Here at Mirantis, we’re just like you; we look forward to the semi-annual OpenStack summits and all of the new ideas and energy they bring, as well as the opportunity to present our own ideas. Unlike in Vancouver, where we had a remote-controlled robot that enabled our reporting staff who had to stay home to still experience the summit, the logistics of the Tokyo venue meant that the only robot we had was Megarantis printed at the booth’s backdrop, and we were limited to reporting provided by those who were on the ground.
Fortunately, instead of robots we had a glorious team of volunteers among our engineers, so we are pleased to present, “Voices from Tokyo: The white-collar diaries”. Many thanks to Kirill Zaitsev, Evgeniya Shumakher, Alexander Tivelkov, Fabrizio Soppelsa, Alexander Adamov and Boris Bobrov for their contributions.
We brought our own stories to Tokyo, of course, and they can be divided into three groups: comparative studies, such as Mirantis OpenStack 7.0 vs. RHEL OSP7, big data, such as Sparkhara (OpenStack service controllers produce large amounts of log data, and processing these logs can be a time consuming and difficult task), and sessions related to the Community App Catalog, including a number of Murano-related sessions and The Evolution of the Glance API: From v1 to v3, and over the next few weeks you’ll see articles about those ideas here on the blog.
But that’s not what we’re here to talk about today. Today we’re here to give you a view of the rest of the Tokyo summit from some of those who were lucky enough to have been there.
Tuesday, October 27
Even though we’re a global team, the first day was still a bit jarring for most of us; those who’d traveled were a little jet lagged, and those who hadn’t were running on Tokyo time anyway, sometimes after a full day of other work. Why? Because we were looking forward to covering the summit live, in real time — especially the keynotes, which had been livestreamed from Vancouver.
So we were a bit confused to find out that in fact there would be no livestreaming, (though the OpenStack Foundation did make the video available shortly after). Thank heavens for Evgeniya Schumakher, who used Periscope to show us what was happening in Tokyo at the keynotes — as they happened. We were excited to see Mirantis mentioned by two of the keynote speakers
Bitnami’s COO Erica Brescia (see 1h09m of the Keynotes official video) mentioned Murano, of which we’re very proud (having contributed more than 90% of the code by LoC during the Liberty cycle). “I especially liked the Bitnami keynote,” Fabrizio Soppelsa wrote, “especially where they underlined that application developers now want something like click-and-install, instead of wasting time on configuration (which should be left to experts like us).”
But Brescia had a more urgent note, as Evgeniya Schumakher pointed out: “Erica shared her view on what OpenStack needs to be successful.” In her mind, this ability to provide applications that can be deployed with the same ease as adding an app to your phone is crucial to OpenStack adoption. Users, Brescia said, will always take the path of least resistance.
Intel’s VP and General manager of the Open Source Technology Center Imad Sousou (see 1h25m of Keynotes official video) talked about the partnership with Mirantis in the context of Intel’s Cloud for All initiative, which we’re very excited about.
Of course, we’re not all about us. Our experts also noticed the presentation by Lithium’s Lachlan Evenson (see 27m33sec). “I liked the Lithium presentation: the Kubernetes-powered cross-cloud CI/CD solution sounds really familiar and useful,” Alexander Tivelkov mentioned. “Moreover, they know what we are doing in this direction and even used some of our code.”
Evgeniya also found Yahoo! Japan’s talk about the company’s experience in running OpenStack-managed bare metal environment by Simon Chung to be particularly interesting; the company serves more than a billion users (with a b) with more than 64.99 billion page views a month, with 31.9 billion of those to mobile devices, and operates over 100 services.
There was more going on than just the keynotes, and as our experts settled in to Tokyo, they zeroed in on several projects, including OSA (openstack-ansible) (“Looks like a nice thing to try,” Kirill said), the user stories session, and the introduction of Kuryr by IBM’s Mohammad Banikazemi. (“However, sometimes I wasn’t sure if I was attending OpenStack Summit or a DockerCon”, Fabrizio said).
