Jump to: navigation, search

Difference between revisions of "ProductTeam/User Stories/HA VMs"

(History: solutions aren't just on github)
(redirect to new Storyboard story)
 
(2 intermediate revisions by 2 users not shown)
Line 1: Line 1:
 
==  High Availability for VMs ==
 
==  High Availability for VMs ==
  
=== User story ===
+
This page has been superseded by https://storyboard.openstack.org/#!/story/2002124
 
 
The [http://specs.openstack.org/openstack/openstack-user-stories/user-stories/proposed/ha_vm.html official user story] explains the high-level motivation and requirements for this feature.
 
 
 
=== Current work ===
 
 
 
In Austin the following key points were agreed:
 
 
 
* All stakeholders wanted to converge on a unified upstream solution.
 
* That solution should consist of modular components with clearly-defined interfaces. This will give a clean separation of concerns, and facilitate gradual transition of all existing vendor-specific implementations to a unified upstream one, by incrementally replacing custom code with upstream components.
 
** These modular components are described in  [http://etherpad.openstack.org/p/newton-instance-ha this etherpad].
 
* Specs should be written for each component, to ensure consensus on the feature sets and high-level implementation plan.
 
 
 
This work has started, but is still in its early stages.
 
 
 
=== History ===
 
 
 
Initially, various factions within the OpenStack community built solutions mostly independently from each other.  The architectural challenges and several of these solutions were presented in [https://www.openstack.org/videos/video/high-availability-for-pets-and-hypervisors-state-of-the-nation a talk at the Austin summit]. 
 
 
 
The code for three of these solutions can be found online:
 
 
 
* [https://github.com/gryf/mistral-evacuate the mistral-based POC] (Intel)
 
* [https://launchpad.net/masakari masakari] as used by NTT
 
* [http://aspiers.github.io/openstack-summit-2016-austin-compute-ha/#/ocf-pros-cons OCF RAs], as used by Red Hat and SUSE
 
 
 
 
 
However, around the time of the [[Mitaka Release Schedule|Mitaka]] Design Summit in Tokyo, an OpenStack HA community was formed with [[Meetings/HATeamMeeting|weekly IRC meetings]] ([http://eavesdrop.openstack.org/meetings/ha/ minutes of all HA meetings are available]). [https://etherpad.openstack.org/p/automatic-evacuation All known existing approaches were summarised and compared], and discussion on a best-of-breed solution commenced.
 
 
 
Since then, the community has been working towards converging on a single solution.  In Austin, many long meetings were held, and minuted in [http://etherpad.openstack.org/p/newton-instance-ha this etherpad].
 
 
 
Concurrently, the [https://wiki.openstack.org/wiki/ProductTeam OpenStack Product Working Group (PWG)] identified this feature as missing functionality in OpenStack which should be addressed with high priority.  As a result the user story was co-authored by the PWG and members of the OpenStack HA developer community.
 
 
 
=== Meeting Information ===
 
See [[Meetings/HATeamMeeting]]
 
 
 
=== Team Members ===
 
{| class="wikitable sortable"
 
|-
 
! Name !! Company !! IRC Handle
 
|-
 
| Pete Chadwick (user story owner within PWG) || SUSE || pchadwick
 
|-
 
| Adam Spiers || SUSE || aspiers
 
|-
 
| Andrew Beekhof || Red Hat || beekhof
 
|-
 
| Dawid Deja || Intel || ddeja
 
|-
 
| Sampath Priyankara || NTT || samP
 
|}
 
 
 
=== Additional Artefacts ===
 
 
 
* [http://lists.openstack.org/cgi-bin/mailman/listinfo/product-wg Subscribe to the PWG Mailing List]
 
* [http://lists.openstack.org/pipermail/product-wg/ PWG Mailing List Archives]
 

Latest revision as of 23:09, 30 May 2018

High Availability for VMs

This page has been superseded by https://storyboard.openstack.org/#!/story/2002124