We are going to take down all Public Cloud GPU hosts that are currently in the Beta for a hardware upgrade:
- Add a third GPU for better consolidation,
- Upgrade the watercooling,
- Replace the disks to make the hosts 2017-flavors-ready.
All instances are going to be deleted at the beginning of the maintenance. You can take snapshots of your instances before the maintenance and respawn them when the hosts come back. However, you will need to create a 400GB Ceph volume and boot from that, since the new flavors are local-raid only and smaller than the 2016 ones. Contact us if you need a voucher to pay for the Ceph storage.
The maintenance starts on Thursday (23/3/2017) in the afternoon and the hosts should be back online on Monday (27/3/2017).
Update(s):
Date: 2017-03-28 09:23:22 UTC Everything looks good. You should all be able to see and use the new flavors:
If they don't appear in your 'nova flavor-list', just send me your tenant_id so I can grant you access.
Enjoy!
Date: 2017-03-28 08:58:45 UTC All hosts are installed, performing last-minute manual checks.
Date: 2017-03-28 08:02:50 UTC The installation system is fixed. Launching reinstall of all hosts.
Date: 2017-03-27 15:38:20 UTC The hardware upgrade is done, but we need to update our automatic installation process to take the new partition scheme into account (2017 format).
The hosts should be back online tomorrow.
We'll keep you posted.
Date: 2017-03-23 13:10:16 UTC All instances have been deleted and hosts have been disabled in openstack to prevent new instances creation.
All hosts are going back to prod for hardware upgrade. We'll keep you posted as soon as they come back.