Matthew Lengenfelder
unread,May 26, 2010, 12:15:15 PM5/26/10Sign in to reply to author
Sign in to forward
You do not have permission to delete messages in this group
Either email addresses are anonymous for this group or you need the view member email addresses permission to view the original message
to bism...@googlegroups.com
I'm thinking about setting up 2 Ubuntu servers, then using DRBD, heartbeat, and iscsi. Those 2 machines will basically be redundant storage for virtual machines, if one goes down, the other one will keep ticking and all the virtual machines stored on both of them will continue to be accessible by the other machines running the virtual machines. Then I'll have 2 or more other machines that are running KVM virtual machines and those virtual machines (web servers, an email server, some windows Terminal Servers, etc). Basicly the idea is that if one of the KVM servers go down, We'll have either a backup server that we can use to replace the downed server and get back up and running in moments (because they're running on iscsi/remote storage, if one machine goes down, it's as easy as telling a different machine to start the virtual machines that the downed machine was running, or perhaps even have some sort of heartbeat/pacemaker type script that manages that automatically.
I think this is a perty decent plan, and some of the virtual machines can be load balanced and whatnot depending upon the type of virtual machine (we won't be load balancing our windows terminal servers, but we could do that with the linux web servers... etc)
I ran across some articles on Ubuntu Enterprise Cloud but they don't really talk about what a cloud really gives you. It appears that there is one master cloud computer that basicly controls the cloud, if it goes down does the whole cloud go down... or perhaps a bigger concern is how is the data on each individual cloud node being backed up. If I'm running a Windows terminal server on one node and that node goes down, does the terminal server continue to run? I'm not sure how a cloud would work without some sort of remote storage solution, unless everything is being replicated to multiple machines, but if that's the case and I have one node that has 1TB of HDD space and another node that has 8TB of HDD space it obviously can't replicate all of the data from the 8TB of storage to the 1TB of storage... do all of the nodes need to be running on identical hardware? If so what happens a year after it's built and 16 core(or whatever) cpu's are out and 4TB HDD's are out and you want to start using the new hardware, do you have to build a second cloud with the better hardware and basically scrap the first? I guess I just have a lot of questions and not much experience about private clouds and why you would choose that over a DRBD/Heartbeat/KVM solution... anyone here played with them or have some good links where some of my questions might get answers, I haven't been having much luck finding them myself.
Matt L.