Today was the last day of the OpenStack Design Summit. It wrapped up with a change of pace this time around, each project had their own contributor meetup which was used to continue hashing out ideas and getting some work done. I think this was a really brilliant move. I was pretty tired by the time Friday rolled around (one of the reasons the later Ubuntu Developer Summits were shrunk to 4 days), so I’m not sure how useful I would have been in more discussion-driven sessions. The contributor meetup allowed us to chat about things we didn’t have time to run sessions on, or do in-person follow-ups to sessions we did have. We also had nice in-person time to collaborate on some things so that some of our projects got to a semi-working state before we all go home and take a vacation (my vacation starts next Thursday).

I spent my day meeting up with with people to talk about our new translations tools and did the first couple drafts of the infrastructure specification to get that project started. Given the timeline, I anticipate that my real work on that won’t really begin until after I return from Jamaica on November 21st, but that seemed to sync up with the timeline of others on the team who are either taking some time off post-summit or have some dependencies blocking their action items.

There was also time spent on talking about the Infrastructure User Manual as a follow up to the session earlier in the week. We decided to host a 48 hour virtual sprint on the first couple days of December in order to collaborate on fleshing out the rest of the document (announcement here). As we all know, I love documentation, so I’m glad to see this coming together. I was also able to have a chat with a contributor later in the day who is also looking forward to seeing it finished so he can build upon it as the foundation for more project-specific developer documentation.

Also, the topic of third party testing came up during one of my chats and was overheard by someone nearby – which is how we learned there were at least three teams talking about creating a more automatic mechanism for determining the health of the third party testing systems. That’s approximately two teams too many. Kurt Taylor was able to get us all on an email thread together so I’m happy to say that a specification for that project should be coming together too.

Late in the afternoon James E. Blair did a demo for developers of gertty. I wrote about the tool back in September (here) and I’m a big fan of CLI-based code review, so it was fun to see others excited and asking questions about it.

As things wound down, I realized that this was probably the best OpenStack summit I’ve attended. The occasional snafu aside (like the over-crowded lunch on Thursday – I ate elsewhere), for a conference with over 4,600 attendees it felt well-managed. The Design Summit itself had a format I was really pleased with, as in addition to having the Friday work day, Tuesday was devoted to much-needed cross-project summit sessions. As OpenStack grows and matures, I’m really happy to see everyone working to fine tune the summits like this to keep pace.

Tonight I joined several of my OpenStack colleagues for an early dinner, retiring early to my room so I could re-pack my suitcase (and hope it’s not over 50lbs) and get some work done before my flight tomorrow morning. As exhausting as this trip was, it sure flew by fast and I am quite sad to be leaving Paris! Alas, my sister in law’s wedding in Philadelphia on Sunday awaits and I’m looking forward to it (and finally seeing my husband again after almost 2 weeks).

As the OpenStack Summit continued for those of us on the development side, Wednesday and Thursday were full of design sessions.

First up for me on Wednesday was a great session about the Infrastructure User Manual led by Anita Kuno. A pile of work went into this while we were at our mid-cycle Infrastructure sprint in July, but many of the patches have since been sitting around. This session worked to make sure we had a shared vision for the manual and to get more core contributors both reviewing patches and submitting content for some of the more complicated, institutional knowledge type sections of the manual. The etherpad for the session is available here.

The session on AFS (Andrew File System) for the Infrastructure team was also on Wednesday. In spite of having a lot of storage space at our disposal and tools like Swift (which we’re slowly moving logs to), there are still some problems we’re seeing to solve that a distributed filesystem would be useful for, enter the AFS cell set up for the OpenStack project. The session went through some of the benefits of using AFS in our environment (such as read-only replicas of volumes, heavy client-side caching support and more comprehensive ACLs than standard Unix filesystem permissions). From there the discussion moved on to how it may be used, some of the popular proposals were our pypi mirror, the git repos and documentation. Detailed Etherpad here.

