[NCLUG] Failing drive or controller...

jeff jeff at themoes.org
Mon Dec 18 14:37:06 MST 2006


Benson Chow wrote:
> It's not enough information from just seeing a sector or so fail to 
> pinpoint whether it's a dying harddrive or controller.  Plus the fact 
> that recent IDE controllers have some intelligence on-disk that the 
> drive itself has part of the controller.  But usually it is the disk 
> module that failed though -- if you never had any disk failures before, 
> here's a good rule of thumb - blame the disks first.
> 
> There appears to be two errors in that screen shot - thermal 
> recalibration failed and timeout while read sector.  While something as 
> silly as the hard drive getting suddenly disconnected could possibly 
> cause both, having a disk that went badly out of alignment and 
> triggerred the disk's onboard microcontroller to give up is the more 
> likely scenario.

Ok. Thanks for the info. :)

> Either way, regardless if it's the drive or the onboard controller, it's 
> likely time to have someone visit the sick machine and someone who's 
> authorized to visit the machine at FRII is the only real answer :-( 
> What's the colo contract say about dying machines - and do you have 
> someone locally authorized to visit the machine?

Yes, I do, but he's much more confident with GNU/Linux than OpenBSD.

> (For this type of situation where I'm far away from my box and must need 
> high uptime I'd probably have set up the machine with some sort of hot- 
> or cold- swappable RAID system that can automatically figure out and 
> rebuild any replaced redundancy... and have someone who does not need to 
> know anything about the system that could replace disks available...)

Oh for sure--it's just not all perfect yet. ;)  What I really wanted was to set 
up CARP...
http://en.wikipedia.org/wiki/Common_Address_Redundancy_Protocol

The system is still marching on, which is cool. I just can't log in and change 
any rules.

Perhaps this thread can turn into: what's the "best" 1U system for an OpenBSD 
firewall? Something that can hold at least 6+ ethernet (presumably 2 onboard + 
4 port) 256 megs ram, Gig Mhz or so. I've tried some of the OpenBSD recommended 
vendors in the past, but haven't been too wowed. No puffycomputing.com that 
I've seen. ;)

Thanks again,

-Jeff



More information about the NCLUG mailing list