Note:  this file is meant only for users of ulab, the Sun system
which supports the departmental microprocessor lab for courses
CS341, 444, and 644.

Problems related to reset (reboot) of online SAPCs.

1. When ~r stops working in mtip.

2. Severe performance problems (load factors of over 15) on ulab 
which can be due to "chattering boards", that is, online boards left 
sending streams of data into ulab's kernel buffers.

Solutions

If "~r" in mtip stops working, while mtip itself continues to work,
it means the reset server is not working.  The reset 
server is a special purpose computer running a program
burned into its memory, so it doesn't need downloading itself.
It can be power-cycled if necessary.  It resides in or near the
rack of online SAPCs.

The reset server can stop working because of the UNIX program
(rserver_client on ulab) that sends commands to it.  If this program
opens the line to reset server and then hangs, no other instance of
the program can open the line, until system reboot or killing the
first process.  Any user can find and kill all such processes by running
mkill.  Type "mkill" at the shell prompt on ulab and answer "y" to all
questions about killing rserver_client processes.  To just print a
list of such processes, type "musers". 

Resetting all the online SAPCs at once.

If the system load on ulab rises abnormally, it can be because of
multiple SAPCs outputing constantly:  all this output has to be
accepted and handled by ulab.  To reboot all the online PCs to
stop this behavior, type

mreset all

at the shell prompt.  Be aware that this is not done lightly, since
it resets all the online SAPCs, some of which might be in use.
On the other hand, this does not damage anyone's files, only an ongoing
debugging session. Clearly not much debugging can go on if ulab is
unusable!

If your path is not set up to find mreset, use its full path:
/groups/ulab/pcdev/bin/mreset all

To perform a manual reset, power-cycle the SAPCs in the hardware lab.

If this doesn't help, or you suspect another system performance problem is
involved, notify the operators, for example by "mail operator".  A reboot
of ulab may be helpful.



next up previous
Next: SAPC problems Previous: The mtip Command: examples