March 2008 Archives

keep on feeding it tomorrow.

| No Comments
Worked for a while on the "about the authors" content over some delicious pastor tacos.

It's silly, but I guess this is the sort of thing that's associated with the dead-trees publishing.  You have a physical thing, and so you've got to interact with it in certain ways.  These require, in ways so obvious that I won't bore us both by enumerating them, that we make some more information available in advance.  We've got to sell this damned thing, and in the process sell ourselves.

Anyway, that's over and done with now.  I wish I had more to write for the "tell us a bit about yourself" question.  It really just seems like an invitation to indulge in self-loathing.

hvm device emulation model.

| No Comments



The struggle between abstraction, technical accuracy, and aesthetics.  I find the whole thing depressing, no matter which comes out ahead.

Anyway, you can see here my attempt to depict the HVM emulated devices.  It's a slightly-simplified (and dramatically-changed) version of a diagram in http://www.cmg.org/measureit/issues/mit41/m_41_1.html .  (I think mine is more attractive and clearer, but theirs is most assuredly more complete.)  The biggest difference between this one and that is that I've focused on the write/interrupt aspect as a cycle, ordering the other elements based on that interaction with the devices.  Hence the overall circular aspect.  I think the concept is good, but the implementation could use some work.  The big aesthetic problem with this diagram is that it wastes a huge amount of space in the "emulated device" box, because that box needs to span the entire circle.  Maybe I'll revise it later.

Most likely we won't include anything like this in the book.  But I did rough it into earlier drafts, so it's good to at least get something in there.

shrews evolving into parsnips.

| No Comments
Now I've tweaked this diagram slightly.  I decided that three semantic layers were too many, and the mental model implied by the previous diagram was incorrect ("hardware" appeared both above and below the layer of interest.)




Also, expanding the domU boxes allows me to emphasize their role as 'container' (I don't fully encapsulate the domU because the entire point is that they're somewhat removed from the hardware.)

At a certain point I'll just fall of the deep end and start talking about reifying paradigms or somesuch. Until then, my art blather is barely comprehensible. Look grateful.


I also did a diagram, in my increasingly spare style.  This one's meant to express the workings of iSCSI.  I think all the boxes are too small in this one, though.  Plus, it looks like one of the enemies from centipede.

information-seeking behavior.

| No Comments
A line from the wikipedia!  "In technical writing, recursive organization, where parts resemble the organization of the whole, helps readers find their way."  Accurate and just a touch poetic -- I thought it was very good technical writing, myself.

Today I finished the migration chapter.  I think, anyway.  The truth is that a lot of the diction and syntax feel clunky to me, and I haven't actually tried to migrate a machine with the latest version of Xen. . . but that's what tech review is for.

Next I want to back-fill and make sure that some of the earlier chapters are in shape to send to copyedit.  After that. . .  well, I'm not really sure.  There's not that much left to do, right?  Right?

things man was not meant to know.

| No Comments
In response to Luke's suggestion to keep pervasive snapshots of Windows using LVM -- here's a good article on direct use of the device mapper.  We've come across this stuff before (I remember my shock at the Xen LiveCD's COW system,) but this is a nice overview.  http://linuxgazette.net/114/kapil.html

Since I tend to use defaults whenever possible, I don't have free space in an LVM group to use for domUs.  Instead, I use file-backed instances.  The ability to trick the device mapper into using them like LVs is therefore invaluable.

days and doughnuts.

| No Comments
Time passes.

I don't know whether it's my malaise, or that the Windows chapter is harder than expected, but I'm thinking that it might have take another couple of days to finish.  It's depressing, really.

Another issue is that I'm not sure whether it wouldn't be easier to work on the migration chapter for a while -- but if that turns out harder than expected, I'm still stuck on Windows, but more time has passed and I've got no perceptible progress to turn in.

And I'm sleeping too much.  Life is depressing.  Weep!  Wail!  The rain pours over my wizened heart.

GPL PV drivers

| 1 Comment
Well, so I've blown away my first Windows install on the PV drivers.  Big shock there.

The problem seems to have been that the same device was being enumerated as both a QEMU emulated device, and as a PV SCSI device.  This led to immediate irreparable filesystem corruption.  (Thanks, Microsoft, although truthfully I don't know if any of the Linux filesystems would have done better.)

Interestingly, this seemed harmless across several reboots as long as the backing devices were of type file:.  In that situation, the PV device showed up in the disk manager, but wasn't mounted.  (The INSTALL file with the PV drivers offers a clue: only phy: devices provide the right sort of hard-drive oriented metadata.)  Only when I used the device mapper to access the files as physical devices did Windows automatically map a drive letter -- and incidentally destroy the filesystem.

Anyway, we have to get this working.  I'm not ruling out failure as an option, but neither are we going to give up without throwing a truly herculean effort into the task.  Windows is reinstalling now.  I think I'll back it up this time before I do anything stupid.
http://xenamo.sourceforge.net/

Found a good reference on storage clustering.  Be handy for the migration chapter.
So yesterday we picked up another hard drive, so that I can start over with RedHat and figure out just how much damage this virbr / libvirt emulated bridge nonsense is doing to our networking advice.

I don't know.  Sensible configuration management -- that is, practicing what we preach -- might go some way toward improving our lives here.  Luke issued some gentle chiding on that count.  But it's an awful lot of work, with no guarantees.

And I guess that's my todo list, once I get back to the test machines: fresh CentOS install, install Windows, test GPL PV drivers, and then send off that chapter.

on the road.

| No Comments
This is a blog about Xen, and about the book that Luke and I are writing on Xen.  (Hereinafter the book.)

A little about me:  I'm a student, English major, certified sane by the highest authorities.  I like computers and computers like me.  But sometimes they have odd ways of showing it.
Welcome to my new blog powered by Movable Type. This is the first post on my blog and was created for me automatically when I finished the installation process. But that is ok, because I will soon be creating posts of my own!

About this Archive

This page is an archive of entries from March 2008 listed from newest to oldest.

April 2008 is the next archive.

Find recent content on the main index or look in the archives to find all content.