Buildworld fails in 5.1
Tillman
tillman at seekingfire.com
Fri Jul 18 09:53:18 PDT 2003
On Wed, Jul 16, 2003 at 02:24:18PM +0000, Bosko Milekic wrote:
>
> Same here, remove the -j N. Right now there seem to be some
> dependencies which you fail against with a parallel build.
And it works here as well.
Building with -j highlights a problem I'm seeing on a sparc64 though
... the stock Ultra 5 hard drive is abysmally slow. Bonnie++ reports:
Version 1.93c ------Sequential Output------ --Sequential Input- --Random-
Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
Caliban 512M 8 95 1995 10 927 5 15 94 1893 6 54.9 17
Latency 1878ms 445ms 2894ms 1074ms 272ms 5349ms
Version 1.93c ------Sequential Create------ --------Random Create--------
Caliban -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 918 72 3008 82 2649 97 697 55 3112 89 2825 97
Latency 1665ms 51352us 2797us 1622ms 345ms 2241us
Compare this to a NFS mount (with options rw, intr, nfsv3, -w=32768,
-r=32768) on the same box (an Ultra 5) from my vinum fileserver:
Version 1.93c ------Sequential Output------ --Sequential Input- --Random-
Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
Caliban 512M 4076 18 1764 17 4102 15 79.5 77
Latency 294ms 184ms 44630us 1699ms
This made the buildworld time /really/ slow:
real 1068m1.246s
user 904m49.768s
sys 80m51.158s
(buildkernel took about an additional ten hours).
I'm hoping that replacing it with a more modern hard-drive should fix
this, though I'll be poking through atacontrol and dmesg to see if it's
a simple configuration problem.
In the meantime it's actually faster for me to build off of an NFS
mount. Heh.
-T
--
Page 461: Tools that are simple enough to use the first day are often a
real pain after the first month.
- Harley Hahn, _The Unix Companion_
More information about the freebsd-current
mailing list