master.kernel.org status

master.kernel.org status

[H. Peter Anvin]

Whew...

Thanks to a tip from Leonard Zubkoff I've successfully reconstructed the 
data on master.kernel.org.  However, before putting the system back 
online or even mounting anything readwrite, I'm doing a full backup of 
the contents.  I'm going to let the backup run overnight, as I'm about 
to head home to sleep.

Clearly, expanding capacity on this system is no longer being 
considered; we need to replace the server.

	-hpa

Re: master.kernel.org status

[Ville Herva]

On Wed, Jan 30, 2002 at 12:07:27AM -0800, you [H. Peter Anvin] claimed:
> Whew...
> 
> Thanks to a tip from Leonard Zubkoff I've successfully reconstructed the 
> data on master.kernel.org. 

After the well-deserved sleep, would you (or Leonard) care to share the tip?
I've been involved in (fortunately, not responsible for) a situation where
attempt to extend a hardware raid set led into state where be had to
reformat the whole system and restore all from backups (what backups we
had). I wonder if that piece of information would have been applicable.

And I do wish to express my respect for the hard work you are doing on
kernel.org.


-- v --

v@iki.fi

Re: master.kernel.org status

[H. Peter Anvin]

Ville Herva wrote:

> 
>>Whew...
>>
>>Thanks to a tip from Leonard Zubkoff I've successfully reconstructed the 
>>data on master.kernel.org. 
> 
> After the well-deserved sleep, would you (or Leonard) care to share the tip?
>


It's rather card-specific, but here goes:

a) On a DAC960PRL, use the "daccf" utility, not "ezsetup" which is what
the Mylex web page recommends (THIS ONE IS THE REASON FOR THE FAILURE IN
THE FIRST PLACE.)

b) After "ezsetup" screws over your configuration, you *may* be able to
recover it by running "daccf -o" and restoring the configuration EXACTLY
AS IT WAS ORIGINALLY SET UP.  DO NOT START A REBUILD OF THE ARRAY.

c) After (b), DO NOT WRITE TO THE ARRAY.  (Boot from a floppy or
SuperRescue CD, then try to access the drives readonly.)  Apparenly the
firmware won't actually write to the disks and start any kind of
reconstruction (which would be fatal to your data if you got step (b)
wrong) until any write operations happen to the disk set.

	-hpa