The next thing

It has been a couple of months, so I feel that perhaps its time that I mentioned more publicly where I ended up after Google. I am now a systems administrator at Canonical, the makers of Ubuntu. That's a pretty good fit for me in the sense that I have been a Ubuntu user for a very long time. For reference, I don't love the job title "systems administrator" because it doesn't really match what I do, but it isn't something I'm fixated on.

One of the things I help manage at Canonical is our Openstack infrastructure. Along the way I've been finding a few things there I think can be improved, which is why I've been hacking on Openstack in my spare time. I've had a couple of patches merged already, and am generally having fun contributing to an open source project which I think stands a very good chance of being the Apache of cloud management. It is a lot like the stuff I was doing at Google in the sense that I like working on things which I think will affect the quality of life for a large number of people, and Openstack is clearly in that space.

Canonical is also much more open about contributing to open source projects than Google was, so expect me to be able to talk more about what I do in my work life than I did before. I think its already noticeable that I am blogging more than I did during my six years at the big G.

Tags for this post: work canonical google
Related posts: Taking over a launch pad project; Slow git review uploads?; Further adventures with base images in OpenStack; Wow, qemu-img is fast; Are you in a LUG? Do you want some promotional materials for LCA 2013?; Announcement video; linux.conf.au Returns to Canberra in 2013; Openstack compute node cleanup

Comment

Syndicated 2012-02-04 15:50:00 from stillhq.com : Mikal, a geek from Canberra living in Silicon Valley (no blather posts)

Rise of a Merchant Prince

ISBN: 0006497012 CollinsVoyager (1996), Paperback, 496 pages LibraryThing

I didn't really like this book, but I persisted with it because I want to keep reading the series. I thought the previous Serpent War book Shadow of a Dark Queen was weak, but this book was weaker. The book follows the rise of Roo as a merchant, and is improbable at best -- Roo's wealth is generated by cornering the food market for Krondor and I see weaknesses in the analysis there -- surely the Duke wouldn't allow such a manipulation of the market when it harms his citizens, why weren't there food riots when the cost of basic staples jumped to record levels overnight? Basically, it just doesn't seem believable to me. The sequence at the panthian lair on the other hand is much better, and the best bit of the book. Its a pity it is only about 50 pages long. Tags for this post: book raymond_e_feist midkemia combat crime fantasy sword_and_sorcery serpent_war Related posts: Shadow of a Dark Queen; Daughter of the Empire; The King's Buccaneer; Servant of the Empire; The Riftwar Series; Silverthorn; A Darkness at Sethanon; Mistress of the Empire; Prince of the Blood; Magician: Master; Magician: Apprentice; Raymond E Feist's Empire Trilogy; Polar City Blues; Dawnspell: The Bristling Wood; Caves of Steel; Currency; Tipping point: windscreen washers; You Can Be The Stainless Steel Rat; Belgarath the Sorcerer; The Adventures of the Stainless Steel Rat; Mona Lisa Overdrive

Syndicated 2012-02-04 15:30:00 from stillhq.com : Mikal, a geek from Canberra living in Silicon Valley (no blather posts)

An update on Catherine's health

In the last week we've now seen the two specialists that we needed to see to learn more about Catherine's pituitary adenoma. The first was the opthamologist, who kindly saw us at very short notice. Even better, he's Andrew Tridgell's brother and a lovely guy. He did a great job of answering our questions and generally reassuring us, and the short of it is that Catherine's vision is not current disturbed and barring another hemorrhage or a significant growth in the tumor it shouldn't be. He of course couldn't rule these things out, but that's because all things are possible even if they are unlikely. The ongoing strategy here appears to be a series of MRIs and visual field studies done every six months or so for the foreseeable future. The only real wart here was that if there is a hemorrhage, which is something we can't control, the prognosis here could change rapidly for the worse with very little warning. There is evidence of a previous hemorrhage on the MRI.

The second specialist was the endocronologist, who we saw on Monday in Sydney. Again he was a lovely guy and put up with our two pages of questions. As best as he can tell the adenoma is not cancer, but he's not sure if it is functional or not (controlling the level of prolactin in Catherine's body). The first steps are that he's going to take the MRI films to a radiologist specializing in cranial scans, and has put Catherine on a drug which should control her prolactin levels. Then it will be a blood test in a month to see if the drug is working, and we'll take it from there. He was talking about the possibility that this whole thing is related to Catherine's sarcoidosis from a decade ago, but he thinks that only a biopsy of the tumor will confirm that. I feel that if they're going to do brain surgery for a biopsy they may as well just take the darn thing out while they're there, but we'll have that argument when we get there.

So, overall not as horribly bad as it could have been. There are still risks if there is a hemorrhage, and its possible that we'll end up seeing a neurosurgeon to have the tumor removed, but we'll cross those bridges when we come to them. The next step is either that the radiologist will see something on the MRI that he thinks needs more information, or that Catherine will have a blood test in a month. We'll keep you posted.

