Booting Problem After Power Loss (fsck)?

[Rishi Chopra]

Please see my reply below:

Ion-Mihai Tetcu wrote:
> On Sun, 08 Feb 2004 21:12:51 -0800
> Rishi Chopra <rchopra at cal.berkeley.edu> wrote:
> 
> 
>>Here's a summary of my problem so far:
>>
>>Server was idle (e.g. absolutely no processes running aside from
>>csh, ttyv0 and ps) when power was cut; server reports a problem mounting 
>>/usr partition upon reboot.
>>
>>I have since tried the following:
>>
>>(1) Booted into single-user mode and ran 'fsck' - the latest output to 
>>the terminal says:
>>
>>****     FILE SYSTEM MARKED CLEAN     ****
>>     /dev/da0s1e
>>     Last Mounted on /usr
>>     Phase 1 - check blocks and sizes
>>
>>After letting the system 'do its thing' for 5+ days, the output did not 
>>change.
>>
>>(2) I tried an 'fsck -p' and got the following message:
>>
>>/dev/da0s1a: 1128 files, 36058 used, 47059 free (261 frags, 58771 
>>blocks, 0.1% fragmentations)
> 
> 
> Do you get the prompt back ? Try fsck -p on / then on /var /tmp and last
> /usr. At least you will know what partitions are ok. Better yet I
> suggest you boot from the second aka Fixit CD and run fsck from there;
> you fsck binary may be broken. Also boot verbose (I don't know if
> safe-mode applies to SCSI, but if it does, try that also).

This is exactly the problem.  The 'fsck' command does not return to a 
prompt.

>>The display has been stuck with that same output for countless hours now.
> 
> 
> Do you have disk activity when fsck seems to be stuck ?
> 
> 
>>Questions I have:
>>
>>(1) Have I suffered a total loss or is this still some way to revover my 
>>filesystem?  After suffering a similar loss with a hardware raid-0 
>>failure under win2k, I was assuming the FreeBSD setup would be more 
>>durable.  I would hate to walk away thinking that a simple power loss 
>>could wipe out a freebsd server under nothing more than one terminal login.
> 
> 
> Generally this doesn't happen. From my experience, it happens if either
> there are problems with the disk access infrastructure (a la timeouts,
> etc. on ata) or something bad elsewhere in the kernel.
> 
> 
>>(2) Why would a simple fsck of the filesystem not work in my case?
> 
> 
> If you have the kernel with 
> options         DDB
> options         BREAK_TO_DEBUGGER
> 
> and no disk activity I suggest that you break to debugger hitting
> Ctrl+Esc and try to gather some info from there. Note that in case fsck
> is actually running this could further damage you fs, but since you
> can't do anything else I would say to give it a try.
> 
> To summarize:
> 
> 1. See if you have disk activity when fsck seems to be stuck.
> 
> 2. Try fsck (-p) first on root, the on tmp, /var, /usr, /home.
> 
> 3. Esp. if fsck / doesn't go ok try booting verbose with Fixit CD and
> run fsck from there.
> 
> 4. If 1 gets you the same results try putting the disk in another
> machine where you have debugging options in the kernel, break to
> debugger and gather info from there (esp. if you're running 5.x try
> asking on current@ what exactly to look for in the debugger).
> 
> 
> 

-- 
Rishi Chopra
http://www.ocf.berkeley.edu/~rchopra