OpenStack Summit Tokyo October 2015
Tuesday Oct 27, 2015
Keynote
Johnathan Bryce (Executive Director, OpenStack Foundation)
Certified OpenStack Administrator course
Egle Sigler (Principal Architect, Rackspace)
OpenStack navigator for tracking and displaying information about various OpenStack components (openstack.org/software)
Lachlan Evenson (Cloud Platform Engineering Team Lead, Lithium Technologies)
container stuff, Crocodile hunter game
Takuya Ito (Sr. Manager of Infra Engineering and OpenStack Blackbelt, Yahoo! Japan)
64.99 billion page views per month
31.9 billion are from mobile devices
270 million apps downloaded
100+ services
running 50000+ instances on OS
20 PB data storage
6 times traffic density compared to physical environment
20+ clusters
they've doubled OS use in the last year
experience massive spikes when certain events (e.g. natural disaster) take place
mission critical workloads are run on OS
it is very important to have consistent APIs (i.e. OpenStack), so that the underlying hardware doesn't matter
Tsugikazu Shibata (OpenStack Foundation Board member, NEC)
Erica Brescia (COO, Bitnami)
How to Banish the Shadow Cloud
provide easy-to-use, complete solutions for users
OpenStack I18n team
Daisy (Ying Chun Guo), Carlos Munoz (Red Hat Zanata Dev Lead), KATO Tomoyuki (Sales Engineer, Fujitsu)
translate code, documentation, community collaboration (wikis, etc.), promotional material (website, emails, etc.)
14 languages translated for dashboard
6 languages translated for docs
10 languages for user messages of 9 projects
Carlos Munoz
Zanata is now the official translation platform for OS
upcoming improvements in glossary translation, per-project permissions, statistics, request management
Kato Tomoyuki
Zanata demo
similar strings are presented by the tool, so you can copy them if useful
Ops Liberty Issues
https://etherpad.openstack.org/p/TYO-ops-major-liberty-issues
Rackspace public cloud is viewed as a benchmark for adoption of new OpenStack versions
nova Liberty release notes are apparently awesome
Kilo release notes were poor; new buggy features were on by default and not mentioned in release notes
conclusion: Release Notes are very important to ops and decisions to upgrades; they should be given closer attention and need more involvement from projects so they are accurate
suggestion that we update Release Notes following mid-cycle meetups (nova has done this once before)
upgrades are a key concern; solid docs for updating each service is very important
we are lacking database migration docs
Docs Cross Project Workshop
Cloud Admin and Admin User Guides
maintain split? combine?
Publish user-task matrix for OpenStack
Networking Guide
this is useful, but people want more
Many projects are BYOW: Bring Your Own Writer
Packaging
Fedora packing will be essentially nightly builds
Automated OpenStack Deployment: A Comparison
Florian Haas (Principal Consultant, hastexo)
the room was full, but this would be worth watching online later
Upgrading OpenStack Without Breaking Everything (Including Neutron!)
Sean Lynn (Principal Engineer, OpenStack DevOps, TWC) and Clayton O'Neill (Principal Engineer, Time Warner Cable)
Automate everything - they use Ansible to drive Puppet
Keep up (upgrade every 6 months)
track AMQP heartbeats
Requirements
API outage is ok
no compute outage
no storage outage
no network outage
Consider time of day when upgrade occurs
test production data from all regions
test network impact
add load to the dev/staging environments
shut down services in such a manner that no Juno services are trying to talk to Kilo services
change to configuration for nova and cinder was not well documented
Juno deprecations were not documented as removed in the Kilo Release Notes
a new Kilo default was not documented in the Release Notes
new features were not always well documented