Analysis of 2019 User Survey Feedback

Introduction

The OpenStack Technical Committee (TC) added questions to the User Survey for the first time in 2019. The following questions were asked by the TC:

  • How do you upgrade your version of OpenStack?

  • Once on a given release, do you use stable branches for bugfix upgrades?

  • To which projects does your organization contribute maintenance resources, such as patches for bug fixes and code reviews on master or stable branches?

  • How do members of your organization contribute to OpenStack?

  • What prevents you or your organization from contributing more maintenance resources, or makes contributing difficult?

  • Other ways to participate?

The intention of these questions was to understand how users are maintaining their OpenStack clouds and how they are interacting with the community. In the case that they weren’t currently interacting with the community it was hoped that the questions would spark thoughts on how they could participate in the future.

In general the TC was pleased with the number of responses they got to the questions and the information provided. As a result it was decided that the TC won’t revise the questions before the next User Survey. The plan is to look at the responses to the next survey and then decide if any questions need to be revised or changed.

Below we will summarize the responses that were received to each of the questions:

How do you upgrade your version of OpenStack?

This is the question to which we got the greatest response. The following were the available responses to choose from and the associated number of responses:

Question

Number of Users

Percentage of Responses

Upgrade all coordinated releases (once every 6 months)

73

22

Skip/Fast-forward releases

91

27

Not Upgrade

119

35

Deploy Regularly from master branch

30

9

Deploy all intermediary releases and all coordinated releases

24

7

So the responses here were spread pretty widely. The number of users who respond that they don’t upgrade explains why there are so many people still running on releases like Mitaka and Newton. The number of users that do skip or fast-forward updates supports the efforts that have gone into making such an upgrade path possible.

For a future survey it would be interesting to ask the users who are not upgrading why that is the case. Would be good to know if this is due to it being seen as too dificult a thing to do or if it is just that the users don’t see a need. At this point, however, the TC doesn’t feel that the current results warrant a revision to the questions.

Once on a given release, do you use stable branches for bugfix upgrades?

This question also had good participation with about two thirds of the users responding to the question. The responses were also pretty evenly spread across four of the five options:

Answer

Users

Percentage of Responses

I do not do bugfix upgrades

40

19

Yes, backporting specific fixes

45

21

Yes, deploying every commit on the stable branch

9

4

Yes, upgrading at various points in time depending on fixes

56

26

Yes, using only official point releases

66

31

Not surprisingly, deploying every commit being the least popular option. The fact that the majority of users reported using the official point releases speaks to the importance of continuing to do stable releases.

Similar to the previous question, it would be interesting to know why the people who do not do bugfix upgrades choose that route. It would also be interesting to correlate if these are the same people who responded that they don’t upgrade at all or if this is a different group of users that go from major release to major release only.

To which projects does your organization contribute maintenance resources such as patches for bug and reviews on master or stable branches?

About one third of users responded to this question. The results were as one would expect with core projects having the greatest participation. The top 5 were as follows:

Project

Users

Nova

45

Neutron

43

Cinder

27

Keystone

26

Glance

25

In discussing these results the TC wanted to know how this compared to the number of users who report they are using the project. To get this information we considered the number of users who reported that they were using the project in production with one exception. Karbor had 4 respondents report they were using it but only 3 in production so the number of respondents that indicated they were testing it was used. Here are the results of that investigation:

