[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Broken Things...
As Chris knows, the tape drive hooked up to Ultra1 hung, hanging the SCSI
bus, and resulted in massive file system damage to /userspace.
It took about four hours of hair pulling to get the file system to where
it would fsck clean. Numerous files were deleted and/or moved to lost+found,
over 5400 files in lost+found, so there is a hell of a mess to clean up.
I knew we'd have a lot of stuff to restore from tape so my first thought
was to take the old eskimo drives, make a big partition out of them (4x 9gb
would have been enough), and restore the entire dump into that so we could copy
disk to disk things that were needed.
Unfortunately, after Eric left, 400MB into that restore approximately, one
of the old drives crapped out, some sort of mechanical problem with the head
actuator (it's highly audible to put it mildly).
Long and short of it is we're going to have to restore things from tape.
I've got a restore going now of our www site and these users whose directories
were entirely blitzed:
bhc, baldy, farrago, hshinn, neverstp, and eborders.
Other users who have files that ended up in /userspace/lost+found by inode
name, but for whom I haven't figured out where they belong yet include:
6wood, amcc, avery-ec, badger, baldy, bhc, biged98, bloo, bpentium,
cek, crick, dalus, dempt, earthcng, erict, farrago, ghawk, hideaway, htak,
jessamyn, jragon, milla, nanook (lucky me), ncpa, neverstp, nobody, ravensys,
redfoot, root, sandy, smartlst, tweiler, user.
Keep track of anything users say is missing doesn't work and we'll grab
those things from tape on additional passes after the initial restore is
complete.
Customers: Please be patient while we try to put this back together.
Please let us know any files(s) missing and we'll add those to the list of
files to restore on the next tape pass.
Per a recommendation on Exabytes website (this is a bug they knew of),
I've added a second SCSI controller to the machine just for the tape drive so
that if it hangs it won't affect the disks in the future.
I've got to tell you I'm feeling majorly bummed right now. Last month and
this have expenses have exceeded income. Problems with MegaPOP, losing DSL,
etc, has created a lot of additional short term expenses.
And then as we've been experiencing all of the economic problems customers
have been subscribing for much shorter terms greatly reducing the immediate
income.
At the same time we've lost hardware (the entire server) which needed to
be replaced, had overlap in port wholesalers as we moved people from MegaPOP
mid-month, and I've ordered a third T1 to address data congestion issues (which
is due to be installed June 11, 2001).
What I really want to do with Eskimo and what is seeming like a more and
more distant goal, is to reproduce the original interface we used to have on a
single-line BBS, except make it available via the web and telnet as well with
Internet extensions, including the games, etc.
The Internet has huge potential to bring people together that I believe is
almost totally unrealized, and I want to see that potential realized, but damn,
it just seems like everytime I get one foot forward I'm blasted back ten.
Folks, I'd really really appreciate your support at this time in any way
you can, bringing new users to us, subscribing for longer terms so we have
some much needed immediate income to cover expenses and anyone wishing to help
develope code we can use it.
I'm really getting bumbed about this kind of marginal existance. If I had
my drathers, we'd have a second file server as a hot spare, constantly
mirroring the data on the main server so if the first server died we could just
tell the second to go active, UPS for everything, etc. But right now things
are extremely tight.
Barring any sudden rush of new users and income I'm going to be forced to
make some adjustments upwards to the rates which I know will lose a percentage
of customers, but we need some redundancy and cushion for disasters that
invetiably happen.