There were also a couple QA/Infra sessions, including one on Gating Relationships. At the QA/Infra mid-cycle meetup back in July we touched upon some of the possible “over-testing” that may be done when a change in one project really has no potential to impact another project, but we run the tests anyway, using up testing resources. However, there isn’t really any criteria to follow for determining what changes and project combinations should trigger tests, and it was noted that many of what seem like unnecessary testing was actually put in place at one point to address a particular pain point. The main result of this session was to try to develop some of this criteria, even if it’s manual and human-based for now. Detailed Etherpad here.

We also had a QA and CI After Merge session. Currently all of our tests are pre-merge, which makes sure all code that lands in the development repository has undergone all official tests that the OpenStack CI system has to offer. This session discussed whether heavier, less “central” tests to the projects be tested post merge or with periodic tests, with what I believe was some consensus: We do want to split out some of the current gated jobs. Several todo items to move this forward were defined at the bottom of the etherpad.

I also attended the “Stable branches” session (lively etherpad here). Icehouse’s support is 15 months and the goal seems to be to support Juno for a similar time frame. Several representatives from distributions were attending and giving feedback about their own support needs and there seems to be hope that there will be work from folks from distros committing to do some of the maintenance work.

There were also a couple sessions about Tempest, the integration test suite. First there was “Tempest scope in the brave new world” which focused on the questions around what should be in Tempest moving forward and what the project should consider removing as the project moves forward. Etherpad for the session here. There was also a “Tempest-lib moving forward” session, which discussed this library that was created last cycle and various ways to improve it in the coming cycle, details in the Etherpad here.

Wednesday evening I made my way over to the Core Reviewer party put on by HP at the near rooftop event space of Cité de l’Architecture et du Patrimoine. We were driven there by what was described as “iconic, old French cars” which turned out to be the terrifying Citroën 2CV. And our drivers were all INSANE in Paris traffic. Fortunately no one died and it was actually pretty fun (though I was happy to see buses would be taking us back to the conference venue!).

The night itself kicked off with a lecture on the architecture of the Sagrada Família Basílica in Barcelona by one of the people currently working on it, and which drew some loose parallels between our own development work (including the observation that Sagrada Família is not complete – a 140+ year release cycle!). They also brought in entertainment in the form of several opera singers who came in throughout the night. Some food was served, but I spent much of the night outside chatting with various of my OpenStack colleagues and drinking so much Champagne that the outdoor bartender learned to pull out the bottle as soon as he saw me coming. Hah!

My favorite part of the night was the stunning view of the Eiffel Tower. It’s a beautiful thing on its own at night, but at the top of the hour it also sparkles for 5 minutes in a pretty impressive show. I was so caught up in discussions that I didn’t manage to go on the museum tour that was offered, but I heard good things about it today.

Then it was on to today, Thursday! I had a great chat with Steve Weston about the third party dashboard we’re working on before Anita came to find me so I wouldn’t be late for my own session (oops).

My (along with Andreas Jaeger’s who I saved a seat for up front) session was an infrastructure session on Translations Tools. We’re currently using Transifex but we need to move off of it now that they’ve transitioned to a closed source product. As I mentioned in my last post, we decided to go with Zanata so the session was primarily to firm up this decision with the rest of the infrastructure team and answer any questions from everyone involved. I have a lot of work to do during the Kilo cycle to finally get this going, but I’m really excited that all the work I did last cycle in getting demos set up and corralling the right talent for each component has finally culminated in a solid decision and action items for making the move. Next week I’ll start working on the spec for the transition. Etherpad here.

I attended a few other sessions, but the other big infrastructure one today was about Storyboard, the new task and bug tracker being written for the project to replace Launchpad. Michael Krotscheck has been doing an exceptional job on this project and the first decision of the session was whether it was ready for the OpenStack Infrastructure team to move to – yes! The rest of the session was spent outlining the key features that were needed to have really good support for infrastructure and to start supporting StackForge and OpenStack projects. The beautiful Etherpad that Michael created is here.

Tonight I went out with several of my OpenStack colleagues to dinner at La maison de Charly for delicious and stunningly arranged Moroccan food. I managed to get back to my room by 9PM so I could get an early night before the last day of the summit… but of course I got caught up in writing this, checking email and goofing off in IRC.

