Subject: Re: S L O O O O W MV2 source tarball unrolling?????
To: None <>
From: Havard Eidnes <>
List: port-vax
Date: 10/25/2000 17:48:16
> I was unrolling the source tarballs of current on an MV2 last
> night, and started the tar -xpzf yadayada thing, and 7 hours
> later it was still spinning the head actuator motor back and
> forth and back and forth.  An abort and a check of the thing
> indicated it had unrolled only about half after 7 hours.  I
> know an MV2 is slow, but this is entirely rediculous, at best.
> Does it normally take this long, or should I unroll from
> unzipped tarballs on a separate HD for speed.  Or, should I
> unroll on my M76 and selectively port back only the desired
> VAXbits or NFS mount or ....?

If you can tolerate the risk, you could try to remount your file
systems with "async" before unpacking the tarballs.  Please
remember to remount "noasync" afterwards, and pray that you don't
get a power outage while unpacking is ongoing, as the danger is
that you'll end up with an inconsistent file system.  (It'll
still take a considerable amount of cpu to run gunzip, so don't
expect a miracle...)

...or if you have the kernel and user-land support, be brave and
try the "softdep" mount option, which should ideally give you
similar speed improvement, but maintain a consistent file system
image on disk at all times.  There are of course two hitches: one
is that this may be a "chicken and egg" situation, in that if you
come from a 1.4.X-based system, your user-land mount binaries
won't know how to do this, and secondly: I have heard people
claim that the softdep code is not entirely stable.

> Out of curiosity, I ran up 1.4.1 originally on the beast, and
> it was relatively fast.  The 1.5A1 suite seems much slower by
> comparison.  Any idea of what might be going on?  Boots and
> shutdowns are very significantly slower.

That's probably due to the new /etc/rc.d based startup scripts,
which may create quite a few more processes now than the old
single-file startup script system did.

Best regards,

- H=E5vard