Things I’ve been focusing on:

Learning more about agile methods, especially as it pertains to to dealing with teams that have operational requirements as well.

Working more with my OpenSolaris NAS. Suprisingly, one of my most popular ever blog articles.

Playing lots of music: I’m playing bass and guitar in my cover band’s summer concert this weekend. (Can’t wait!)

Planning an epic road trip for the summer: 5 weeks up the West Coast, starting in July.

And, of course, loving my little 5 month old!

So, more blogs to come.

It was a ‘5×5 Chefs Collaborative’, meaning 5 local chefs (and 1 guest chef, Alex Stratta) hung out together and handled one course each.

They gave us a menu to take home that I’ve attached. All the chefs signed it, which was a nice touch I’d never seen before.

The menu: Providence 5×5 Signed Menu.pdf

Every course was incredible. Our favorite courses were probably the amuses, the hamachi sashimi, and the best testament to the power of sous vide I’ve had in a long time — Alex Stratta’s duck.

Yum.

To make things easy, I use the current timestamp for this, as I’ve never seen that number go backwards.

Ok, not the normal kind of thing you do with Puppet, but it can be done. We just need to know what time it is in the template.

Thankfully Puppet lets us define custom facts, so here we go…
in the module’s plugins/facter directory, I created ‘epoch_time.rb’.

Facter.add("epoch_time") do
        setcode do
                Time.now.to_i
        end
end

Then I could use that in templates/heartbeat/cib.xml I said

 <cib admin_epoch="<%= epoch_time %>">

Well, that works fine, except for one little issue.

When puppet manages a file it does this:

1. Renders out the templates and takes an md5sum of that
2. Sends that to the client, which checks it’s own md5sum
3. If the sums are different, the client overwrites the file with the old one
4. The client then updates anything ’subscribed’ to that file (for example, restarting a service.)

So since puppet runs about every 30 minutes, that’s not good. I don’t want to be reloading the CIB database every 30 minutes. That’s bad.

This is what we get:

debug: File[/etc/ha.d/ha.cf]/checksum: Initializing checksum hash
debug: File[/etc/ha.d/ha.cf]: Creating checksum {md5}18530322762561ce59f1d414340b4c43
debug: File[/etc/ha.d/cib.xml]/checksum: Initializing checksum hash
debug: File[/etc/ha.d/cib.xml]: Creating checksum {md5}a02f0ca8a3cce64d7913faa3268e530b
debug: File[/etc/ha.d/cib.xml]/content: Executing 'diff /etc/ha.d/cib.xml /tmp/puppet-diffing20090415-3486-ska653-0'
1c1
< 
---
>  
debug: File[/etc/ha.d/cib.xml]: Changing content
debug: File[/etc/ha.d/cib.xml]: 1 change(s)

So here was my solution, and as inelegant as it is, it pretty much works.

1. Manage a file called ‘cib.xml-puppet’ instead
2. In this file, set admin_epoch=”0″
3. When THAT file changes, do a chained action which sets the epoch time correctly, and writes out the actual cib.xml file
4. Run the command to reload the CIB database in that and only that case.

So the puppet code to make this happen:

        exec { "add-epoch-cib-xml":
                command => "sed 's/admin_epoch=\"0\"/admin_epoch=\"${epoch_time}\"/' /etc/ha.d/cib.xml-puppet > /etc/ha.d/cib.xml && cibadmin -R -x /etc/ha.d/cib.xml",
                path => ["/bin", "/usr/sbin/"],
                subscribe => File["/etc/ha.d/cib.xml-puppet"],
                refreshonly => true,
        }
        file { "/etc/ha.d/cib.xml-puppet" :
                type => 'file',
                ensure => 'present',
                content => template("ha_nfsroot/heartbeat/cib.xml");
        }

So I still get to use my very first custom fact (it gets interpolated into the ’sed’ command) but I also don’t run that terrifying update every time.

If any puppetmasters or Linux HA gurus want to tell me exactly how I’m doing it wrong I would love to hear that. This certianly doesn’t feel “elegant.”
However, since it feels “functional”, functionality trumps elegance for today.

That took the form of a very very sweet hack that I will publish at a later date, but for now I’ll just share some of the results.

In the sidebar I am using pretty standard WordPress plugins to load in

• my latest images from Flickr
• my latest tweets from Twitter
• my recent bookmarks from del.icio.us

These plugins are all implemented as server-side fetches.

Here’s the thing, those services aren’t always very fast. For example, on a recent run I got these results:

I guess I need to either switch to using an in-browser widget model, or cron a fetch of that data locally on some reasonable interval. That might be a nice service for WordPress to have internally for widget authors: ‘cron/cache mode.’ Provide a URL to curl periodically that updates the caches on all your active widgets, as well as a global configuration about how fresh you want your data to be.