Alexander Tivelkov also had a fruitful discussion of Murano’s roadmap with key external contributors and partners, and Bruce Matthews and Joseph Yep gave an impressive session comparing Mirantis OpenStack and Fuel with Red Hat Enterprise Linux OpenStack Platform and Director, which was fun and very useful.
Still, it was a summit, which meant a time to blow a little bit of steam; Mirantis upped the ante from Vancouver’s air hockey table with a virtual race simulator at the Mirantis booth, and it was a source of real fun for all visitors of the Marketplace. Even various members of the Foundation’s Board Directors enjoyed the race.
Our experts also enjoyed Tokyo itself; Fabrizio went to the Sengakuji Temple and Tokyo imperial palace, but it was night and the garden was closed. But the Gonpachi restaurant was open — a beautiful restaurant, recognized by Tarantino fans as location for the climactic fight between Uma Thurman and Lucy Liu in the Kill Bill films.
And speaking of food, as our volunteers settled in, they also tried out some of the food. Fabrizio had ramen with all the toppings (but wasn’t sure what they were called), Kirill had octopus pieces in rice balls, and Evgeniya zeroed in on the Japanese sweets, including “Matcha sponge cake and some marmalade candies.” (Editor’s note: Evgeniya brings awesome Russian candy when she comes to the US, for which those who receive it are immensely grateful; this woman knows her sweets.)
Wednesday, October 28
On day 2, we settled down to the serious business of project planning and idea exchange. But that doesn’t mean it’s all business.
Kirill: Who’s the best Pony?
“The Murano meetings were pretty productive,” Kirill said. (For more details, see Alexander Tivelkov’s more extensive report, OpenStack Glance, Murano, Community app catalog: links in a chain. ) “Then I’d been to the OpenStack client session — some takeaways are that we should claim the object names ASAP. Glance V3 should only be in OSC. Also, the current ‘murano-’ prefix is a bad thing, and we should fix that.
“Kolla is a nice project, that is going to request a kolla-mesos repo for mesos-specific configs, which fits the Mirantis direction pretty nicely. Mirantis was going to make those things open anyway,” he noted, which isn’t surprising given that Fuel has been using containers for several releases already, “but having been guided by Kolla folks here is a nice thing.”
“The Congress meetings went really nicely, with them agreeing to have a murano-dsvm job. Also, the operator session lead by Chris Aedo was influenced by the murano guys, You can get more details on the session Etherpad.”
Kirill also enjoyed the evening, but not for the reasons you might expect. “The HP-sponsored party was nice. I had severe nostalgia with the cosplayers, and the J-POP group was good, although I guess most of Mirantis guys left right around then.
“Oh, and I finally met someone, whose answer to ‘Who’s the best pony?’ is ‘Fluttershy’, not ‘What the *** are you talking about?’ So I kind of had the time of my life this evening.”
Alexander Tivelkov: the cameraman
“In the morning I went to take some pictures at the ‘Women of Openstack’ session,” Alexander said. “The pictures may be found here. Ladies, you are welcome to find yourself in this set of photos!” Alexander, who at times seems like Mirantis’ unofficial photographer (as well as an engineer) also took pictures of most of the Mirantis team at the summit.
“I spent the day doing serious work on Murano, Glance and the Community App Catalog; you can see my notes in this separate blog post. Finally, there was a great party in the evening: samurai, geisha, cosplayers, cool music, nice food and plenty of drinks. I took some pictures there as well.”
Evgeniya: meetings and more meetings
“I spent most of the day in meetings with partners and on booth duty. It’s great to see how many technology partners are excited about the Fuel Pluggable Framework, the level of automation it makes possible, and its close integration with Fuel. Some partners have already developed and validated their plugins (there’s a list here) while some are only starting out this journey. It was great talking about how the Mirantis Unlocked team supports everyone, and it was interesting to see how many partners were interested in the Mirantis NFV initiative.
“I also got to attend the HP party, which was not only an awesome venue and a great program, but also a terrific opportunity to meet other OpenStackers, and ex-colleagues.”
Fabrizio: Italian man in Tokyo
Adrian Otto presented with his 5-grade son a nice application of running clusters of containers on Rackspace – introducing Carina, but I could have used more talk, rather than Q&A. In practice they want that people test Carina and give feedback — for free.
