[chef] Opscode Chef survey results

[Bryan McLellan < >]

A week ago we put out a survey asking some questions to help drive the
work Opscode does stewarding Chef releases. We'd like to share some of
the results with everyone in the community.

The survey was initially released to twitter and the mailing list,
pulling in 59 and 55 responses before I closed them this morning. Upon
request, I provided another link to IRC which pulled in 4 responses
for a total of 108 responses. Unfortunately, Survey Monkey only gives
you access to the first 100 responses. I've talked to our marketing
folks about sharing a tool with them in the future. We tried to target
this survey to current Chef users and it was intentionally more
grassroots than a survey that marketing put out a couple weeks ago. I
believe they offered free t-shirts whereas we only offered our undying
love and goodwill. I've seen the results of the survey that marketing
did, but I'm not at liberty to speak directly to its results. I will
say that people clearly love Chef, or maybe they just love t-shirts.
Some of the results are interesting, but they're mostly target market
type questions and not directly relevant to the open source projects.

We did well targeting the community, with over half of the responders
running their own Chef server. We were psyched to see that while 65%
of you were using the 0.9 release, 60% were also using 0.10. This
probably indicates that most people are trying 0.10 but haven't
upgraded all of their nodes yet.

* How do you currently install Chef?
69.0% Rubygems
16.0% Debian packages
8.0% Redhat packages
7.0% Other (Ports, from source, etc)

This wasn't particularly surprising as we've always focused on
Rubygems being our primary release mechanism and treated Debian
packaging as a release format for those who couldn't live without
traditional packaging. We've been shepherding some community work on
Redhat packages lately, and are considering committing to something
official.

* What role or job description most closely matches what you do?
32% Devops
(other answers omitted)

This question was also in the marketing survey, in which 40% chose
Devops. I included this question to see the difference in responses
between the community and those who answered the other survey. Surely
there was crossover, and this provided some indication of that. This
is interesting as some believe that devops is strictly a cultural
change where development and operations work more closely together,
whereas others believe it represents a new job title for someone who
does both. As a sysadmin, I've always been a
sec/dev/sys/net/hvac/ops/janitor, so I enjoy being a fly on the wall
amidst these debates.

* What is the primary operating system or distribution in your environment?
54.2% Ubuntu    
17.7% CentOS    
12.5% Debian    
7.3% Redhat     
3.1% FreeBSD    
2.1% Other      
1.0% Fedora     
1.0% SuSE       
1.0% Windows    
0.0% Gentoo

As expected, this somewhat mirrors what Chef currently supports. There
are a few differences between Ubuntu + Debian, as well as CentOS +
Fedora + Redhat, that Chef providers, Ohai plugins and Cookbooks have
to take into account, so we're gauging where we should be spending
time testing across these platforms. It is interesting to ask
ourselves to what degree low numbers are due to a cultural difference
between the users of that operating system or distribution and the way
that Chef works, or because we haven't provided enough support to
encourage those users to join our community yet.

 * How important is it to you that Chef can manage these operating systems?

                Not important   Fairly important        Important             
  Very important  Essential
Debian  20.2% (17)      8.3% (7)                        23.8% (20)      14.3% 
(12)      33.3% (28)
Ubuntu  6.8% (6)                2.3% (2)                        11.4% (10)    
  10.2% (9)               69.3% (61)
Redhat  28.2% (24)      14.1% (12)              20.0% (17)      12.9% (11)    
  24.7% (21)
Fedora  45.0% (36)      17.5% (14)              21.3% (17)      10.0% (8)     
          6.3% (5)
CentOS  19.8% (17)      11.6% (10)              18.6% (16)      12.8% (11)    
  37.2% (32)
SuSE    72.2% (57)      16.5% (13)              5.1% (4)                2.5% 
(2)                3.8% (3)        
Gentoo  71.3% (57)      17.5% (14)              5.0% (4)                2.5% 
(2)                3.8% (3)        
FreeBSD 59.3% (48)      17.3% (14)              11.1% (9)               3.7% 
(3)                8.6% (7)        
Windows 54.7% (47)      20.9% (18)              7.0% (6)                9.3% 
(8)                8.1% (7)        
Other   65.7% (46)      17.1% (12)              10.0% (7)               4.3% 
(3)                2.9% (2)        

        | Not important | Fairly important | Important  | Very
important | Essential
Debian  | 20.2% (17)    | 8.3% (7)         | 23.8% (20) | 14.3% (12)
  | 33.3% (28)
Ubuntu  | 6.8% (6)      | 2.3% (2)         | 11.4% (10) | 10.2% (9)
  | 69.3% (61)
Redhat  | 28.2% (24)    | 14.1% (12)       | 20.0% (17) | 12.9% (11)
  | 24.7% (21)
Fedora  | 45.0% (36)    | 17.5% (14)       | 21.3% (17) | 10.0% (8)
  | 6.3% (5)
CentOS  | 19.8% (17)    | 11.6% (10)       | 18.6% (16) | 12.8% (11)
  | 37.2% (32)
SuSE    | 72.2% (57)    | 16.5% (13)       | 5.1% (4)   | 2.5% (2)
  | 3.8% (3)
Gentoo  | 71.3% (57)    | 17.5% (14)       | 5.0% (4)   | 2.5% (2)
  | 3.8% (3)
FreeBSD | 59.3% (48)    | 17.3% (14)       | 11.1% (9)  | 3.7% (3)
  | 8.6% (7)
Windows | 54.7% (47)    | 20.9% (18)       | 7.0% (6)   | 9.3% (8)
  | 8.1% (7)
Other   | 65.7% (46)    | 17.1% (12)       | 10.0% (7)  | 4.3% (3)
  | 2.9% (2)

I'm including this table twice, once with spaces and once with tabs.
Text based tables are hard. Hopefully one will look okay for you.

The responder could vote once for each operating system /
distribution, or choose not to at all. What's important from this data
to point out is that Ubuntu, CentOS and Debian, in that order, were
all considered essential by the majority who voted for those lines.
The remainder got a majority vote of "not important." I suspect we're
not seeing a lot of heterogeneous shops out there in our community.

I know the Opscode services team has been working toward adding and
verifying CentOS support in more cookbooks and we've been discussing
how we're going to ramp up our CentOS support in core Chef and in the
packaging. You can expect to see progress in these areas in the near
future.

* Which packaging task would be most meaningful to you?

18.7% "Fat" binary installer that includes Ruby + all deps      
53.8% Releasing Debian packages simultaneously with Gem releases        
27.5% An official Opscode RPM repository        

This is kind of a trick question, because we've been talking about or
actually working on all of these. Generally, there are reasons for
doing so to increase the community, but it is very interesting to see
what current users want.

* Is there anything else that we can do for you?

This was an open format question. I've read all the responses and
appreciate them. Unfortunately the survey as anonymous so I can't help
out the person who wanted us to give them "one million dollarzzz :)"
but I'll work on more "Chocolate chip cookies." =D

The serious answers were pretty much along the lines of what I've
heard and thus expect. We're working every day on tasks like improving
documentation and making Opscode + Chef easier. I'm sure those of you
who have been around for a year or two have noticed the changes.

Thanks to everyone for taking the time to respond.

-- 
Bryan McLellan | opscode | senior systems administrator
(c) 206.607.7108 | (t) @btmspox | (b) http://blog.loftninjas.org