[OpenAFS] Memory cache question

[Ray Link]

On Thu, 7 Nov 2002 Martha.Cofran@cexp.com wrote:

> my web-servers to cache the most frequently used images (done automatically
> at boot so users aren't affected) of the 1Gig that we have.  Management
> wants to tar up all of the images on AFS every night and copy them to the
> web servers, so that if AFS disappears, recovery would only consist of
> untarring the current data and reconfiguring the directory the web server
> pulls from.  I figured the best way to do this is to run a cron routine to
> tar it to a volume.  Then each client can cp the tar from the AFS directory
> to a local directory.  But, if I do this, will the 1Gig tar file completely
> flush all of the real data out of my memory cache?  Is there a better way to
> accomplish this?

[Note: odd requirements sometimes require odd solutions...  ;-) ]

The cache is a non-removable part of the actual transport of data
housed in AFS.  If you want to avoid hitting the cache, change the
transport mechanism...

You could have a separate machine running nothing but an AFS client and
sshd.  Then, you can pull the tarball down to local disk on the
webservers via scp, from the AFS space mounted on the intermediary
machine.  That way, the 1GB tarball hits the cache on the intermediary,
rather than on the webservers.


==== Ray Link === University of Pittsburgh CSSD === rlink@pitt.edu ====

A bad random number generator:  1, 1, 1, 1, 1, 4.33e+67, 1, 1, 1