Starting TCPIP (UCX) and SYSLOGD
Introduction
These notes are mainly aimed at computing staff
who
may need to restart TCPIP communications
or modify where an Alpha writes it's syslog information. Since TCPIP
supports
mounting of NFS disks restarting TCPIP can fix NFS mount problems
that
can occur
after NFS server machine maintenance. One of the TCPIP services is
SYSLOG
which underlies the observing system talker logs.
The TCS writes messages to SYSLOG and we configure these to be
written into
the local file:
[SYS0.UCX_SYSLOGD]SYSLOGD.LOG
and also the talker log for the local ICS. Currently such logs are
maintained for intics,
taurus. The spare/development machine
should be
configured to write to all log machines, but this may need to be
reconfigured when
it is used to control WHT or INT for a long period.
On each alpha two targets for logging are
defined: SYSLOGHOST1 and SYSLOGHOST2.
The local definitions can be set with commands such as
TCPIP SET HOST SYSLOGHOST
/ADDRESS=161.72.6.24
Once changed, SYSLOGD should be restarted. There
are specific instructions here or one
could restart TCPIP (see further below).
SHOW SYSTEM
Look for process name such as SYSLOGD or
SYSLOGD_1 and note process id
Stop the process with
STOP /ID=n
Disable the service
TCPIP DISABLE SERVICE SYSLOG
(note that disabling the SYSLOG service does not
not stop the process)
Start the service
TCPIP ENABLE SERVICE SYSLOG
Check that messages can be written to
syslog on the target host e.g.
LOGGER_D "DEBUG TEST MSG"
LOGGER_I "INFO TEST MSG"
LOGGER_N "NOTICE TEST MSG"
Note that these definitions of these commands (DCL
symbols) are made at login and can be found in
SYS$COMMON:[SYSMGR]SYLOGIN.COM
Shutting down and restarting TCPIP
Before starting, shut down the TCS safely,
putting Telescope in to ENGINEERING
and then exiting TCS with TCSEXIT command. Failure to observe
precautions could
leave telescope moving without proper control.
One way to restart is to reboot the
machine.
It can be quicker just to restart TCPIP but
care needs to be taken.
Typically we use telnet to log into an Alpha.
However telnet sessions will close when
TCPIP is shut down. One was around this is to log into another alpha
using telnet and
the use a DECnet connection to the target machine. e.g.
If DECnet is available then it is
convenient to use DECnet to create a remote session e.g.
session on lpas3 to
connect to lpas2 for TCPIP work:
$ SET HOST LPAS2
log in as SYSTEM
If DECnet isn't available use the
system console.
$ @SYS$STARTUP:TCPIP$SHUTDOWN.COM
Note that some devices may be
left
in use. The shutdown suggests using
SHOW DEVICE BGn /FULL
You can then kill process by
process id
$ STOP /ID=m
You could now update the
syslog configuration file
[SYS0.UCX_SYSLOGD]SYSLOGD.CFG
or create a new version
of
log file (to stop log getting too big).
CREATE
[SYS0.UCX_SYSLOGD]SYSLOGD.LOG
New file created by
FJG<control><Z>
Now restart TCPIP
$
@SYS$STARTUP:TCPIP$STARTUP
Now enable (start) SYSLOG
TCPIP ENABLE SERVICE SYSLOG
Check this is working by
looking
for message
syslogd:
restart
in both alpha and ics log
files.
Note that it can take 5-10 minutes for message to be
visible in Alpha SYSLOGD log
file.
When TCS is started at the INT we see message
NOTICE
Starting
D-task for INT
After starting TCTIP mount the
ics
disk(s) with
@SYS$MANAGER:TCS_STARTUP
One can check what's mounted
with
UCX SHOW MOUNT
TCS Software
Manager
Last modified: FJG 16 Jul 20189