With the integration of the SMP lockup patches in pre-5
can I expect less lockups?
We get a 100% repeatable problem:
Boot system, start a database load . This works OK.
load goes up to 1.3 Load of all 4000 (very wide table)
records goes well, completes. If we start database
loads of two separate tables at the same time, the
system will crash within a minute of the second IO bound
process starting. Load goes to 2.5, then system wedges,
sometimes with no messages, sometimes with "Couldn't get
a free page" Even 2 bonnie's running side by side to
the same disk will do it. We get similar crashes
in 2.0.29, 2.0.30 and we got it with 2.0.31-pre3.
I'm building a 2.0.31-pre5 now, but we are testing
with a uniprocessor kernel.
Database (KE Texpress) is very stable on uniprocessor
2.0.30/IDE machine sitting beside it running 2.0.30 w/no patches.
all tcp/ip fancy stuff turned on in kernel, using de4x5.c
on the same brand ethernet card.
-- Jay Thorne The Net Result System Services jay@result.com Http://net.result.com Zoom 505 Effect page http://net.result.com/~jay