represented by the path \\.\pzbytx
. So if you have a PCI bus and put Adaptec twin adapters AHA 3940 into
bus slots 0 to 2, you have 6 SCSI buses, and ID 4 on bus 1 of adapter 2 is represented as \\.\p2b1t4
.
Each enumeration begins with 0.
Within an ID, LUNs are represented by appending a comma and the LUN. So \\.p2b1t4,1
is LUN 1 of
SCSI ID 4; \\.\p2b10t4
and \\.\p2b10t4,0
are the same.
If you know which paths represent your devices, you can use the inquiry
command to verify the paths. For
example, if you have a CDR 100 Recordable from NSM that is connected to the only SCSI bus on an NT system using SCSI ID 5,
the command
inquiry \\.\p0b0t5
will return something like
0000002 \\.\p0b0t5 is HP's CD-drive "C4324/C4325"
where 0000002 is the time in 1/100 seconds. The message
0000000 Can't open \\.\p0b0t5
occurs if the SCSI driver doesn't run, the path is wrong, the device doesn't work, or the cable is bad. It is also possible that the device may be unknown if it did not run when you booted the system.
If you don't know the paths that represent your devices, call
scsidevs
which returns a complete list of all known and working SCSI devices.
If you add a new controller, you can use these commands to test if the path names have changed.
In UNIX, all devices are located in a subdirectory of /dev
. iXOS-JUKEMAN creates directories
such as /dev/iXOS_SCSI0
, /dev/iXOS_SCSI1
, and so on.
Each directory represents a SCSI bus. Devices are represented by numbers. So if your bus is represented by
/dev/iXOS_SCSI1
, the device using SCSI ID 4 is represented as
/dev/iXOS_SCSI1/4
.
If you use LUNs, add a comma and the LUN number. dev/iXOS_SCSI1/4
,1 is LUN 1 of the
device using SCSI ID 4 on the bus represented by /dev/iXOS_SCSI1
.
On Solaris, if you add devices, either reinstall the SCSI drive with jmsetup
or /etc/halt
the operating system and reboot with boot -r
. You may need to reconfigure
vold
so it does not try to control the CD drives in your jukeboxes. See man vold
and
man vold.conf.
On IRIX, free SCSI IDs are 1-7, not 0-6 as on other computers, because the controller occupies ID 0.
AIX uses a multiplexed driver. You cannot see the devices, but you can use them. The driver generates the device paths automatically upon access.
If installation was successful, the writer is set up and ready for use. See Chapter 5, "The Writer", for information.
If you have a license key for the writer, insert the key in the file named writer.lic
. For example,
version=2
writer
license=abcdefgh
Without a license key, the writer runs in demo mode, which means you can send only 128 MB to a recorder drive even if you use
cdglow
or the writable file system.
Server setup requires that you decide how the file system should be represented and which devices should be used.
The server.lic
file contains the number of supported physical CDs and a license key. For example,
version=2
volumes=500
license=ijklmnop
If this file is missing or the key is invalid, the server runs in demo mode, which means it supports only a few disks (usually five) and runs for a maximum of two hours. When the time is up, you can restart the server.
For additional information, see "License Policy and Demo Versions' in Chapter 1.
The server maintains a single file system containing all CDs as subdirectories, but different clients may require different views of this file system. For example, PCs prefer the 8.3 file name format with no mixed cases, while some UNIX clients want to see long file names stored with the optional Rock Ridge extensions. In other cases, some clients may need to see all CDs, while others need only a small subset.
iXOS-JUKEMAN contains two standard views that are adequate for most applications. If necessary, you can customize the views (see "Exported File System Views" in Chapter 3). The following instructions tell you how to run the default view configuration.
The NT version presents drive x: , containing all CDs in a format for PC clients. It simultaneously exports two additional file system views in PC (z: ) and RR format (y: ) to NFS clients. PCs access x: as a network drive and NFS clients can mount either
jukeman_hostname:/views_pc
or
jukeman_hostname:/views_rr
for the PC or RR view.
Example of a UNIX mount command:
# mount -o timeo=99,retrans=14,soft jm_hostname:/views_pc /cds In directory '/cds' are visible then the CDs controlled by Jukeman.
You can configure additional file system views using the Jukeman Administrator. (After you change the device or file system configuration, press F5 from the File Manager to refresh the screen.)