Tomorrow the summit wraps up with a working day with an open agenda for all the teams, so I’ll be spending my day in the Infra/QA/Release Management room.

Saturday morning I arrived in Paris. The weather was gorgeous and I had a wonderful tourist day visiting some of the key sights of the city. I will write about that once I’m home and can upload all my photos, for now I am going to talk about the first couple of days of the OpenStack Summit, which began on Monday.

Both days kicked off with keynotes. While my work focuses on the infrastructure for the OpenStack project itself and I’m not strictly building components of OpenStack that people are deploying, the keynotes are still an inspiration. Companies from around the world get up on the stage and share how they’re using OpenStack to enable their developers to be more innovative by getting them development environments more quickly or how they’re putting serious production load on them in the processing of big data. This year they had BBVA, BMW (along with a stunning i8 driven onto the stage), Time Warner Cable, CERN, Expedia and Tapjoy get up on stage to share their stories.

CERN’s story was probably my favorite (even if the BMW on stage was shiny and I want one). Like many in my field, I hold a hobbyist level interest in science and could geek out about the work being done at CERN for days. Plus, they’re solving some really exceptional problems around massive amounts of big data produced by the LHC using OpenStack and a pile of other open source software.

Tim Bell of CERN

It was exciting to learn that they’re currently running 4 clusters using the latest release of OpenStack, the largest of which has over 70,000 cores across over 3,000 servers. Pretty serious stuff! He also shared some great links during his talk, including:

• OpenStack in Production blog by CERN
• CERN Operations on Github
• CERN OpenLab

I was also delighted to see Jim Zemlin, Executive Director of the Linux Foundation, get on stage on the first day to share his excitement about the success of OpenStack and to tell us all what we wanted to hear: we’re doing great work for open source and are on the right side of history.

In short, the keynotes spoke to both my professional pride in what we’re all working on and the humanitarian and democratization side of technology that so seriously drew me into the possibilities of open source in the first place.

All the keynotes for both days are already online, you can check them out in this youtube playlist: OpenStack Summit Paris 2014 Keynote Presentations

Back to Monday, I headed over to the other venue to attend a session in the Ops Summit, “Top 10 Pain points from the user survey – how to fix them?” The session began by looking at results from the survey released that day: OpenStack User Survey Insights: November 2014. From that survey, they picked the top-cited issues that operations are having with OpenStack and worked to come up with some concrete issues that the operators could pass along to developers. Much of the discussion ended up focusing on problems with Neutron (including problems with the default configuration) and gaps in Documentation that made it difficult for operators to know that features existed or how to use them. The etherpad for the session goes further into depth about these and other issues raised and added during the session, see it here.

Monday afternoon I met up with Carlos Munoz of Red Hat and Andreas Jaeger of SUSE who I’ve been working with over these past couple of months to do an in depth exploration of our options for a new translations system. We have been evaluating both Pootle and Zanata, and though my preference had been Pootle because of it being written in Python and apparent popularity with other open source projects, the Translations team overwhelmingly preferred Zanata. As Andreas and I went through the Translations Infrastructure we currently have, it was also clear that Zanata was our best option. It was a great meeting, and I’m looking forward to the Translations Tools Session on Thursday at 11AM where we discuss these results with the rest of the Infrastructure team and work out some next steps.

Me, Carlos and Andreas!

From there I went down to the HP Sponsored track where lighting talks were being run during the last two sessions of the day. The room was packed! There were a lot of great presentations which I hope were recorded since I missed the first few. My talk was one of the last, and with a glowing introduction from my boss I gave a 5 minute whirlwind description of elastic-recheck. I fear the jetlag made my talk a bit weaker than I intended, but I was delighted to have 3 separate conversations about elastic-recheck and general failure tracking on CI systems that evening with people from different companies trying to do something similar. My slides are available here: Automated failure aggregation & detection with elastic-recheck slides (pdf).

On Tuesday morning I was up bright and early for the Women of OpenStack breakfast. Waking up with a headache made me tempted to skip it, but I’m glad I didn’t. The event kicked off with some stats from a recent poll of members of the Women of OpenStack LinkedIn group. It was nice to see that 50% of those who responded were OpenStack ATCs (Active Technical Contributor) and many of those who weren’t identified themselves as having other technical roles (not that I don’t value non-technical women in our midst, but the technical ones are My Tribe!).

