UFS2+Softupdates Corruption Regardless on Seven various systems

[Edgar Martinez]

Scott,

Undoubtedly you're in the midst of several things on top of the 6 release
however this issue is continuing to appear pretty regularly. It occurs on
all systems after approx. 30 days of light use. I am in the process of
generating a LiveCD so that I can use read-only media to address the
problem. I have had to remotely walk someone thru a reinstall at least 3
times so far. The idea is that I can dismount and fsck the storage container
independently. This should correct my issue, but perhaps mask a larger
thing. 

We have verified both UPS on the unit and placed units into an EMI cage to
prevent Power/EMI from being an issue. Yet after approximately 30 days, the
entire container requires a single user mode fsck to fix (if you catch it in
time). Background fsck will only correct the data partition, but the OS
partition is not setup to perform the background fsck.

Systems are only running samba (recent change) to provide data storage to
the local users. Other then that, nothing else outside of the standard log
writes and such change. Seven gateway/nat/fw/dhcp/dns/samba units scattered
thru the world. All started life running a vanilla 5.3-STABLE load.

Anything I can do to help out or get you more information. I know that
there's a summer of code thing out for journaling of UFS, so does that
include UFS2? I haven't gotten into the specifics yet..

Thanks!

-----Original Message-----
From: Scott Long [mailto:scottl at samsco.org] 
Sent: Sunday, June 19, 2005 8:37 PM
To: emartinez at crockettint.com
Cc: freebsd-fs at freebsd.org
Subject: Re: UFS2+Softupdates Corruption Regardless on Seven various systems

Edgar Martinez wrote:
> All,
> 
>  
> 
> I have a network of FBSD boxen running 5.3 w/ 2x PATA WD1200JB Drives and
a
> Promise Fastrack TX2 controller in mirror. The systems mainly just pass
> internet traffic and rarely ever touch the disks. After running for a few
> weeks -> months.the disks become corrupted forcing a manual fsck from
single
> user mode. And since the system is thousands of miles away, it can become
> painful to walk someone with a language barrier thru that.
> 
>  
> 
> Question is WHY does this occur?
> 
> How can you avoid this? 
> 
> What can you do to remotely fix the issue? 
> 
> Any proactive maintenance I need to be doing?
> 
> Did I mention I would like to know WHY?
> 

This certainly sounds like a bug, and is not something that people 
normally see.  When and how do you notice the corruption?  Does it
have a particular pattern?  Would it be possible to try a different
brand of disk controller in order to rule out the driver being
buggy?

Scott
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 4543 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-fs/attachments/20050822/83125b28/smime.bin