Something rotten in the OS?

Jez Hancock jez.hancock at munk.nu
Fri Aug 22 14:53:25 PDT 2003


On Fri, Aug 22, 2003 at 10:13:30AM -0700, Steve Warwick wrote:
> This problem seems to start after about a month and is indicated by there
> being fragments of the kernel config data in the daily kernel log messages.
> I asked about that on this list but people seemed to think it was just some
> kind of log rotation.
Someone else mentioned the same problem with dmesg and periodic daily
kernel log message output very recently on this list.
What the person replied to your original
message about this said sounds like the most likely reason for the
seemingly garbled output from the daily periodic kern message run.

You can get the problem occuring very easily if you have a stack of
messages written to the kernel message buffer - say for example you fill
up a disk inadvertently (or otherwise) and the kernel constantly
complains about 'not enough disk space' to write files.  This happened
to me just the other day:

output from dmesg:
pt: file system full
pid 83041 (mysqldump), uid 0 on /opt: file system full
<snip 555 lines>
<snip>
fxp0: promiscuous mode disabled

So what Dan Nelson said to your earlier question here:
http://marc.theaimsgroup.com/?l=freebsd-questions&m=105890452032133&w=2

sounds to be correct.

> I caught the machine a couple of months ago with nearly all the swap used
> (800meg out of 1 gig) and rebooted which kept the machine happy.
In that case you would want to be looking at what is eating away at all
your memory using tools like vmstat perhaps or using a graphing utility
to keep track of memory over time?
 
> I recently noticed the kernel log messages had the config fragments again
> and was going to to a reboot - but alas too late. The machine is down, HD is
> damaged and we are trying to get data off the drive...
Again I think the 'fragments' issue is probably a side issue.  Such
'fragments' can be caused by something as (relatively) harmless as a
disk being filled up or a large number of ICMP packets being limited by
the kernel for a few minutes.

> This is exactly the problem that occurred 6 months ago. That time I put it
> down to lousy overheated hosting but now I don't have that excuse.
Or just a badly behaved daemon that doesn't handle memory too well...?

<snip>
> Does FreeBSD have to be re-booted every month for safety?
NO! :)

 
> Should I give up and use Linux?
NO!! :)

As you and others have said this kind of thing is very frustrating and I
know what it's like.  A lot of the time though if you take away all the
major services and just revert back to the minimal OS you started out
with, things go along fine.  It's just hard work figuring out which
component is going bad on you - stick with it! :)

-- 
Jez

http://www.munk.nu/


More information about the freebsd-questions mailing list