If there is no NFS running, your file system installation is complete. But if NFS is running, you must set up the standard
rpc.mountd
to cooperate with the server.
rpc.mountd
monitors the file systems. Clients that want to mount a Network File System ask
rpc.mountd
for a root file handle, which is the key for later requests to either the standard
nfsd
or the jukebox NFS server. Without this key, you cannot use the Network File System.
There cannot be a root file handle without a file system root. So you must create an empty directory for each standard view. These
views are normally /view/rr
for RR format and /views/pc
for PC format, though other
file names are possible as well (although other names do not comply with the iXOS standard server.cfg
file). Create the directories by typing:
# mkdir /views_rr
# mkdir /views_pc
Next export a file system by telling rpc.mountd
to give the file handles to clients. Most flavors of UNIX maintain a file named
/etc/exports
containing all exported file systems. If you add the line /views
to this
file, rpc.mountd
exports all views of the iXOS-JUKEMAN file system.
Under Solaris, exported file systems are listed in the file /etc/dfs/dfstab
. Add the lines
share -F nfs /views_pc
share -F nfs /views_rr
On some versions of UNIX (Solaris and others), the change does not take effect immediately. On Solaris, call the commands
unshareall
and shareall
to update the export list. Other UNIX systems offer the exportfs
command,
so you can use exportfs -a
to export all file systems. Another method is to reboot the server.
You can easily test if the directories are exported by using showmount -e
. When you start the server, it
tests whether the directories for all views are exported by requesting the file handle from rpc.mountd
.
It stops if it cannot get a file handle.
Once the directories are exported, start the server. In the installation directory call cdnfsd
. The server
runs and exports file systems. However, they are empty because you have not yet defined devices, and you cannot see them because
they are Network File Systems and you must mount them before you can see anything. The CDs never appear in
/views_rr
or /views_pc
; these directories only satisfy the rpc.mountd
requirements. You must mount them as any NFS client does.
Each device that the server controls requires a file that describes the device type, the SCSI addresses of the drives, and the SCSI address or the RS232-address of the changer (also addressed as robot in the following examples). The file can also describe which jukebox storage slots the server should use (the default is all), and which file, if any, contains information about which CDs occupy which slots of the jukebox.
Once you create a device description file named device
, the command
cdadm attach device
attaches the device to the server. The server inspects the slots and presents the CDs in all file system views. This inspection is normally only performed if you attach the device for the first time or change CDs manually without informing iXOS-JUKEMAN. Now clients should be able to see the CDs.
For additional information, see the description of save files in "Device Descriptions" later in this chapter. For details on
cdadm
, see "Controlling the Server" in Chapter 3.
Device descriptions are discussed generally here. Chapter 4, "Supported Jukeboxes," contains individual descriptions for each type of device.
For several jukebox types, the changer is controlled through a serial line, which allows you to save a SCSI ID. For the NSM Mercury jukebox the serial interface allows the software to fully exploit the features of the jukebox and its ability to execute several movements in different states simultaneously. iXOS-JUKEMAN uses this parallel capability; even under high load the server can satisfy 14 client requests for different CDs in a single Mercury per minute. Moreover, you can connect 16 NSM changers to a single serial line (even mixed jukeboxes for 100 CDs and 150 CDs) and all can move simultaneously.
Each serial line is represented in the file system by a name that depends on the operating system. The first two serial lines are:
NT | com1:
| com2:
|
SunOS | /dev/ttya
| /dev/ttyb
|
Solaris | /dev/ttya
| /dev/ttyb
|
HP-UX 10.* | /dev/tty00 or /dev/tty0p0
| /dev/tty01 or /dev/tty0p0
|
AIX | /dev/tty0
| /dev/tty1
|
DEC UNIX | /dev/tty00
| /dev/tty01
|
IRIX | /dev/ttyd1
| /dev/ttyd2
|
For IRIX, also see the output of "man serial."
For AIX and SGI (IRIX), the special serial
cables that are required are available from IBM.
A device description file contains lines in the format key=value
. Some keys must be specified for
all devices, others depend on the type of device. The most important parameter is the device type:
device=type
Supported types include: grundig200
, cdr100
, kubik
,
nakamichi
, nec
, mercury
, pioneer6
,
pioneer18
, pioneer500
, sony_cdr
, sony_cdl,
single
(for a single SCSI CD drive), tower6
, tower7
,
image
(for an ISO image on a hard disk), and standard
, which you can use for all standard
SCSI 2 jukeboxes. Contact our web site ("Writable File System" in Chapter 3).
drive=.
.
