[volunteers] The 2230 formerly (and presently) known as brie

Edward Cherlin echerlin at gmail.com
Tue Dec 18 17:18:47 PST 2007


On Dec 17, 2007 1:36 PM, Daniel Gimpelevich
<daniel at gimpelevich.san-francisco.ca.us> wrote:
> Beginning a new thread with a quote:
>
> > The Marc Merlin-donated
> > machine in question, a bog-standard VA Linux Systems model 2230 2U, a
> > 2x700MHz PIII, built with an Intel L440GX+ motherboard with all four RAM
> > slots filled with 128 MB ECC PC100 sticks, and 2x9GB + 2x18GB SCSI hard
> > drives, is reported (1) unbootable, (2) sitting in _Ed Cherlin's_ house
> > (not Heather's), and (3) having nothing at all done to it, with no work
> > planned.
>
> The only correction that needs to be made to this is to the second half
> of point #3. The following work is planned:
>
> 1) Attempt to boot the thing with "init=/bin/bash" in its present
> state. All indications are that this will succeed.

That didn't work.

After a bit of thrashing and some pursuing of error indications,
Clement created a new initrd. brie boots now.

> 2) Go through the entire filesystem hierarchy, looking for stuff worth
> keeping, which won't be much, and put it temporarily on another
> machine.

There was nothing but autogenerated files in /home.

Daniel reports copying 12 files. Where to, Daniel? Who should know about them?

Done, as far as I know.

> 3) Unceremoniously repartition all the internal drives and their RAID
> slices

Either Daniel will do this, or he will provide the desired map to
Clement to do it. Unless somebody else wants to join the discussion.
We won't wait long to hear from you, though.
.
> 4) Reinstall from scratch Debian onto the machine. Whether this should
> be Etch or Lenny has not been fully examined, and upgrading Etch to
> Lenny, if Etch is ultimately chosen, would be trivial; however, there
> have been changes to the Linux md driver since the Etch kernel that
> would be a must for use of the 9008 JBOD device, and the third-party
> safte-monitor package, IIRC, more cleanly fits on Lenny than Etch,
> although I may have that backwards.

Discuss. Lenny is in testing, with of course no release date set. Do
we want to be conservative, and use the current stable version until
Lenny comes out, or are we in a tearing hurry to get the JBOD up and
running?

> 5) Take advantage of Heather Stern's offer of many moons ago to provide
> a 1' SCSI ribbon cable ending in a slot cover. During the machine's
> brief stint at Via.net, the external SCSI cable reached into the case,
> directly to the MB, and was held in place solely by the top of the
> case. At some point, a Via.net employee brushed up against the cable
> while working on a non-SVLUG machine, and everything stopped working.
> This prompted the first of two visits to Via.net, with the second one
> resulting in the machine being pulled. The slot-cover-with-ribbon would
> eliminate this half of this problem. Now, for the other half: The
> printed manual for the 9008 clearly states that the end of the cable
> that attaches to the 9008 MUST be secured with the screws on the
> connector, because the vibration of eight hard disks will inevitably
> dislodge it in short order. Our 9008 is a prototype, and so there is
> nowhere for those screws to attach. As a result, periodic visits to
> Via.net will be required to inspect the connector.

Heather said recently that she will be able to provide the ribbon
cable and connector assembly in a few days.

I suggest that we find a way to mount a connector on the 9008 with
screw holes. Epoxy? Drill holes in the case?

> 6) Put the machine back in the Via.net rack.

Did I mention how noisy the fan on this thing is? Hooray in advance.

> 6½) (optional) Fill the hole left by the dead drive in the 9008 with
> another 36GB drive.

Do we have one, or the offer of one, or would we need to buy one?

> 7) Rebuild the md RAID array on the 9008 as a RAID5 with one spare
> drive. IIRC, this is NOT how the 9008 is currently set up, because
> although this setup was agreed upon, it never got the opportunity to be
> turned into reality before the machine was pulled.

Any questions?

> 8) Do useful stuff with the results.

So, 9) We set up one or another form of virtualization. We have
proposals for User-Mode Linux, chroots,  or Xen. I'm not involved in
that choice. I believe that Rick, Daniel, and Paul have something to
say. Something polite and technical that does not rehash what has gone
before.

10) Mailman with spamassasin; Moinmoin; your suggestion here

> Rick's contention that no work is planned is only accurate if one
> considers "no immediate work" to be identical to "no work." He has
> graciously offered to effect the above-explained previously-made plans,
> and more. The primary immediate dispute has been that of timing, with
> Rick favoring diving into the brie plans forthwith. I foresee possible
> snarls when implementation reaches step 8, but Rick feels that not to
> be an obstacle to the first seven. And so, the next time anybody hears
> of this, implementation may well already be on step 8. Thank you all
> for your interest.

Daniel reckoned without me putting my oar in.

So, Rick, whaddya think?

> --
> "Consider that two wrongs never make a right, but that three do."
>                                                        --National Lampoon

-- 
Edward Cherlin
Earth Treasury: End Poverty at a Profit
http://www.EarthTreasury.org/
"The best way to predict the future is to invent it."--Alan Kay


More information about the volunteers mailing list