Date: Thu, 6 Jan 2005 13:31:57 +0000 (GMT) From: Robert Watson <rwatson@freebsd.org> To: Mike Tancsa <mike@sentex.net> Cc: freebsd-stable@freebsd.org Subject: Re: unstable 5.3 boxes Message-ID: <Pine.NEB.3.96L.1050106132955.85805C-100000@fledge.watson.org> In-Reply-To: <6.2.0.14.0.20050106001244.04b5abc8@64.7.153.2>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 6 Jan 2005, Mike Tancsa wrote: > At 06:29 PM 05/01/2005, Mitch Parks wrote: > >It seems both my test and production boxes are not stable with > >5.3-Release. I'm hoping for some insight in how to make at least the > >production box not crash. > >The crashes have been at different times of the day and I've found no > >suspicious log activity. At the moment it's been 9 days since the last > >crash, but it has also crashed in less than a day. ACPI and HT have always > >been enabled on this box without problems. > > I would update to RELENG_5 and disable in the BIOS HTT. HTT for the > most part works against you with the current scheduler and seems to add > more instability. But there have been quite a few fixes in RELENG_5 > that you might benefit from. FYI, I had something of an odd surprise recently (about two weeks ago) when I re-benchmarked MySQL between HTT and non-HTT on a dual Xeon and got a faster result with HTT on 6.x-CURRENT. I've not yet investigated sufficiently to figure out why I got that result, or if it's a consistent result. It may be the product of recent libpthread changes but I hope to investigate further in a couple of weeks. There was a slight regression in the normal SMP numbers (maybe a couple of percent) but a massive improvement in HTT... My first goal will be to identify the source of the SMP regression, the second to confirm and reason about the HTT change. Robert N M Watson
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.NEB.3.96L.1050106132955.85805C-100000>