Project

Contributors

Users

% Participation

Aodh

12

48

25%

Barbican

9

48

19%

Blazar

3

3

100%

Ceilometer

21

153

14%

Cinder

27

285

9%

Cloudkitty

3

13

23%

Cyborg

2

2

100%

Designate

15

61

25%

Glance

25

296

8%

Heat

18

212

8%

Horizon

17

271

6%

Ironic

21

77

27%

Karbor

4

7

57%

Keystone

26

291

9%

Kolla

24

46

52%

Kuryr

5

7

71%

LOCI

2

5

40%

Magnum

14

48

29%

Manila

9

39

23%

Masakari

1

6

17%

Mistral

9

26

35%

Monasca

5

22

23%

Murano

3

17

18%

Neutron

43

294

15%

Nova

45

297

15%

Octavia

23

66

35%

OpenStack Client

15

192

8%

OpenStack Ansible

24

63

38%

OpenStack Helm

3

13

23%

Panko

9

20

45%

Rally

9

57

16%

Sahara

5

24

21%

Swift

15

141

11%

Tacker

6

8

75%

Trove

4

27

15%

TripleO

9

34

16%

Zaqar

3

13

23%

It is interesting to note how the rate of participation in the core projects is generally lower than other projects. As we don’t have this data from previous surveys we can’t tell if this rate of participation has been consistent over time or if it has changed. It will be worthwhile to continue to look at these numbers in future surveys.

Another interesting thing to note in the results is the fact that users who responded, generally contributed to more than one project. There were a few examples where contribution to only one project was indicated, but this was not the majority case.

How do members of your organization contribute to OpenStack?

More than half of the users responded to this question. For the most part the answers were evenly spread with the exception of submitting bug reports which was the clear winner for participation. Here is the breakdown:

Contribution

Users

Percentage of Responses

Bug reports

123

86

Participate in forum sessions at the summit

70

47

Pariticpate in ops meetups

57

39

Bug fixes on master

54

36

Documentation improvement

49

33

Code review on master

46

31

Participate in PTG sessions

38

26

Backporting bug fixes to stable branches

34

23

Feature design review

33

22

Code review on stable branches

30

20

Sponsor in-person events

30

20

Host third-party jobs downstream

13

9

Contribute resources to run CI jobs upstream

12

8

Keeping in mind that this was a user survey, these results are very interesting. Over one quarter of the users that responded submit bug fixes on master and nearly as many also do code reviews. Many of the users are also taking advantage of the Forum Sessions and Ops Meetups. As with the previous question, it seemed that users who participated indicated participation in multiple ways.

This would seem to support one of the things that we highlight as being unique about our community. We are users and developers collaborating together.

What prevents you or your organization from contributing more maintenance resources, or makes contributing difficult?

This question elicited a response from 19% of the participants. The field was also a free-form field, rather than multiple choice which seems to generally get fewer response.

Of the 69 user responses, the majority of them had to do with a lack of time or human resources. Other responses indicated that they were busy running their data centers, going along with the theme of insufficient time.

There were a few surprising responses with regards to it not being clear how to contribute. Hopefully the user survey got them thinking about other ways to contribute. There has been a good focus in OpenStack on making how to contribute easier to understand both through documentation and through education opportunities. Perhaps there is a need to better socialize these opportunities?

Other ways to participate:

This was another free-form field that only got responses from 1% of participants. There were a few responses that are worth noting as they show other ways that users work to participate.

There were responses indicating that they particpate in OpenStack User Groups. Doing such things is important to keep communication in the community flowing. Another user indicated that they write blogs on how to do things. We talk about documentation as a way to contribute but forget to mention that blogs can also be a great way to contribute and share information. Similarly, another user indicated that they help write troubleshooting guides. Another great way to help the wider community. Finally, one other user indicated that they work with their distributor to communicate and create requirements for future enhancements. This was interesting as it is an indication that we may not directly see the ways that people are participating with the community.

All-in-all these responses continued to support the collaborative nature of the OpenStack community.

Summary

The TC was pleased with the first round of answers we got from the User Survey. We don’t feel a need to change the questions for the next survey. This will give us a chance to see if responses are consistent between surveys or if there appears to be variation. After that round we may choose to refine the questions.

There weren’t any really surprising responses this time around. The collaborative nature of OpenStack Users is aparent in the results. We will want to ensure that we don’t see a decline in those numbers. In the mean time, we should be proud of the unique and diverse nature of the community we have helped to develop.

Additional Resources

For those interested in more details please see the mailing list thread that includes the results that were used to create this analysis. The OpenStack Survey Report also provides a graphical overview of the OpenStack Survey results.