In my recent talks with Jamie, he gave me some info about the actual mini roadmap of new great features:

1 - Better Xen HVM support, including installing from a CD into an empty Xen instance

2 - XenSource support

3 - Automatic failover

4 - Improved monitoring for specific services being down on virtual systems

The features will be available in the next 4.4/4.5 releases, not necessarily by this order.

In my opinion, the roadmap order should be 3,2,1,4 :smiley:

How are you expecting the automatic failure to work? Even with shared storage, you have to be extremely careful or else its pretty easy to corrupt the guests file system.

well, how does all the others do it? :wink:

If there’s a shared NAS / SAN connected to 2x CloudMin servers, then the Cloudmin servers could very easily do automatic fail over without any service interruptions.

My suggestion:

Let the CloudMin sync the config files between the 2 servers, then run a watchdog or Linux HA script to check and see if the active server goes down, then start-up the instance on the other server. This won’t give 100% uptime though since most VPS’s will take a minute or 2 to boot-up.

But, how does other companies do this? Maybe my suggestion isn’t the best, there could be other options which are better.

well, how does all the others do it? :wink:

They don’t! Very few cloud computing products offer automatic failover…because it is a problem that pretty much always requires participation of your applications, in order to avoid data loss.

I’m pretty sure failover of virtual machines is not coming in 4.4 or 4.5. That’s a huge bunch of problems to be solved.

Jamie may have been speaking of failover of the Cloudmin management server, which we recently discussed, and this one is much more addressable by us, because the application in question that has to protect data is ours. Since we can’t control your applications, failing over for your apps will never be something we can fully automate (though we can provide tools to make it easier, and we will).

VMware does this very well. Download the 60 day trail and check it out :slight_smile:

Vmware’s solution to this is called VMotion. Two host servers connected to san running at the same time.

Resources are dynamically allocated between the two hosts. When an outage occurs, the virtual instances (hosts) are transfered to the other functioning server seamlessly.

VMotion is used for other things as well, such as server maintenance. Admins can now perform hardware upgrades and maintenance durring the day without effect to the end users.

VMware did a really cool demonstration video on this showing an end user playing solitare on a virtual desktop while being “vmotioned” to another host. Seamless. (okay, so it was solitare, maybe not THAT cool…)

Citrix XenServer offers the same functions within their HA Enterprise version of XenServer.