Experimental Physics and
Industrial Control System

Ned Arnold <[email protected]> · Mon, 24 Apr 2000 08:39:33 -0500 (CDT)

Steve,

Here's a few more ideas to add to your pondering ...

- In some of our custom designed modules, we have a register bit that
  indicates if a soft reboot or hard reboot occured (basically it's a bit
  that gets set by software and only gets reset if a SYSRESET occurs). 
  If it would help you recover properly, you might want to add some hardware
  like this.

- The save/restore tool logs the state of a list of PVs and attempts to
  restore them to that state during reboot. Thsi would take the place of
  the dbpf in the st.cmd. I've had mixed success with this, but in many cases
  it works perfectly.

- If your external hardware state can be read back, you can set the record
  state to the hardware state at boot time with a sequence record that 
  reads the readback and sets the output. This sequence record would be
  executed one time, probably by poking it in the st.cmd.

  	Ned

> 
> Dear Andrew, Ned and William
> 	Thanks very much for your quick responses, and on Good Friday at
> that.
> 	I already observed the nice bumpless reboot feature of the
> XVME-240 -- if I do a "soft" restart (like if you type cntl-X or hit the
> "abort" button), then the output values of the XVME-240 do not change
> while the EPICS system restarts.  This is great, and I wouldn't have a
> problem if "soft" reboots were the only thing that ever happened...
> 	Unfortunately, I have to anticipate the occasional "hard" reboot
> (like if the processor hangs and you have to do a power cycle or
> press the "reset" button).  During a "hard" reboot, the XVME-240 is
> cleared, and so I need for it to come to a failsafe setting when
> everything is restarted.  In my system, the failsafe position will be "all
> outputs high" -- hence I am trying to find a way to force that condition
> when the database starts up.
> 
> 	Of course, it would be nice to have the best of BOTH worlds:
> bumpless reboot during a "soft" reset, and failsafe condition after a
> "hard" reset.  That requires a method to know what kind of reboot is going
> on.
> 	Ned's suggestion to execute explicit commands (dbpf "mybo","1") at
> startup is a possibility for me, since I have automated database creation
> scripts.
> 
> 	Hmmm... Another idea might be for me to have my very own
> definition of the bo record, with initial value "1" instead of "0".  I
> think that would allow me to have the "best of both worlds" as described
> above.  But now I am beginning to speculate, always a good time to end the
> email.
> 						thanks,
> 							Steve
> 
> ()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
> Stephen Pate		Department of Physics		office:   505-646-2135
> 			Box 30001, Dept. 3D		dept:     505-646-3831
> 			New Mexico State University	fax:      505-646-1934
> [email protected]		Las Cruces NM 88003		home:     505-382-6880
> 
> 
> 
> 
> 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System