Tags for this post: health catherine brain tumor pituitary adenoma
Related posts: It hasn't been a very good week; JJJ's hack

Comment

Syndicated 2012-02-04 02:02:00 from stillhq.com : Mikal, a geek from Canberra living in Silicon Valley (no blather posts)

Wow, qemu-img is fast

I wanted to determine if its worth putting ephemeral images into the libvirt cache at all. How expensive are these images to create? They don't need to come from the image service, so it can't be too bad, right? It turns out that qemu-img is very very fast at creating these images, based on the very small data set of my laptop with an ext4 file system...

mikal@x220:/data/temp$ time qemu-img create -f raw disk 10g
Formatting 'disk', fmt=raw size=10737418240 

real	0m0.315s
user	0m0.000s
sys	0m0.004s

mikal@x220:/data/temp$ time qemu-img create -f raw disk 100g
Formatting 'disk', fmt=raw size=107374182400 

real	0m0.004s
user	0m0.000s
sys	0m0.000s



Perhaps this is because I am using ext4, which does funky extents things when allocating blocks. However, the only ext3 file system I could find at my place is my off site backup disks, which are USB3 attached instead of the SATA2 that my laptop uses. Here's the number from there:



$ time qemu-img create -f raw disk 100g
Formatting 'disk', fmt=raw size=107374182400 

real	0m0.055s
user	0m0.000s
sys	0m0.004s



So still very very fast. Perhaps its the mkfs that's slow? Here's a run of creating a ext4 file system inside that 100gb file I just made on my laptop:



$ time mkfs.ext4 disk 
mke2fs 1.41.14 (22-Dec-2010)
disk is not a block special device.
Proceed anyway? (y,n) y
warning: Unable to get device geometry for disk
Filesystem label=
OS type: Linux
Block size=4096 (log=2)
Fragment size=4096 (log=2)
Stride=0 blocks, Stripe width=0 blocks
6553600 inodes, 26214400 blocks
1310720 blocks (5.00%) reserved for the super user
First data block=0
Maximum filesystem blocks=0
800 block groups
32768 blocks per group, 32768 fragments per group
8192 inodes per group
Superblock backups stored on blocks: 
	32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208, 
	4096000, 7962624, 11239424, 20480000, 23887872

Writing inode tables: done                            
Creating journal (32768 blocks): done
Writing superblocks and filesystem accounting information: done

This filesystem will be automatically checked every 36 mounts or
180 days, whichever comes first.  Use tune2fs -c or -i to override.

real	0m4.083s
user	0m0.096s
sys	0m0.136s



That time includes the time it took me to hit the 'y' key, as I couldn't immediately find a flag to stop prompting.

In conclusion, there is nothing slow here. I don't see why we'd want to cache ephemeral disks and use copy on write for them at all. Its very cheap to just create a new one each time, and it makes the code much simpler.

Tags for this post: openstack qemu ephemeral mkfs swap speed canonical
Related posts: Further adventures with base images in OpenStack; Openstack compute node cleanup; Taking over a launch pad project; Speed limit; Slow git review uploads?; My machine was thrashing a lot; Large inodes = faster samba; Are you in a LUG? Do you want some promotional materials for LCA 2013?; Announcement video; linux.conf.au Returns to Canberra in 2013; Linux USB quandary

Comment

Syndicated 2012-02-03 17:16:00 from stillhq.com : Mikal, a geek from Canberra living in Silicon Valley (no blather posts)

Slow git review uploads?

jeblair was kind enough to help me debug my problem with slow "git review" uploads for Openstack projects just now. It turns out that part of my standard configuration for ssh is to enable ControlMaster and ControlPersist. I mostly do this because the machines I use at Canonical are a very long way away from my home in Australia, and its nice to have slightly faster connections when you ssh to a machine. However, gerrit is incompatible with these options as best as we can tell.

So, if your git reviews are taking 10 to 20 minutes to upload like mine were, check that you're not using persistent connections. Excluding review.openstack.org from that part of my configuration has made a massive difference to the speed of uploads for me.

Tags for this post: openstack git review gerrit ssh voices
Related posts: More reviews; Book reviews; Contact details for the Canberra LCA 2013 bid; Working on review comments for Chapters 2, 3 and 4 tonight; A ssh quickie; Review; LCA 2013 bid process opens - Canberra at the ready!; Andrew's SSH filtering causes me pain; Further adventures with base images in OpenStack; Twisted conch; The Wild Palms Hotel; The mechanics of bidding for LCA; Status of the book; Are you in a LUG? Do you want some promotional materials for LCA 2013?; Announcement video; linux.conf.au Returns to Canberra in 2013; Canberra officially expresses interest in hosting LCA in 2013; clusterssh; Openstack compute node cleanup

Comment

Syndicated 2012-02-02 16:53:00 from stillhq.com : Mikal, a geek from Canberra living in Silicon Valley (no blather posts)