For example, in NT if you want the server to control a single
drive that the generic SCSI driver
presents as \\.\p0b0t4
, the description file should look like:
device=single
drive=\\.\p0b0t4
Under UNIX a similar description is:
device=single
drive=/dev/iXOS_SCSI0/4
Similarly, the NT description for a pioneer6
changer should resemble:
device=pioneer6
drive=\\.\p0b0t4
and for UNIX,
device=pioneer6
drive=/dev/iXOS_SCSI0/4
This causes the server to export all disks.
You can specify a few disks by adding a line such as disks=1-3
or disks=1,2,4-
6
. When the server attaches the device, it inspects only the specified disks. This is useful during installation when you start
and stop the server often.
If you don't want the server to inspect the disks upon each startup, use the parameter save=savefile
to
specify a save file in which the server stores information about which CDs reside in which slots of the jukebox. The server uses the
save file if present and creates it if it is not present.
In NT the save file might look like this:
device=pioneer6
drive=\\.\p0b0t4
save=p6.sav
In UNIX the save file would resemble:
device=pioneer6
drive=/dev/iXOS_SCSI0/4
save=p6.sav
When you name a save file, instead of appending .dev
, the usual suffix of a device description
file, append the suffix .sav
. This is what the server assumes if you specify the save file as
save=*.sav
. So if your device description is p6.dev
, the save file should be
p6.sav
.
Note that pioneer6
, like single
, does not need a robot specified.
single
drives have no robot, and pioneer6
addresses it through the drive.
All other device types except nec
, tower6/7
, and image
must specify the robot: robot=robot.
For most devices this is another SCSI target. For
sony
and pioneer18
, this is a LUN of the drive target. An NT description file for a
pioneer18
using slots 1-9 must look like:
device=pioneer18
drive=\\.\p0b0t3
robot=\\.\p0b0t3,1
disks=1-9
The UNIX description file for a pioneer18
using slots 1-9 would look like:
device=pioneer18
drive=/dev/iXOS_SCSI0/3
robot=/dev/iXOS_SCSI0/3,1
disks=1-9
For the kubik
, the robot uses a serial interface. The NT description looks like:
device=kubik
drive=\\.\p0b0t1
drive=\\.\p0b0t2
drive=\\.\p0b0t3
drive=\\.\p0b0t4
robot=com2:
The UNIX description looks like:
device=kubik
drive=/dev/iXOS_SCSI0/1
drive=/dev/iXOS_SCSI0/2
drive=/dev/iXOS_SCSI0/3
drive=/dev/iXOS_SCSI0/4
robot=com2:
For NSM CDR 100 and Mercury jukeboxes a serial line controls up to 16 jukeboxes. A new line, robid=id
,
specifies the ID of the robot on the serial line. Each jukebox has an ID from 0 to 15, which can be checked and set by the Mercury's
menu, or directly in the CDR 100 (see the NSM manual). An NT description for a Mercury 20 with robot ID 7 should look
something like:
device=mercury
drive=\\.\p0b0t0
drive=\\.\p0b0t1
robot=com2:
robid=7
The UNIX description for a Mercury 20 with robot ID 7 should look something like:
device=mercury
drive=/dev/iXOS_SCSI1/0
drive=/dev/iXOS_SCSI1/1
robot=/dev/ttya
robid=7
Alternatively, the following syntax can be used to define the serial interface and robot ID serial controlled jukeboxes.
robot=/dev/ttya,7
For more examples of device descriptions, see Chapter 4.
Note: The drives in the device description file must be in the same order as the drive numbers and not in the order of the SCSI IDs. This is valid for all jukeboxes with several SCSI IDs for the drives. On the back of newer Mercury 31/40(20) jukeboxes there are four (two) switches to set the SCSI IDs. Older Mercury 31/40 models use IDs 3 to 6 corresponding to drives 1 to 4; Mercury 20 models use IDs 3 and 4 for the two drives. The first drive in the device description file should be drive 1 of the jukebox, and so on.
You have installed the drivers, defined views, started the server, defined devices, and attached them to the server. Next you need to know how to see the CDs on your client computers.
The NT version of iXOS-JUKEMAN fits many different environments, including those that are mixed. It supports the NFS protocol, making it a true NFS server. UNIX clients simply mount it as they mount any Network File System. Simultaneously, it presents all CDs in a native file system for NT, which can be exported through all available protocols.
After you call the manual startup script cdstart.bat
, a local drive x: presents the
standard PC view of the server's file system. You will find the CDs of all attached devices in x: as subdirectories. You can access the
CDs locally through the File Manager; it is shared as "ixos" as the default. PCs connect to it in the same way they connect to any other
network drive.
