Date: Tue, 25 May 1999 21:04:42 +0100 (BST) From: Doug Rabson <dfr@nlsystems.com> To: Nicole Harrington <nicole@nmhtech.com> Cc: Andrew <mynet@uq.net.au>, freebsd-stable@freebsd.org, Sergey <serge69@nym.alias.net> Subject: Re: kernel panic in 3.2 WAS Re: [Q] How stable is FreeBSD 3.X ? Message-ID: <Pine.BSF.4.05.9905252058520.447-100000@herring.nlsystems.com> In-Reply-To: <XFMail.990525030941.nicole@nmhtech.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 25 May 1999, Nicole Harrington wrote: > > On 25-May-99 My Secret Spies Reported That Andrew wrote: > > A friend of mine who runs a web hosting and anonymous ftp > > server has been running 3.1 Release for about 2 months without > > even so much as a reboot . The box is on a 100mbit connection > > at a data center. He also has a shell service on the same box. > > The box recently set a record for the amount > > of data served from any client at the data center. The second closest > > client which has a whole room full of SGI and SUN machines was > > only using about half as much data. > > > > The machine is a P2-450 with x3 16 GIG IBM ide drives > > and 512 meg ram. > > > > He has just changed another box in a different country to FreeBSD > > 3.2 release from Solaris 2.7 ( x86 ) because of problems. It is now > > running 3.2 Release and everything seams great. Solaris 2.7 dosent > > like ide drives over 8 gig which isnt good considering IBM make 22 gig > > IDE drives now. > > > > Andrew > > I too agree that 3.1 was Very stable and 3.2 for most things seems to be OK so > far as I have used it for many things at the ISP I work for. However I have > come across a reproducable problem with 3.2. just as Sergey has spoken about, > that does not seem to occur in 3.1. > > I am testing the new FreeBSD port of the Inktomi caching server. The size of > the processes that are used can grow quite large and are multithreaded. > I have found that if I set MAXDSIZ and DFLDSIZ too small (like 512*1024*1024) I > can make the server page fault and reboot by making it work hard. (IE make the > process grow) > > SO far a setting of (2*1024*1024*1024) for both seems to be quite stable under > .any load, but it seems unsettling that rather than kill a process that is > demanding too much memory, the kernel page faults. Especially since the process > runs as a user process not as a system process. > > I am no expert on these things, but I welcome the assistance of anyone willing > to help identify the root of this problem to make FreeBSD more stable. Flames > and put downs about a lack of detail or knowlege like I have seen so far, please > send to the linux list of choice to save us both some time please. I use > FreeBSD as it is a professional OS with mostly professional people willing to > lend a hand. What message is printed when the kernel page faults. If there is an instructions pointer in the message, use 'nm -n kernel' to find which function it crashed in. If possible, compile a kernel with DDB, run it until it crashes and get a backtrace. Getting a full kernel dump from a kernel compiled for debugging (use config -g CONFIG for this) is invaluable as it allows a developer to gather as much information as possible. Once you have a backtrace and/or a kernel dump, you need to get someone interested in fixing it. Find out roughly what part of the system is involved and see if you can figure out who worked on it last. If you talk to that person and provide them with all the information you have gathered, then you have a chance of getting a fix (or maybe only a workaround). -- Doug Rabson Mail: dfr@nlsystems.com Nonlinear Systems Ltd. Phone: +44 181 442 9037 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.05.9905252058520.447-100000>