[RC5] Possible bug

bdragon at distributed.net bdragon at distributed.net
Mon Jul 14 19:32:17 EDT 2003


> 
> I've got a number of clients set up to all work from a common set of
> buffers accessed over NFS, so that I can run a process to do the
> fetch/flush while the rest of the systems crunch the packets:
<snip> 
> The working dir of all the clients is the same as the
> alternate-buffer-directory.
> 
> Did I tickle something in the fetch/flush code that caused this due
> to my configuration of the clients?
> 
> Any comments or recommendations would be welcome.
> 
> 	-Michael Pelletier.

"  5.3  Flushing and fetching using remote buffers ----
  
     "Remote buffers" are simply buffers that are serviced by another
     client. When fetching/flushing from/to remote buffers, the client
     opens the remote files and moves what it needs from/to its "local"
     buffer files. The difference betwees "remote buffers" and "shared
     buffers" is of course the fact that with "remote buffers" each
     client has its own files. Lock contention is thus minimized and
     they can work without the user having to worry about the network
     (if one exists) between them failing." -- dnetc.txt

if using remote-buffers, that directory should be the working-directory
of only one of the clients, and said client should not have the
alternate-buffer-directory directive. In this way, only one client will
be actively fetching and flushing, and the other clients will siphon
off workunits when needed. This _should_ also allow the individual clients
(unless configured not to) to initiate direct requests for blocks in the
event the remote-buffer is empty.

in either way, you should really have separate working directories for
all of your individual clients.

As someone else mentioned a personal proxy might be a better avenue
for a variety of reasons.



More information about the rc5 mailing list