Why the Public Cloud Rebooted
-
Over the weekend the majority of the world's public clouds rebooted. Most notably were Rackspace, Amazon and IBM Softlayer. Why? Because of a patch that needed to be applies to Xen, the hypervisor that powers effectively the entire global cloud provisioning space except for the obvious exception of Azure.
At MangoLassi we were impacted by this too, as we run on Rackspace. That was the cause of our weekend blip. But an important security update that needed to be applied very quickly. And the downtime was minimal.
-
How come they don't fail over the VMs to another server, then update the original server move them back and so on as to make it an update with no down time?
-
@thecreativeone91 said:
How come they don't fail over the VMs to another server, then update the original server move them back and so on as to make it an update with no down time?
Cloud is not high availability. That's a very common misconception. Cloud is beneath the HA layer. Anyone who had HA stayed up and running, of course, but cloud itself is not HA.
-
@scottalanmiller said:
@thecreativeone91 said:
How come they don't fail over the VMs to another server, then update the original server move them back and so on as to make it an update with no down time?
Cloud is not high availability. That's a very common misconception. Cloud is beneath the HA layer. Anyone who had HA stayed up and running, of course, but cloud itself is not HA.
They easily could have chosen to do so, but the workload would not offset the cost I am sure.
-
@JaredBusch said:
@scottalanmiller said:
@thecreativeone91 said:
How come they don't fail over the VMs to another server, then update the original server move them back and so on as to make it an update with no down time?
Cloud is not high availability. That's a very common misconception. Cloud is beneath the HA layer. Anyone who had HA stayed up and running, of course, but cloud itself is not HA.
They easily could have chosen to do so, but the workload would not offset the cost I am sure.
Not really, because you can only do so much HA at that layer. Not every workload can be made HA in that way. But more importantly, HA is always an option and most importantly, HA is something you do, not something that you buy. But anyone who had an outage, like us, chose to not pay for HA because it is cheaper to have blips than to have the cost of HA.
It's really the end users who opted out of HA.