Project Kuryr is about the joining of forces between Neutron (OpenStack) and libnetwork (Docker), with compliments to Thierry for facilitating collaboration between the communities.
Thursday, October 29
Day three was hard core, with the keynotes over and our volunteers getting into the real work of design sessions intended to determine direction for the next six months of OpenStack development.
Kirill: Mitaka process changes, Searchlight and Kolla
Kirill started by talking about changes to the OpenStack release process itself. Some important takeaways he noted:
- Pre-versioning, in which setup.cfg is used for versioning, is about to be dropped in favor of a tagging system. There is a small issue with that — development versions may run into a problem because Pep 440 (Version Identification and Dependency Specification) only supports pre-versioning — but ultimately the release management team agreed to let this go, as the problem would only affect development versions — and even then, only under very specific circumstances.
- Synced milestone releases, the process where milestone releases such as mitaka-1, mitaka-2, and so on, must fall on the same date, are to be dropped. Projects may still synchronize their pre-releases, but the tagging system makes it optional;
- Adoption of Reno, a Release Notes tool, is targeted for mitaka-1 in early December;
- The Liberty cycle eliminated stable point releases, instead considering every backport to the stable branch to do its own release. Unfortunately, this provides the potential for an unmanageable flood of releases. So rather than do a release every time a change was backported to stable (which could prompt teams to hold back changes until they have a batch) release management agreed to enable “on demand” releases with Reno. Almost every commit to stable should be a release, but this enables greater control.
- Release models, which define the relationship between a project’s releases and the releases of OpenStack in general, are hard and many do not understand them, so it was resolved to work harder at educating the community on such issues.
- Another interesting idea discussed was to drop use of blueprints (use either specs or bugs instead).
“Searchlight seems to be a surprisingly mature project,” Kirill noted, “with a couple of Horizon cores involved. Looks like the project is in a very good shape, and Horizon participation is definitely a plus. We can hope for the UI/UX to be good/consistent there. Some of the challenges include cross-region/cross-cloud search and policy-based access to indexed resources”
Alexander Tivelkov noted, “Searchlight is setting priorities for Mitaka, so the team is investigating the possibility of adding more objects into their index. They are looking to extend Nova support by indexing not just servers, but other nova resources such as flavors. Also, they are gathering opinions on Neutron, Swift and Cinder. They’ve asked us about Searchlight support for Artifacts, and we’ve promised to come back to that as soon as the Glare API (that is, the API for Artifacts) is stable. When they asked about other types of indexable entities, I’ve mentioned Murano Environments: using Elasticsearch for them may good, as it is efficient in searching within large JSON files. The same should apply to deployed Heat stacks, and there was interest in that from somebody from the Heat team.”
As far as Kolla, Kirill said, “Mirantis had very productive sessions there. The biggest takeaway is the new kolla-mesos repository, governed by both Kolla and Mirantis folks.”
Evgeniya’s lucky strike
Another day of booth duty for Evegeniya, but “I won an Apple iWatch from Citrix. Love these guys!”
Alexander Adamov: documentation and security
The Documentation Contributor Guide made by the Mirantis Doc team went down well and will be used as a reference point for all kinds of documentation in the community. Most of my time, though, was spent on security issues.
Boris Bobrov: Keystone
When it comes to Keystone, it was all about deprecations, deprecations, deprecations! PKI tokens are really close to be deprecated in favor of Fernet tokens, the LDAP assignment driver will be either removed or set to read-only mode, and the v2.0 API will be partially deprecated, leaving only authentication-related parts non-deprecated. In addition, running keystone with an eventlet will be removed in this cycle. Boris has put it all together in a blog post about Keystone in Tokyo.
Friday, October 30
Alexander Tivelkov: Last, but not least
It may have been the last day, but the work was still going on. “One of of the most important conversations happened today,” Alexander reported, “during our Murano Contributors’ Meetup. This actually turned out to be a cross-project work session for Murano, Glance and AppCatalog: we worked on defining the borders between each project, agreeing on shared responsibilities and reuse of common functionality. You can see the deep-dive here.”