SUMMARY::in the middle of a crash. (please please help)

From: Piotr Grzybowski (merlin@alek.if.uj.edu.pl)
Date: Fri Aug 22 2003 - 08:09:35 EDT


ehlo!

 few days ago i had a problem; advfs domain gone crazy.
the domain was panicking and so was i and kernel. every
time i was about to mount the domain i recived panic, trap and
reboot, same thing happened when i was not mounting it, but
it existed via directory in /etc/fdmns.
 just after i panicked i posted to the list. i never do it
so quick after crash, but this is more or less production server,
and about 100 users are holding data from many years of scientific
work.
 i recived with the speed of light responses from Dr Thomas.Blinn,
Chris Bryant,Fred N. van Kempen,David Knight,Jose Ramirez,Alan Rollow,
Paul Thompson. i would like to thank you all! it is beyond
my ability of writing to express how gratefull i am for your help!
 i exchanged tons of mails with David [4] and dr Tom [1].
dr Tom suggested that the domain is damaged and thanx to him by
moving it away from /etc/fdmns i was able to boot the system.
together with dr Tom [1], David [4] and Allan [6] we were tracking down
the problem. unfortunately nothig could be done. in version 5.0
nothing really can be done with the damaged domain
(i am not sure what verify really does, but it was panicking the
system allright) since fixfdmn is available in later versions.
 what i did was: get the salvage to work, put every thing i can
recover from the domain to the tape, and then restore from that tape.
i needed to do it fast, people started to come to work, and were
finding themselves with homes of size 8*1024 bytes.
salvage works, when verify does not! i salvaged in sigle user mode,
with the domain in /etc/fdmns. by the way: there is an ERROR
in the manual page of salvage (!!) the correct way of
salvaging from domain is:
salvage [options] domain fileset file
so, for example:
salvage -F tar -f /dev/ntape/tape1c -x -S users_domain u1 /u/merlin/mail/i_love_tru64_managers_mailing_list
and NOT (!!)
salvage ... users_domain u1/merlin/mail/i_love_tru64_managers_mailing_list
 any way, from 317046 files 1903 were not recovered. salvage does not
recover symbolic links, there were about thousand of them, but even
without that, only 0.6% were not recovered. some files were corrupted,
i.e. had pieces of other files, or just trash inside, but in
general i was happy with salvage.
 i would like to thank specially dr Tom and David they were with me
all the time!
 thank you all dr Tom, Chris, Fred, David, Jose, Alan, Paul!

yours,
pg

references

[1] Dr Thomas.Blinn <tpb@doctor.zk3.dec.com>
[2] Chris Bryant <Chris.Bryant@DTAG.Com>
[3] Fred N. van Kempen <Fred.van.Kempen@microwalt.nl>
[4] David Knight <dknight@fitzandfloyd.com>
[5] Jose Ramirez <iwrbpbbtw@yahoo.com>
[6] Alan Rollow <alan@desdra.cxo.cpqcorp.net>
[7] Paul Thompson <paul.thompson@thedacare.org>



This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:49:33 EDT