** Next PR Triage Wednesday, October 29th @ 10:00 am Pacific. **
*Priorities*
1. Puppet 3.7.3
2. CFacter on the march
3. New puppet doc implementation
4. Code removal for puppet 4
*Commentary*
Puppet 3.7.2 is released! There are a couple tickets that didn't make the
boat (PUP-3302, PUP-3500, PUP-3505, maybe others) and so expect a puppet
3.7.3.
Peter and Michael are continuing on cfacter. Solaris support is still in
the works.
Puppet strings 0.1.1 has been released! It fixes some typos, a bug when
future parser is already enabled, and provides some examples of how to use
the markup.
Code removals are plowing ahead. The master branch's lib directory is now
at 98 KLOC whereas stable is at 101 KLOC.
I'm working on a plan to transition from YAML and PSON to JSON for
serialization (network and on disk). There are several changes around
encoding that will need to be handled in order to keep some functionality
(file content with binary data, non-UTF-8 data in reports). I'm going to be
sending out an email to puppet-dev soon.
Puppet-dev conversations of note:
* Behavior of apply + ENC
The discussion continues at a trickle. Felix has summarized the pros
and cons of the various options. Please help drive it to a conclusion.
* Merits of directory environments and opt-in fallback mode?
The thread has fizzled out. I'm interested in whether others see value
in the proposal.
*Data*
I have some data about profiling!
I got two different data sets. They were pretty comparable in many ways.
Data Set 1 (DS1) catalogs took ~11 seconds on average, Data Set 2 (DS2)
catalogs ~14 seconds. One of the big differences in the setups where that
DS2 was using the MessagePack serialization whereas DS1 was using the
default PSON. The average serialization time for each dataset showed a very
strange outcome: PSON took an average of ~0.03 seconds, and msgpack took ~0.05
seconds. I plotted a histogram of the serialization times and that showed
that there was a huge spike in both datasets on the very low end. This
skewed the average times substantially.
By excluding times <= 0.003 seconds we get
MessagePack
[image: Inline image 1]
PSON
[image: Inline image 2]
This indicates that there is actually a benefit to using msgpack. The high
end of serialization is cut off. There is a large difference in the number
of measurements in the two datasets, so it is still possible that msgpack
is not as large of a difference as we expect. We are not going to switch to
msgpack however. Instead we are going to switch to JSON, but the
performance differences should be similar. According to various benchmarks
that I've found online (and don't have links to right now), the difference
between msgpack and json is negligible.
--
Andrew Parker
a...@puppetlabs.com
Freenode: zaphod42
Twitter: @aparker42
Software Developer
*Join us at **PuppetConf 2015, October 5-9 in Portland, OR - *
http://2015.puppetconf.com
*Register early to save 40%!*
--
You received this message because you are subscribed to the Google Groups
"Puppet Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to puppet-dev+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/puppet-dev/CANhgQXvovGK9qcOTG%3DAAejWBiOmc5701Tyz0hx4XVXviQ66KHQ%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.
