2

If I were to set up physical servers for running a business critical website, I would absolutely build redundancy into that environment. I would strive for no single point of failure and as a minimum run two servers of each role (two web servers, two database servers etc.). Even if I chose to add a virtualization layer to those servers, I would still strive for redundancy (and in this case also ensure that e.g. databases were distributed on different hosts).

However, I'm uncertain if this way of thinking applies to cloud server offerings based on e.g openstack. First, I could always run two web servers and two database servers but how do I know that my two database servers do not run on the same physical hosts? Second, I'm in the dark when it comes to how open stack operates virtual machines. Is a single VM even bound to a specific host, or is there a physical hardware pool which VMs are shuffled around in?

Should I think of open stack cloud servers as just another way VPS/VM flavour? Should I plan open stack cloud servers just as I plan with physical or plain old virtual servers

5

If you want redundancy you still need to plan for it. Cloud providers (regardless of the platform) still have outages on both the host and datacenter levels.

Usually there is a pool of hosts and you don't have much say over what VM goes where. If a host dies you have a small outage while the VM is moved and reboots. Many providers let you setup anti-affinity rules on your machines to ensure you don't end up with all your web servers on the same host. Bigger ones let you put them in different geographical regions to keep running even when there is a datacenter level outage.

Openstack makes it easier for them to move VMs between hosts so a host failure generally causes less downtime, but doesn't eliminate failures. If high availability is something you need you still need to plan for it.

| improve this answer | |
  • In reply to "If a host dies you have a small outage [...]" : what if a MySQL ran on a cloud instance on this host; would the database here be vulnerable to corruption just the same way as a MySQL which runs on physical hardware is? – sbrattla Feb 14 '16 at 14:37
  • @sbrattla yup. Cloud instance can crash and corrupt files. Or the host can crash and corrupt files. That is why both replication and backups are needed if you want HA. – Grant Feb 14 '16 at 15:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.