ftp-master down for a few hours

[Peter Wemm]

The new ftp-master hardware is online and data is being copied.  Things 
should be much much much faster once this is finished.

The old old ftp-master was a 3ware-based 8x80G raid5 box with roughly 
480G space (1 spare, which got used).  It had a multi-drive failure and 
died.

The temporary ftp-master was serving the ftp archive from a netapp via 
nfs.  This was far from satisfactory because the machine had nowhere 
near enough ram to cache stuff.  Any rsync operations caused a huge 
amount of nfs traffic.  The netapp for the cluster was running at 100% 
load nearly 24x7.  I wrote a cvsup scan generator a while back and that 
helped the cvsup clients considerably, but it still sucked in general.  
It had 1 pentium4 cpu with 1G ram.

The new machine is a dell 2850 with 4 cores, 6x300G scsi drives set up 
as a 1.2TB raid5 and 4G ram.  It should be several orders of magnitude 
of an improvement over the temporary arrangement.

So, there are a couple of things to watch out for.

1) I expect all the timestamps and attributes to be preserved, but it is 
possible that some might get touched.  There may be an initial round of 
rsync/touch activity.  I hope not though.

2) the old array was 480G, and is now 1.2TB. I suspect that the archive 
will bloat a bit more as we have future releases, ports packages, etc.  
7.x is coming up soon etc.  There won't be a hard limit to force the 
poor release engineers to juggle things around to cram it on anymore so 
I'd expect it to tend to expand over time.

3) there might be a separate ftp-archive for old releases being able to 
be served from here, but that depends on what Ken wants to do.

So anyway, apologies for the lack of notice for the outage, but it 
should be worth it.

-Peter