amd64-list Digest, Vol 7, Issue 13

[Matthew E. Lauterbach]

On Tue, 28 Sep 2004 18:00:24 +0200, Eugen Leitl wrote
> Does anyone knows what's up with error messages like
> 
> Downloading Packages
> error: rpmts_HdrFromFdno: MD5 digest: BAD
> Expected(3c92110ada4549cf6818835a7c69e7f0) !=
> (37a261901ff63c1c711f112d17ff82d6)
> Damaged RPM 
> /var/cache/yum/updates-released/packages/libpng-1.2.5-8.i386.rpm,
> removing.
> 
> ?
> 
> I'm getting these frequently when running yum update on Athlon64
> or Opteron (Sun Fire) systems. Ctrl+C and rerunning yum 
> usually fixes this.
> 
> Are these timeouts due to overloaded servers?

My guess would be a corrupted mirror.  Can you tell if all the bad packages
are coming from the same mirror?  I wouldn't think that with an overloaded
mirror you would get far enough for yum to do a checksum on the package.
> 
> Why doesn't Fedora use Bittorrent or other P2P means of 
> update, anyone knows?
> 
Not sure.  Probably not trivial to implement.  Plus, you would have to
maintain a seed for each package all at one location to harness the power of
bittorrent.  This could be feasible with the main repository hosting the
seed, the mirrors bittorrenting the packages, and leaving the connection
open to help with downloads.  However, you would have to implement some sort
of failover communication between the main repository and the mirrors so
that one of the mirrors could take over hosting the seed if the main
repository failed.  Then, the client would also have to have some method to
failover and try different seed.  Maybe some other type of P2P would be
better suited to the task?

> (Otoh, FC2 did install cleanly on the Sun Fire, while SLES9 
> did leave a
> broken install, and SLES8 hung on boot).