Re: [Bug-gnubg] Multiprocessing queries

[Olivier Baur]

Le mercredi, 4 juin 2003, à 16:51 Europe/Paris, Ian Shaw a écrit :

> I've been following the discussion about the multi-processing, and I 
> thought I'd make a couple of comments, since I expect to make 
> extensive use of it when it's ready.
>
> I noticed there was some discussion of using hostnames rather than 
> octal IP addresses. I'm on a DHCP network so the latter are liable to 
> change.

Yes, this is a must do.


> Will the "master" be able to tell when new "slaves" are available? I 
> am likely to be able to start a rollout, then later set up more 
> processors when my colleagues go home.

This is not implemented yet. However, processing units (local or 
remote) can be added or removed while they're being used (in theory 
only, since it is not possible now through the actual user interface -- 
you cannot type in commands like "pu add" or "pu remove" while rolling 
out :-)

It can be done in three different ways:
1. Have the master look for available slaves from time to time, by 
using IP broadcasting, but this will work only on a local area network 
(not across the Internet)
2. Have slaves signal their availability to a known master when they 
are switched to slave mode; this could be done by expanding the "pu 
slave" command to "pu slave [<master-address>]"
3. On the master, create a remote processing unit for each known slave 
gnubg (even if it's not available at that time), eg by adding "pu add 
remote" commands in the .gnubgrc file; rpu's which are not available at 
that time will be created anyway, but with a "stopped" status; they can 
later be activated when the slave host is running by issuing the "pu 
start <rpu-id>" command; what would be necessary is to have an 
"auto-start" process which tries to start (reconnect) stopped rpu's 
every n minutes.

Solution 1 is better for LAN's (zero configuration)
Solutions 2 and 3 are better for the Internet



> Will it cope with slaves being switched off mid-operation? My 
> colleagues are happy for me to use their PCs on the understanding that 
> they are free to interrupt GnuBg without warning me.

Yes, the master can already cope with slave disconnections.


-- Olivier