| Previous | Contents | Index |
H.1 Deinstall Instructions
To deinstall RDF on each ULTRIX CLIENT node, follow the steps listed
below and type the commands shown. The ULTRIX system is space conscious
and case sensitive, so be sure to type the commands exactly as
shown.
# cd /tti_kit/tti_rtf
# ls
rtallocate rtf.final rtf.o rtshow
rtdeallocate rtf.install rtf.control rtf.deinstall
config.log kernel.build.log rtf.build.date
|
# rtf.deinstall
|
MACHINE file name [xxx]:
.
.
|
# rtf.final
.
Continue (yes/no)? YES
.
.
|
I.1 Command Overview
The setld command is used for installing and managing software
on ULTRIX systems. The setld command uses two procedures:
# RTALLOCATE node_name::device tape_unit [logical_name] [norewind]
# setld function_key tape_unit logical_name
|
Where node_name is the name of the node that the tape device is locally attached to (the SERVER node).
Where device is the name of the tape device on the server node.
Where tape_unit is the name of the magnetic tape on a particular tape unit.
Where logical_name (an optional parameter) is the name to be used when accessing the remote tape device. If no logical_name is given, a logical name in the form TAPEnn is created.
Where norewind is a required parameter to be used when using the setld command.
Where function_key is a specific function to be performed by the setld command.
I.2 Loading Software Using The setld Command
To load software using the setld command, enter the following
from the ULTRIX CLIENT node:
% rtallocate zeus::msa0 nrmt9h norewind
% setld -l rmt9h ....
|
Where -l is the function key to load software from distribution mounted on location.
Where rmt9h is the magnetic tape on unit 9.
J.1 Can a Node Be BOTH an RDF Client and an RDF Server?
Yes, all RDF server nodes are licensed to be BOTH a server and a
client. When the RDSERVER_STARTUP.COM file is executed on the server
node, the client is started up automatically as well.
J.2 Can RDF Be Used With "Stand-alone" Backup?
Sorry, but, No. The Remote Device Facility requires DECNET to be
running. During stand-alone backup, DECNET is not available on the
client node.
J.3 Does the Order of Startup Matter?
No. You can start up the CLIENT and SERVER nodes in any order. You can
also shutdown the server and client at any time---and in any order.
J.4 Can RDF be Used With OpenVMS INSTALL?
Yes. A previously RDALLOCATED device can be used as the input device
for OpenVMS INSTALL. However, if the installation procedure shuts down
DECNET, RDF cannot be used. This is because RDF requires DECNET to be
running on the client node.
J.5 At What Level is RDF Operating?
The Remote Device Facility operates as a pseudo device. The pseudo
device on the CLIENT is IDENTICAL in behavior to the real device on the
SERVER down to the QIO level! This means that a RDALLOCATED device can
be used ANYWHERE that a real device is used.
J.6 Does RDF Allow Access to My Optical Drive?
The Remote Device Facility works with ANY device that OpenVMS sees as a
disk or tape device. To see whether or not OpenVMS thinks that a given
device is a disk or tape device, enter the following DCL command:
$ show device/full DEVNAME
|
Where DEVNAME is the name of the device in question.
If the device is a tape device, you see something similar to the following:
Magtape ZEUS$MUB0:, device type TU81, is online, record-oriented device, file-
oriented device, error logging is enabled.
Error count 50 Operations completed 37645
Owner process "" Owner UIC [0,0]
Owner process ID 00000000 Dev Prot S:RWED,O:RWED,G:RWED,W:RWED
Reference count 0 Default buffer size 8192
Density 1600 Format Normal-11
Volume status: no-unload on dismount, odd parity.
|
MUB0: is preceded by the word Magtape. This means that device MUB0: can be accessed using RDF.
If the device is a disk device, you see something similar to the following:
Disk $1$DIA0: (RF), device type RF30, is online, mounted, file-oriented device, shareable, served to cluster via MSCP Server, error logging is enabled. Error count 0 Operations completed 1552 Owner process "" Owner UIC [1,1] Owner process ID 00000000 Dev Prot S:RWED,O:RWED,G:RWED,W:RWED Reference count 1 Default buffer size 512 Total blocks 293040 Sectors per track 37 . . |
$1$DIA0: is preceded by the word Disk. This means that device $1$DIA0: can be accessed using RDF.
J.7 Why is RDF Slow When I RDALLOCATE a Local Device?
I started the SERVER and the CLIENT processes on the same node and then RDALLOCATED a device on this system. The performance was terrible. Why?
There are several reasons. One is that we must use smaller buffers when "talking over" the local data link, so we have to do more work to move the same amount of data. Also, when moving data over a real data link such as Ethernet, DECnet's pipelining capability just about eliminates the data link transfer "wait time" from the throughput equation, since this time is available to the sending CPU and we use this time to queue up subsequent buffers for transmission. But with a single CPU both sending and receiving the data, all benefits from pipelining are lost!
Furthermore, data sent through DECnet to the local node traverses nearly all layers of the DECnet software---excepting only the data link driver, which is minimal---and with the same CPU at both ends of the link, that CPU must now do all the DECnet work of both sending and receiving the data. Whereas with a real data link, this work is shared by the sending and receiving CPUs. The CPU is therefore working nearly twice as hard when the link is to the local node!
