> Hundstad, Jeffrey E. wrote:
>> It's too bad that XFS is losing people.
>> [...]
>> If there already isn't too many hurt feelings is it possible to get a
>> report on what TYPE of usage is causing problems. I've been using it
>> on a 700GB server serving FTP, Samba and NetAtalk users to a couple of
>> hundred university folk. Since we've gotten are hardware problems
>> solved we've had no problems. I run it on my squid cache, my local
>> workstation and laptop, no problems there either despite the batery
>> running out and the XYL unplugging the hardware.
One thing I've noticed -- and feel free, guys, to correct me if I'm wrong,
since I'm *not* a kernel hacker -- is that "Built-in Kernel Debugger
support" is enabled by default, at least for the CVS copy I have. The
problem here, I believe, is that if the system has an OOPS, instead of
logging it to syslog and going on its way, it dumps you into the debugger.
While this doesn't crash your system, per-se, it *does* make it unavailable
until someone manually types "go" at the debug prompt. This can create big
problems in a production environment (alas, I proved this myself last week).
I've since re-compiled all my kernels to -not- have the debugger included,
and all proceeds smoothly. Perhaps the debugger should _NOT_ be enabled by
default in the .config file included with XFS' Linux kernel tree?
$.02,
-Ken
|