Sorry for the delay. Man, what a month. 

On Thu, Jun 26, 2003 at 01:10:25PM +1000, Simon Burge wrote:
> kpneal@pobox.com wrote:
> 
> > On Mon, Jun 23, 2003 at 11:02:58PM -0700, Greywolf wrote:
> > 
> > The dump in 1.6 is useless for me because it insists on sending itself
> > signals before it is ready. This causes it to hang. I do backups
> > with pax instead because pax, uh, works.
> 
> Can you provide any more deals on this one in particular?  I'm using
> dump both manually and with amanda dumping 30-oddGB to a DLT drive on
> the 1.6 branch without any problems, but I've heard of others having
> problems.

Sure. 

Small network, four machines. All of them Alphas. 

1x 164LX, 600 MHz
1x AlphaStation 200, 233MHz
2x AXPpci33, 166 MHz

All of them have tlp cards, some are DEC de500-?? cards, two are
combo ethernet/SCSI cards. All are 100mb. The hub (yes, hub) is
a fairly old, inexpensive (I bought it new in college) SMC 10base-T
"EtherEZ" (I told you it was inexpensive).

All Alphas are running NetBSD 1.6. 

The master backup script runs on the 164LX. It logs into the other
boxes via ssh or (lately) rsh. Backups flow back through the ssh/rsh
channel or they are sent over NFS. 

The hangs don't happen with the AXPpci33 boxes, they only happen
with the AS 200.

Backups go to a holding disk. I have a script that runs in parallel
and splats backups out to disk while dump drops the next onto a holding
disk. The fact that I'm using a DLT is probably not relevant. Only
one dump runs at a time -- this is 10base-T, after all. Yes, it is
pathetic that my tape drive is 2-2.5 times faster than my network.

Backups over NFS have to go to a filesystem mounted nointr to avoid having
write() fail with 0 bytes written. I've not written patches for this yet.
Backups over ssh/rsh hang in a predictable way. Backups to NFS -o nointr
hang due to the abovementioned signaling problem.

Backups using pax work fine, the only problem being that I'm using pax
and not a real backup program.

My guess is that I'm the old person alive still using 10base-T, and
therefore I'm the only person who sees these problems.
-- 
"A method for inducing cats to exercise consists of directing a beam of
invisible light produced by a hand-held laser apparatus onto the floor ...
in the vicinity of the cat, then moving the laser ... in an irregular way
fascinating to cats,..." -- US patent 5443036, "Method of exercising a cat"