Following the results summaries, we split into 4 groups to talk about some of the challenges facing us as a minority in the OpenStack community and came up with 4 problems and solutions: Coaching for building confidence, increasing profile and communication for and around the Women of OpenStack group, working to get more women in our community doing public speaking and helping women rejoin the community after a gap in involvement (bonus: this can directly help men too, but more women go through it when taking time off for children). The group decided on focusing on getting the word out about the community for now, seeking to improve our communication mechanisms and see about profiling some women in our community, as well as creating some space where we can put our basic information about what we’re working on and how to contact us. I was really happy with how this session went, kudos to all the amazing women who I got to interact with there, and sorry for being so shy!

After keynotes, I headed back over to the Design Summit venue to attend a couple cross-project testing-focused sessions: “DefCore, RefStack, Interoperability, and Tempest” (etherpad here) and Moving Functional Tests to Projects (etherpad here). One of the most valuable things I got out of these sessions was that projects really need to do a better job of communicating directly with each other. Currently so much is funneled through the Quality Assurance team (and Infrastructure team) because they run the test harness where things fail. Instead, it would be great to see some more direct communication between these projects, and splitting out some of the functional tests may be one way to help socially engineer this.

Following lunch and a quick meeting, I was off to “Changes to our Requirements Management Policy” (etherpad here) and then “Log Rationalization” (etherpad here). There seemed to be more work accomplished on the latter, which was nice to see since there’s a stalled specification up that it would be great to see moved along so that the project can come up with some guidelines for log levels. Operators have been reporting both that they often run logging at DEBUG level all the time so they can see even some of the more basic problems that crop up, AND are frustrated by some “non-issues” being promoted to WARNING and filling their logs with unnecessary stack traces.

Next up was the Gerrit third-party CI discussion session. I wasn’t sure what to expect from this session, but the self-selected group (many were more involved with OpenStack than was assumed, but they did come all the way to the summit…) was much more engaged than I had feared. Talk in the session centered around how to get more third party operators involved with the growing third party community, one suggestion being moving the meeting time to a more European friendly time every other week. There was also discussion around the need for improved documentation and I raised my hand about helping with a more dynamic dashboard for automatically determining the status of third party systems without manual notifications from operators. Etherpad here.

The last session of my very long day was “Translators / I18N team meetup” where the group sought to promote translations to grow the community and recognize translators, etherpad here. As I mentioned earlier, I’m working on some of the new tooling that the team will use, so in spite of only speaking English, I was able to chime in a bit on the technical side of making some of the recognitions and other statistics available once we switch back to an open source platform for translations.

Then it was off to the HP party at Mus̩e des Arts Forains. Open for private events only, the venue hosts a collection of antique/vintage (dating from 1850-1950) games, rides and other fair-related objects. I played a couple of the games and enjoyed snacks and wine throughout the evening. It was certainly busy and some areas were quite loud and crowded, but it was easy to find large areas where the volume was quite conducive to conversations Рof which I had many.

Social events and parties are not really my thing, but this one I really enjoyed. Transportation to the venue included an optional tourguide led tour on the bus past many of the stunning sights of Paris at night. And they began running shuttles back to the conference center at 9PM – which I figured I’d catch then, but it was after 10PM before I made my way back to the bus. I think what I really don’t like are club-like parties with loud music and nothing interesting to occupy myself with when I find myself frequently wandering around solo (apparently I’m a lousy pack animal). The ability to stop and play games, explore the interesting food offerings and run into lots of people I know made the evening fly by.

Huge thanks to my friends and colleagues at HP for putting on such a comfortable and exciting event, this one will be hard to top in my awesome-events-at-conferences ledger.

Tomorrow we begin the hardcore part of the conference for me, kicking off with an Infrastructure session at 9AM and moving through various QA and Infrastructure sessions going on through the rest of the week. Since it’s nearing 1AM, I should get some sleep!