Cluster Stack in Oneiric

Registered by Andres Rodriguez

This session will consist of the following:

 * Continue with the work on the Ubuntu Cluster Stack to complete previously defined work items.

 * Discuss the adoption of the new software versions as there have been several upstream changes, and different distributions are adopting different components.

 * Obtain feature requests and discuss the customization of the cluster stack for Ubuntu.

* Nova Components HA (following discussion from the OpenStack Design Summit)

Blueprint information

Status:
Complete
Approver:
Robbie Williamson
Priority:
Medium
Drafter:
Andres Rodriguez
Direction:
Approved
Assignee:
Ubuntu Server
Definition:
Approved
Series goal:
Accepted for oneiric
Implementation:
Implemented
Milestone target:
milestone icon oneiric-alpha-3
Started by
Andres Rodriguez
Completed by
Dave Walker

Related branches

Sprints

Whiteboard

Work items:
[andreserl] Finish packaging resource-agents: DONE
[andreserl] Finish packaging fence-agents: DONE
[andreserl] Upgrade pacemaker to 1.1.X and submit debian dir to Debian: DONE
[andreserl] Upload fence-agents: DONE
[andreserl] Upload resource-agents: DONE
[andreserl] Keep RHCS for DLM (gfs_controld.pcmk/dlm_controld.pcmk) and gfs2-tools: DONE
[andreserl] Remove installation of resource agents from rgmanager: DONE
[andreserl] Remove installation of fence agents from source: DONE
[andreserl] MIR resource-agents: DONE
[andreserl] Switch dependencies on cluster-agents to resource-agents: DONE
[andreserl] Upgrade Heartbeat 3.0.5: DONE
[andreserl] Upgrade pacemaker to 1.1.5: DONE
[andreserl] Link up with openstack and upstream to compile a list of HA requirements: DONE
[andreserl] MIR fence-agents: POSTPONED
[andreserl] Upgrade DRBD 8.4 if Oneiric Kernel ships it: POSTPONED
[andreserl] Sort out PPAs (ubuntu-ha-maintainers becomes main PPA, all other go away): POSTPONED
[andreserl] Evaluate upgrade path 1 or path 2: DONE

Path 1:
Update RHCS to latest 3.0.X (3.0.17 no longer maintained by upstream): TODO

Path 2:
Evaluate if RHCS should be updated to 3.1 (needs testing for OCFS2/GFS2 with pacemaker+corosync+cman. Heartbeat will no longer be used for GFS2/OCFS): TODO
Upgrade to RHCS 3.1: TODO
Finish packaging gfs2-utils: TODO
Upload gfs2-utils: TODO
MIR gfs2-utils: TODO
Update default Pacemaker configuration to reflect upstream changes: TODO

Upstartify corosync (upstart does not stop correctly stop corosync threads): POSTPONED
Include new RHCS and related packages into natty: POSTPONED
Look into automated deployment with puppet (Simple: VIP/ Advanced: DRBD, CLVM, etc): POSTPONED
Add meta-package/tasksel to install simple cluster and join it: POSTPONED
Document for HA with OCFS2: POSTPONED
Document for CLVM: POSTPONED
Include documentation in Ubuntu Server Guide: POSTPONED

== Agenda ==

 * Current Cluster Stack:
   - RHCS (3.0.12)
     + dlm_controld.pcmk/gfs_controld.pcmk
   - Pacemaker/corosync/heartbeat (1.0.10):
     + RA's handle above.

 * Discuss the adoption of the new software versions as there have been several upstream changes, and different distributions are adopting different components.
   * RHCS (3.1):
     - RHCS (dlm, cman, etc), fence-agents, resource-agents, gfs2-utils
     - gfs-pcmk, dlm-pcmk dropped.
   * Pacemaker 1.1:
     - Pcmk + Corosync + cman (to use dlm_controld/gfs_controld)
   * resource-agents
     - Linux-HA and RHCS RA's merged in a single source.
   * MIR fence-agents
     - Cobbler uses it.

 * DRBD 8.3.X/DRBD 8.4.X
 * Nova Components HA (following discussion from the OpenStack Design Summit)
 * Continue with the work on the Ubuntu Cluster Stack to complete previously defined work items.
 * Obtain feature requests and discuss the customization of the cluster stack for Ubuntu.
 * Cloud HA - Pacemaker Cloud Policy Engine:

 http://www.redhat.com/summit/2011/presentations/summit/whats_new/thursday/dake_th_1130_high_availability_in_the_cloud.pdf

== Session Notes ==

== Actions ==
 * Move forward with OCFS2 as the supported clustered FS
 * Keep RHCS 3.0 for DLM's only.
    - dlm_controld.pcmk/gfs_controld.pcmk
    - keep shipping gfs2-tools (and make sure it's ok)
    - upgrade to latest 3.0.x that contains .pcmk (3.0.17)
 * Package fence-agents/resource-agents
 * Upgrade pacemaker to 1.1
 * Investigate and resolve ARM related issues within OCFS2, test
 * Move from cluster-agents to resouce-agents
 * Link up with openstack and upstream to compile a list of HA requirements
 * DRBD 8.4 if is shipped in mainline kernel Oneiric ships.
 * ***Documentation***
 * Sort out PPAs (ubuntu-ha-maintainers becomes main PPA, all other go away)

(?)

Work Items

Dependency tree

* Blueprints in grey have been implemented.