On 10/30/2015 4:29 AM, Andreas Jaeger wrote:
Maybe interesting for some of you as well, Andreas -------- Forwarded Message -------- Subject: [cloud-devel] Summit Report - Dirk Date: Fri, 30 Oct 2015 17:46:06 +0900 From: Dirk Müller <dmuel...@suse.com> To: cloud-devel list <cloud-de...@suse.de> Hi, For the last week I’ve been at the OpenStack Summit in Tokyo. If there is one major topic then it probably is Docker, Kubernetes and the related projects on the OpenStack side. For the first time Neutron was the most popular project rather than Nova, and those sessions were generally the most crowded ones. NFV, SDN are therefore also strong themes. General impression is that given this happened in multiple hotels spread across a campus it was much less crowded than the summits before, and usually sessions were not so full that one couldn’t get in. That said of course there were session where it was just impossible to get in. I don’t know the overall attendee number but it was probably not lower than Vancouver. Surprisingly when they asked in the keynote about first time attendees vs first time japan visitors, there were many more hands raised for first time japan visitors (like the majority of the room). Basically it was very hard to walk around without stumbling over a Red Hat employee. I was told over 250 employees attended this summit, which felt higher than IBM. My impression was that attendance from HP was lower than the years before. The HP cloud announcement probably was causing that to some extend, as well as Rackspace raising their attendance level quite a bit. Maybe it was also because RH employees got asked to wear either RDO or Red Hat corporate branding shirts or hats all the time. I did not have any SUSE branding with me since I had to save the shirts from last year’s susecon for the susecon attendance next week. City and people are absolutely amazing, I think this got the 2nd spot on the ranking of openstack summit locations just after vancouver. I“ve had a couple of meetings related to Ruler of the Stack challenge. The talk was reasonably well attended for a small room (maybe 50 - 60 people). Funny side stories are that Canonical seemed to have spent some man weeks on preparing for the challenge but then only realized two days beforehand that the competition will not happen this time. Similarly there was a private attendee from Netherlands attending the Summit just in order to „kick my butt“. He booked the flights and took holidays just in order to win the challenge on his private expenses, and he bought 8 expensive Extremespeed USB drives preloaded with an image that he built by himself (own bootloader, own OS, own package installer and so on, based on RHEL7). From what he told me, he would probably have won assuming he would have been able to comply to the rules since in a test run on his own hardware he managed to install in 8 minutes or so. We also had a long session with Michael from intel on defining the future rules for the competition and got a lot of good ideas out of that session. I’ve not been attending many of the general track sessions since they’re all recorded anyway. There was one about the release process changes for OpenStack Mitaka that is probably a good time spent if you don’t know about it already (done by TTX and dh). There were a couple of good design sessions. In general my impression it is that the decision less crazy than it was before, which is a good thing. Particularly worth mentioning was the session about Live Upgrade which was eye-opening to me. it looks like we’ll have plenty of fun to do that when we’re skipping a release since nova’s implementation is depending on customer to upgrade from N to N+1 to N+2. skipping a release is unsupported. Tom and I discussed probably a few hours about that during the summit already.
This is due to online data migrations which are a key part of live upgrade. Dan Smith has an excellent series of blog posts on that starting here [1].
Another interesting one was about keystone. There were btw several keystone in big deployments/multi site talks that are very worth watching. Major upcoming change is deprecation of v2 (which was expected) and deprecation of PKI, since it has an unfixable severe security bug. We should probably go switching to other backends immediately, even for older products. It looks like they might move off eventlet removal for another cycle although there is ongoing work to integrate with nginx/wsgi, which seems to be a slightly saner choice than apache2 as wsgi container. The reason on why they’re not removing it is because eventlet is the easiest server to use for dockerized openstack services. Besides that I spent a lot of time on the „OpenStack RPM Packaging“ group that I’m currently PTLing. We’ve met with Haikel and Igor as well as a couple of people from TripleO and Red Hat. Today I spent most of the day helping Mirantis to fix their packaging so that they would be able to switch to our common packaging. Luckily we got our first spec file merged (after months of little to no progress) and started on gating implementations. The fishbowl session that I ran was well attended. Also from people other than RH and Mirantis, and they’re all interested in being able to have just RPM packaging that works and where they can throw in their own downstream patches. After a couple of hours talking privately with the RH guys it seems more clear though on why they’re not paying as much attention on this as we’d have hoped though. Summary is since basically everyone uses RDO anyway, they’re pursuing again the opportunity of upstreaming RDO packages under the OpenStack Big Tent, making the „Common Packaging“ project a competing project without any reason for existance anymore (well, other for having packages that work on SUSE of course). The market place was very very big this time and very crowded, but very boring. The same thing as usual happened: every 2nd if not more booth had Ubuntu and RHEL partner logos and stickers and whitepapers and merchandising crap from Red Hat and Ubuntu. SUSE was nonexistant at the other booths. There was only one booth that I found with a SUSE logo and that was Cavium (which was because we’re working closely together with them on SUSE on ARM). Our SUSE booth was apparently well attended and Aya, Peter, Ryo and more did an awesome job of running it. Particularly worrisome was the density of NFV/SDN vendors that partner with Red Hat. It feels like we already lost the Cloud/NFV marked due to our inability to position ourselves there. Same is true for the Neutron 3rd party integration like the load balancers and related. I had also a good meeting with Andy over dinner and he basically confirmed that worst dreams could come true regarding Huawei. Red Hat basically gave them the blanket check that without any questions asked they will match whatever we offer at the same price or lower. Andreas Poeschl’s talk was mentioning SUSE positively. There were several crapfest talks about Mirantis vs RHOSP or Juju against the rest or Chef vs Ansible or Salt vs Puppet and the like. if you like those kind of talks they’re interesting. I was in the Mirantis vs RHOSP talk and I just loved how the last row lost the red hats (which was filled by Red Hat employees wearing at the beginning of the talk their red hats, but with the middle and end not anymore, but rather looking calm and quiet and trying to not be noticed anymore). Very exhausted about to leave tokyo tomorrow morning. Greetings, Dirk
[1] http://www.danplanet.com/blog/2015/10/05/upgrades-in-nova-the-details/ -- Thanks, Matt Riedemann __________________________________________________________________________ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