J.8 I'm using TAPE01: and Get a Device Error. Why?
When a device is RDALLOCATED a logical is created for that device (e.g.
TAPE01). If the RDALLOCATE was unsuccessful, the logical is not
created. To see whether or not a device was successfully RDALLOCATED,
use the RDSHOW CLIENT command.
$ @tti_rdev:RDSHOW CLIENT RDALLOCATED devices for pid 20200294, user DJ, on node MINI:: Local logical Rmt node Remote device TAPE01 ZEUS:: ZEUS$MUC0 |
A successful RDALLOCATE can be detected by checking the symbol RDEV_FINAL_STATUS.
The following DCL code segment illustrates the use of RDEV_FINAL_STATUS.
.
.
.
$ rdallocate:
$ @tti_rdev:rdallocate zeus::mub0: mytape
$ if .not. rdev_final_status then goto try_again
$ backup/image/ignore=label dua0: mytape:sysdsk.bck
.
.
.
$ exit
$
$ try_again:
$ wait 00:05:00 ! wait 5 minutes
$ goto rdallocate
|
OPCOM messages are not being relayed from the CLIENT node to the SERVER. The RDCLIENT log file says the OPCOM relay has been disabled. The OpenVMS status was no such device. Why?
If OPCOM messages are not being relayed to the SERVER node, you should check the RDCLIENT_nodename.log file. The file is located in the RDF directory on your CLIENT node.
In many cases, the log file looks something like:
RDF - Remote Device Facility (Version 4.n) - --rdclient-- Procedure
Copyright (c) 1990-2000 Touch Technologies, Inc.
[30-JAN-2000 19:25:06.06] RDCLIENT startup (V4.n)
[30-JAN-2000 19:25:06.28] OPCOM relay disabled (creating PTY)
[30-JAN-2000 19:25:06.28] SYSTEM-W-STATUS-NOSUCDEV,no such
no such device available)
[30-JAN-2000 19:25:06.28] Startup complete ()
|
This means that the FT (pseudo-terminal) device isn't present in the system.
This can be caused by running under a pre 5.5 release of OpenVMS. Or the FT driver may have been moved from the system for some reason.
The FTDRIVER is normally started in SYS$STARTUP:VMS$DEVICE_STARTUP.COM:
$sysgen connect fta0/noadapter/drivername=ftdriver
|
YES. When you upgrade Ultrix you get a new KERNEL image. Because RDF is linked into the kernel, you must reinstall RDF whenever you have a new Ultrix kernel.
J.12 What does RDF check in determining whether or not to return the RTF_CLIENTSBUSY error message?
I am no longer able to RDallocate a remote device from the CLIENT node.
RDF returns the following error message:
'Error RDallocating node_name::device_name:---RDF_CLIENTSBUSY'
'%SYSTEM-F-ABORT, abort'
Where is this coming from, and what does it mean?
The RDF_CLIENTSBUSY message is returned from an unsuccessful RDallocate when RDF has determined that all of 16 pseudo devices (ie. RDEVA0, RDEVB0, etc.) are busy.
J.13 The server has a new tape device, but I can't RDallocate it. Have I forgotten something?
YES. Whenever a device has been added or removed to
the server node, the configuration file,
config_nodename.dat must be modified. The easiest way
to do this is to shut down the RDserver process on the server, delete
the configuration file, and then restart the RDserver. Otherwise you
can edit the configuration file and add or remove the proper device
line.
J.14 Why can't a disk be mounted anywhere before RDALLOCATE/WRITE/EXCLUSIVE is allowed?
OpenVMS controls concurrent access to files and the file system using
the Distributed Lock Manager. The Distributed Lock Manager does
not operate over a network connection.
Without the Distributed Lock Manager watching over file system synchronization, shared write access to a disk is dangerous.
RDF, therefore does not allow the shared write access to a disk unless only a single node is accessing the disk.
J.15 Does RDF handle the turn of the century?
The RDF dates are all eight (8) digit dates, so the turn of the century
will not have an impact.
| CLIDENY | Access from this CLIENT to the SERVER is not allowed. Check for "CLIENT/ALLOW" in the RDserver's configuration file. |
| CLIENTSBUSY | All 16 pesudo-devices are already in use. |
| DEVDENY | Client is not allowed to the Device or to the Node. This error message is dependent on the "CLIENT/ALLOW", "/ALLOW" or "CLIENT/DENY", "/DENY" qualifiers in the configuration file. Verify that the configuration file qualifier is used appropriately. |
| EMPTYCFG | The RDserver's configuration file has no valid devices or they are all commented out. |
| LINKABORT | The connection to the device was aborted. For some reason the connection was interrupted and the remote device could not be found. Check the configuration file as well as the remote device. |
| NOCLIENT | The RDdriver was not loaded. Most commonly the RDCLIENT_STARTUP.COM file was not executed for this node. |
| NOREMOTE | This is a RDF status message. The remote device could not be found. Verify the configuration file as well as the status of the remote device. |
| SERVERTMO | The RDserver did not respond to the request. Most commonly the RDSERVER_STARTUP.COM file was not executed on the server node. Or, the server has too many connections already to reply in time to your request. |
| Index | Contents |