19. 02. 2014 No comments

Cache PECL builds on Travis

One of my customers, Imagine Easy, is making extensive use of Continuous Testing using Travis CI. Some of the test runs require additional PHP PECL Modules to be compiled and installed, like php-apcu or gearman support. Since Travis provides a new provisioned VM for every test, this happens before every test run, and slows it somewhat significantly down.

While Travis has support for caching similar artifacts between builds for other types of data, it is not yet supported for PHP modules.

So I wrote a small shell script and the accompanying travis setup which copies the modules to a cacheable location after compilation, and tries on start to use those cached modules before compiling it if necessary. A sample setup is to be found at Github.

By the way, Imagine Easy is hiring in New York and Berlin – see some of their job offerings here. I can fully recommend working there, the team and technology is pretty awesome.
24. 11. 2013 No comments

WordPress: Speed up akismets “show approved comments” feature

The WordPress Anti-Spam-Plugin Akismet features the option to show the number of approved comments for each comment author in the comments overview.

Akismet Comment Count

While this is a somewhat interesting feature to see if this is a regular user commenting, it significantly slows down the admin interface on blogs with a large comment database. This can be fixed by adding a new index to the database assisting the underlying query Akismet does:

CREATE INDEX akismet_comment_count_loggedout ON $PREFIX_comments (comment_author_email, comment_author(100), comment_author_url(100));

Please note that you have to replace $PREFIX with your table prefix, obviously.
22. 10. 2013 3 Comments

Vagrant: Cache packages between rebuilds

Sometimes I have to destroy and reprovision a vagrant box several times. This causes a lot of redundant bandwidth – varnish downloads all those ubuntu packages over and over again. My initial idea was to set up a transparent squid proxy on the host machine, but this turned out to be quite a maintenance nightmare, so I abandoned it.

Luckily, there exists a vagrant plugin fixing my problem: vagrant-cachier mounts several cache directories like /var/cache/apt to the host machine, so they will be already available for a newly provisioned vagrant machine. As soon as my PR is being accepted, composer downloads will be cached the same way, too.

PS: I collected some interesting Vagrantfiles over at my pinboard profile.
02. 10. 2013 No comments

Vagrant: Inline provisioning script

Sometimes I use Vagrant not for a full featured chef provisioning setup, but just with a shell script which installs some packages and minor modifications. This can be done using a separate shell script, but I prefer having it inline, so the whole setup is visible within one file:
13. 08. 2013 No comments

Vagrant: Naming your boxes

Being a pretty heavy user of Vagrant, I tend to have multiple Vagrant boxes idling on my machine.

The plugin vagrant list helps to shed some light there by showing a list of all current existing vagrant boxes and the current status in the console. However, the name of the box is by default just the name of the directory the Vagrantfile resides in and a numeric id, resulting in something like var_1370856673.

Thankfully, there exists the --name virtualbox-parameter, where you can provide a better custom readable name for your box:

And suddenly, the list of boxes on your machine make a lot of sense.
17. 03. 2013 No comments

Debugging OpsWorks

Amazon Opsworks, the awesome tool formerly known as Scalarium, stores all its logfiles and json-dumps of every task being executed in /var/lib/aws/opsworks/chef at the instances.

The JSON-dumps are quite handy to figure out what variables are available to be used within the custom cookbooks. To have them somewhat more readable, use python:

cat <timestamp>.json| python -mjson.tool

Update: Or, way simpler: Use opsworks-agent-cli get_json – thanks, till. :)