All the services will be down during this period of time and the following actions will be performed:
- virtual machine servers will be upgraded to Debian 64bits
- heartbeat will be upgraded to peacemaker
- the shared file system (ocfs2) will be removed and replaced with ext3 partitions mounted on demand by the HA system
During the last months we have been suffering oom_killer errors due to poor performance of the ocfs2 partition we have been using to store the virtual machines. During next window maintenance we will get rid of ocfs2 and will move the information to ext3 partitions that will be mounted on demand by the HA software. The result of these changes will be a more stable platform which will involve a better services performance.
Find below a brief schedule of the action we will perform during this weekend:
- Friday at 18:00h the access will be disabled, we will disconnect the firewalls and start creating full backups.
- By Saturday morning the "big" changes will be applied and the Xen servers will be using peacemaker with Debian 64bits and ext3 partitions.
- During Saturday afternoon the SAs will test the platform, performing stress tests to ensure we got rid of the oom_killer errors
- By midnight Saturday the platform will be up and running again
- During Sunday we will keep an eye on the performance. We foresee everything will be ready by this time but the platform may have small network outages whether we find something misconfigured.
Again, thank you for your patience.
The OSOR team.