[OpenAFS] OpenAFS algorithm

Fabien COMBERNOUS fcombernous@kezia.com
Mon, 19 Apr 2010 09:30:24 +0200

Thank you for your input.

Harald Barth wrote:
>> My question is how to permit client in B to use server in B ? I
>> didn't found any document explaining the algorithm used by OpenAFS
>> to decide the server contacted by the client.
> The algoritm is very old and if your IP numbers do not reflect your
> network topology it probably yields incorrect results.
What do you meens by an IP number do not reflect the network topology ?

> Have a look at man fs_setserverprefs and man fs_getserverprefs

I already tried to play with. but without any result.
> a lower value indicates a greater preference
> but first check that the clients really knows that there is more than one copy (something like this):
> $ fs where /afs/pdc.kth.se/home/
> File /afs/pdc.kth.se/home/ is on hosts kinilaw.pdc.kth.se morena.pdc.kth.se sculpin.pdc.kth.se houting.pdc.kth.se 

The fs where command report me that the path is hosted only on one 
server. But i have a RO copy on the two servers. How to use the local copy ?
> As my list from fs gets is quite long, I came up with this to find out
> which one should be the prefered server:
> $ fs gets | awk '$1 ~ "'$(echo $(fs where /afs/pdc.kth.se/home/ | awk -Fhosts\  '{print $2}') | sed "s/\ /|/g")'"' | sort +1 -2n
My fs gets retyurn the two servers. And in the office B the server B is 
with priority 54, server A is with priority 30001.

/unix system engineer/
www.kezia.com <http://www.kezia.com/>
*Tel: +33 (0) 467 992 986*
Kezia Group