Date: Sun, 17 Nov 1996 06:30:03 -0800 (PST) From: Peter Wemm <peter@spinner.DIALix.COM> To: freebsd-bugs Subject: Re: kern/1698: sup from around 21:51 GMT 28th very unstable (mmap, maybe?) Message-ID: <199611171430.GAA28420@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
The following reply was made to PR kern/1698; it has been noted by GNATS. From: Peter Wemm <peter@spinner.DIALix.COM> To: Heikki Suonsivu <hsu@clinet.fi> Cc: FreeBSD-gnats-submit@freefall.freebsd.org Subject: Re: kern/1698: sup from around 21:51 GMT 28th very unstable (mmap, maybe?) Date: Sun, 17 Nov 1996 22:20:43 +0800 Heikki Suonsivu wrote: > November 16 kernel is still unstable. For last two kernels I got a > deadlock within 5 minutes of newsfeed starting up, so something might have > changed, from crash and filesystem corruption to deadlock. August 19 > kernel still runs with no serious problems. > > Has anyone got a news server running reliably with kernels later than end > of August (preferably ones with tag RELENG_2_2) ? It seems odd that I can > repeat this so easily, I would assume a lot of people should see this ? I > have gotten metoos from (only?) three people now. This problem has been > there for about 2 months now. Will this do? FreeBSD gecko2.dialix.com.au 2.2-CURRENT #30: Mon Oct 14 03:41:06 WST 1996 10:07PM up 33 days, 10:33, 2 users, load averages: 0.31, 0.51, 0.62 > > We use 4K blocksize. With the previous kernel from about a month a go we > > have been getting uptimes of about a week, I do not know for sure how much > > average would be as I have been about once a week tried to upgrade to a > > newer current. I haven't seen a single corruption problem with that > > kernel, though some panics, so something bad happened about month ago. Have you tried redoing a newfs on the news partition with the default 8K/1K blocksizes? (Yes, it's painful, I know) The machine above was filesystem corruption within about 3 days of each filesystem clean. We newfs'ed the partition 3 times over the last 6 months, the problem kept coming back after a few days. I am also suspicious of some problem that seems to be related to the kernel size and data space layouts. We had a (different) machine crashing daily for about a week, and that was "fixed" simply by changing a few compile parameters. eg: changing maxusers from 48 to 38, changing "pseudo-device vn 4" to 2, and dropping an unused driver. We don't know if this was coincidence or otherwise. I don't like not knowing, but I'm not game to tempt fate on that machine, it's too important. The config file on the crashing machine was practically identical to other machines with nearly the same hardware that were not crashing. You might like to try doing some arbitary changes to the kernel (adding a few extra unused drivers and changing the maxusers params might do it) and see if that has any effect on your uptime. Cheers, -Peter
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199611171430.GAA28420>