The volume names burned on the CDs can have up to 32 characters. Sometimes PCs are confused by these long directory
names. However, you can rename the CDs using the cdadm
command. For information, see
"Controlling the Server" in Chapter 3.
Some CD software assumes that the CD's root is the root of the drive; they do not expect a CD name to be an additional path component. This may cause an installation message noting that a file is missing. If this happens, use the File Manager to share the corresponding subdirectory so the client PC can connect this CD as a separate network drive.
To execute files directly from the server's file system, you have to use it as a shared network drive even if you are working on the server computer.
Both the UNIX and NT versions support the NFS protocol. NFS clients can mount the server's file system as they mount any Network File System, but they need to add some mount parameters. For example,
mount -o port=4027,timeo=33,retrans=14,soft
hostname:/views/rr /cd
mounts the server's file system on the client's empty directory /cd
. Once this is done, all CDs appear to be
subdirectories of /cd
. The simple command ls-1/cd
shows a list of all available CDs.
Depending on the operating system, some versions of mount
require additional parameters such as
mount -Fnfs -o port=..
or mount -f NFS,port=..
depending on the
operating system. See "man mount"
for details.
The port=4027
option tells mount
that the NFS server uses port 4027
instead of 2049, which is used by the standard NFS daemon. This enables the server to coexist with the standard
nfsd
so clients can use both magnetic disks and jukeboxes on the server computer concurrently. NT does not
include a standard nfsd
; consequently, the NT version uses the standard port, and you do not need to specify the
port number.
The NFS 'mount' command on Solaris 2.5 or DEC UNIX systems has to have a additional option 'vers=2', because without that parameter NFS version 3 would be used. IXOS-JUKEMAN is compatible only with NFS version 2.
Example NFS mount on Solaris 2.5 with iXOS-JUKEMAN installation on WindowsNT:
# mount -o timeo=99,retrans=14,soft,vers=2
jm_host:/views_rr /cds
Example NFS mount on Solaris 2.5 with iXOS-JUKEMAN installation on SGI IRIX:
# mount -o port=4027,retrans=14,timeo=99,soft,vers=2
jm_host:/views_rr /cds
To understand the other options, you need to understand NFS clients. A user level application accesses a mounted Network File System as if it were any local magnetic disk. The kernel of the client computer automatically generates NFS requests and waits for the answers, which in turn are used to satisfy the accesses requested by the application. But networks may drop a request or an answer. Therefore, the NFS client built into the client's kernel not only generates NFS requests, but also retransmits them if it does not receive a reply within a reasonable time.
The timeo=99
option instructs the kernel's NFS clients to retransmit a request if there is no reply after 99
tenths of a second (9.9 seconds). These retransmits are not visible to users, except for messages such as "NFS server not responding,
still trying."
Short timeouts increase the network load because each disk move can cause several useless retransmits. Long timeouts are bad if a packet is dropped by an unreliable network and a user must wait until the kernel's NFS client retransmits the request. After each retransmit, the timeout value is doubled, up to one minute maximum.
The retrans=14
option instructs the kernel's NFS client to automatically retransmit a request 14 times before it
gives up and the file system access that caused the NFS request fails. It makes sense to specify a high value because if several
clients access different CDs located in the same jukebox, the server must move these CDs, and the last client must wait a long time.
You can avoid long wait times if you have enough jukeboxes and enable the server to distribute the load by duplicating the disks and
spreading them over the jukeboxes. This enables you to build failure-tolerant archives with predictable short response times.
The soft
option instructs the kernel's NFS client to give up after all retransmits. You can also specify
intr
or hard
. If you specify intr
, it may interrupt the system call
that caused the NFS requests with a signal; otherwise, the only way to finish the system call is a server response.
If the mount
command generates a "no such file or directory" message, make sure the /cd
directory on the client side and the /views_rr
directory on the server side exist.
If you receive a "permission denied" message, just export /views_rr
on the server side.
If your client computer uses a PC operating system such as DOS or NT, you can install an NFS client on the PC or install a PC file server such as samba on the server computer. You can obtain samba from the iXOS ftp server, ftp.ixos.de, or from nimbus.anu.edu.au, under pub/tridge/samba. The latest version of samba is included with iXOS-JUKEMAN. You can find more information about this package at the samba web site, http://lake.canberra.edu.au/pub/samba.