-
The latest major version of Facter—Puppet's tool for collecting system information—has been out for some time now. However, we've been hard at work fixing bugs ever since. The fact that Facter has to be able to run on a variety of operating systems and architectures makes maintenance quite a challenging ordeal.
-
InfluxDB
InfluxDB – Built for High-Performance Time Series Workloads. InfluxDB 3 OSS is now GA. Transform, enrich, and act on time series data directly in the database. Automate critical tasks and eliminate the need to move data externally. Download now.
-
For each test, rspec-puppet stubs the fake facts using the custom facts API. The fake facts it gets from facterdb, which is a gem that contains "dummy" facts for a variety of operating systems and Facter versions—basically lots of files containing facter --json output. The resulting information is then fed to Puppet for catalog compilation.
-
We started by decoupling Puppet from Facter as much as we could, introducing the possibility of having multiple Facter backends. While Puppet would use the default Facter implementation when running on its own, external users would be able to define and pass their own Facter implementation when initializing Puppet, similar to how puppetserver configures Puppet to use its JRuby-compliant HTTP client.
-
To avoid breaking the Facter API, we ended up implementing an overcomplicated way of interacting with a hash. Using our dumb Facter backend, custom facts were now simply added to a hash, and querying them would just produce them from the hash if available:
-
And because performance improvements mean nothing without showing the numbers, here's how test times have changed for the puppet-nginx module:
-
One thing I haven't mentioned is that running the same tests with Puppet 6 takes a total of 25 minutes, so there's more to improve in Puppet itself as well. However, from a Facter standpoint it's impossible to make the tests any faster, unless Ruby itself improves hash access speed 😜.