I don’t care if any of those results are more than an hour up to date, I’d much prefer to have my page load time cut by 3.24 seconds than have it be up-to-the second accurate.

In any case, here’s the essence:

When you’re cooking something that has it’s own residual moisture, use that moisture (instead of additional fat) to help cook the items by putting a lid on your pan.

2 good examples I’ve used this on recently are mushrooms and onions. Toss a bit of olive oil or butter in a pan, add the veggies and cook for a bit, then lid the pan. Because the environment gets so steamy you’ll actually be able to cook the vegetables more than you would have (for the things like mushrooms and onions, where ‘cooking more’ is more virtue than crime) without them starting to stick and burn.

Mushrooms in particular are known for absorbing any fat you throw in the pan, then, as you cook, spitting it back out, and getting greasy.

Capturing the water vapor (deliciously flavored water vapor, I might add) and using that to keep everything nicely lubricated solves that problem.

Then, at the end of the process, when you’re close to the texture you want, pop the lid off, and let that little bit of fat you added in the beginning do it’s magic and brown things all up.

I’m curious to see if this works with other ’superabsorbers’ like eggplant, but that’s for another day.

Thankfully, within a month I should be
-1 key and keyless from selling the Subaru
-2 keys from my subleased office I am moving out of
-2 keys from selling my scooter
-2 keys from my new office (once converted to use the RFID).

So for now I just have to live with a bloated ring.

I wonder how many years it will be before we look back on keys as a hilarious antiquated concept. There’s already enough hardware in my car and office building that they could at least offer an iPhone compatability mode. And you only have to search ‘bump key’ on YouTube to realize how silly ‘traditional’ locks are.

Oh well. Time to see if I can get back to sleep.

To celebrate his third year with us I got a embroidered coat made.

This little tar contains all the files you need to make this work. Careful with it, as I have provided a minimalist .vimrc so don’t unpack this in your home directory.

puppet_vim.tar.gz

In any case, it’s easy to install on a Mac if you’ve got MacPorts set up:
sudo port install ghc

However, the install takes a LOOOONG time. So long, I was wondering if something had gotten hung. I could see (and hear from the fans doing their best) that the CPU was busy. Running top just showed that a process named ghc (the Haskell compiler itself) was working in some way. But DTrace shows exactly what’s going on.

This trivial example (from Brendan Gregg’s DTrace Oneliners) shows the files getting opened by matching the ‘open’ system call, when it’s entered, and printing out the name of the app that called it (execname) and the first argument to that function — in other words, the file.

sudo dtrace -n 'syscall::open*:entry { printf("%s %s",execname,copyinstr(arg0)); }'


1 17657 open:entry tclsh8.4 /opt/local/var/macports/software/ghc/6.10.1_9+darwin_9_i386/opt/local/share/ghc-6.10.1/doc/ghc/libraries/directory/LICENSE
1 17657 open:entry mdworker /opt/local/var/macports/software/ghc/6.10.1_9+darwin_9_i386/opt/local/share/ghc-6.10.1/doc/ghc/libraries/containers/minus.gif
1 17657 open:entry mdworker /opt/local/var/macports/software/ghc/6.10.1_9+darwin_9_i386/opt/local/share/ghc-6.10.1/doc/ghc/libraries/containers/minus.gif


And there we go. You can see the installer monkeying around in the documentation directories.

This technique works for pretty much any installer/process you might be curious about. In the Bad Old Days, I’ve been able to get this kind of insight from things like strace — watching as it runs a given process. To me, one of the most beautiful things about the DTrace model is that it doesn’t matter if (as happened in this case) there are a lot of processes getting run — you can still find exactly what you need without stressing with PID’s. Strace works on a process, DTrace works on the system.

If you’re on a Mac, open up a terminal and have some fun with the one-liners I linked to. You’ll be amazed by what your computer’s doing when you’re not looking.

In any case, there have been days (like this Friday) where for some reason I can’t make it. No problem, CrossFit can be done anywhere. However, a lot of the workouts, at the very least, require a pullup bar. I really want to get one for my house and office, but haven’t done that yet. Also, I wanted to find workouts I could do on the road as well where the equipment I’ve got on me is extremely limited. (Kettlebells and Packing Light don’t really go that well together.)

I googled for “CrossFit Bodyweight” and found this great document:
Bodyweight WOD’s. (A ‘WOD’ is the Workout Of The Day, the workout a given CrossFit affiliate has their athletes do that day.)

Lots of them still require the ability to do pullups (again, no bar) or include moves that I’m still working on, like handstand pushups or 1-legged squats.

So I clipped out (for my reference, but figured it might help someone else, too) a few of those that I am able to do, and only require one piece of equipment: some ability to keep time. (Well, and in some cases something approximately 2 feet tall that you can jump onto.)

I’ve retained the original sourcing on all of the WOD’s that had them.