On Thu, 23 Sep 2004 10:30:01 +0200, Timo Sch=F6ler
<timo.schoeler@macfinity.net> wrote:
>=20
>=20
> > my ultra10/440 experienced the sleep forever bug twice a couple of
> > weeks ago.  it happens on any sparc64 box.  it seems to be more
> > likely to occur if the disk and network are busy (eg, both the
> > above happened when i started writing and read heavily over NFS
> > at the same time.)
> >
> >
> > .mrg.
>=20
> appears to me that there might be a connection between *really* heavy
> disk and/or network I/O (as your NFS example stated) and sleep forever;
> as i wrote yesterday, i have an U2E running for months now, the most
> important service is IMAP. i.e., there are chunks of data read/written
> from and to disk and network I/O for a *short time*, then silence.
>=20
> the U2E is connected to a managed L3 switch so that i have a great
> overview over it's stats (unfortunately, i'm not running sth. like mrtg
> yet).
>=20
> however, seems to me that there's a dependence between sleep forever
> and the apps run on the machine and their related I/O 'profile'...
>=20

That makes sense. My Ultra 1E Model 170 is nothing but a NAT/DHCP/DNS
box - although it gets heavy network load if I'm downloading something
big, it's never enough to come close to saturating either of the hme
nics in it.

My Ultra 5/330 runs fetchmail, ircd, and spamd. So it's load is rather
sporadic - every few minutes it'll have some mail to chew on, but
other than that, no major traffic or I/O to speak of.

However... I now recall speed testing my LAN by a) using ttcp to send
a gig both ways at the same time, and b) using FTP to grab a huge file
and stick it in /dev/null. Perhaps I should try copying a large file
over NFS (or a large directory?)

If we can narrow this thing down, well, every little bit helps.

--=20
Sean