[OpenAFS] Small suggestion for RH spec file? (keeping CellServDB up-to-date on *nix clients)

Paul Blackburn mpb@est.ibm.com
Mon, 30 Dec 2002 17:34:41 +0000


Normally, the local site AFS admin team knows what CellServDB
to use for their clients. If they are organised, they maintain a "master"
CellServDB file (usually downloaded to clients via http://) from 
something like:

     /afs/@cell/common/CellServDB

This is propogated to clients at install time and via regular updates.
Client CellServDB updates done either via the Transarc AFS "package" or 
local equivalent
triggered via crontab (and at boot time with /etc/rc.d/init.d/afs).

Maintaining a CellServDB on Windoze clients seems to be more of a problem
(maybe it can be done with perl?).

[side note]: The OpenAFS install process on Windoze is pretty neat.
You get prompted for a URL for the CellServDB you need plus the name of 
your cell.
If this could be done non-interactively, it would be even better.

I have a network install process for OpenAFS which  takes two
command-line arguments: one is the URL for the CellServDB, two is the 
cellname.
This enables AFS install on Linux with a single command line which can be
part of a local customised complete Linux network install.

Once AFS is running, we can apply RPM updates straight out of /afs/
very efficiently (thanks to AFS's intelligent Cache Manager and our
mirrored collection of update RPMs in /afs/) with something like:

    cd /afs/@cell/linux/distro/updates/8.0/en/os/i386 \
    && rpm -Fvh $(ls -1 | grep -v kernel)

The last piece of the jigsaw is a crontab script to keep the client 
CellServDB
in sync with /afs/@cell/common/CellServDB. The solution I use is:

http://www.angelfire.com/hi/plutonic/scripts/cron_sync_CellServDB

As far as the RPM install behaviour for CellServDB is concerned,
I believe it should not overwrite the local CellServDB file.

The process for keeping CellServDB current is something that needs
to be done on a regular basis (eg daily crontab).

Relying on RPM install to keep client CellServDB current is not really 
good enough.
--
cheers
paul                        http://acm.org/~mpb



Derrick J Brashear wrote:

>On Sun, 29 Dec 2002, Paul Blackburn wrote:
>
>  
>
>>Not everyone who uses OpenAFS is using it over the Internet.
>>    
>>
>
>We call those people "unlucky" ;-)
>
>  
>
>>You might be surprised to learn that there are non-public networks
>>that span the globe and do not exchange AFS traffic with Internet AFS sites.
>>    
>>
>
>Do you think it's realistic to do one of:
>-figure out how to provide an openafs-config package which is just a
>cellservdb and perhaps anything else assumed to be user modifiable
>
>or
>
>-assume people will edit CellServDB.local and just not create mount points
>for cells they don't want to use
>
>I'd love to hear an alternate proposal that makes everyone happy, but
>remember, when a db server gets renumbered, the terminal server which
>appears in its place shouldn't end up getting vlserver queries for the
>next 6 years. I really do wish I was kidding.
>
>_______________________________________________
>OpenAFS-info mailing list
>OpenAFS-info@openafs.org
>https://lists.openafs.org/mailman/listinfo/openafs-info
>  
>