I'm really new at this, so bear with me a moment please? I'm not really a hardware-focused guy -- I'm more of an Ubuntu fan and a LAMP Developer. However, I am working with some hardware guys in our data center and I was the one who pointed them to Ubuntu Enterprise Cloud (UEC). I need advice on a recommended storage configuration to tell them to arrange. Right now we're reading about UEC and making plans.
What we want to do is use the private cloud platform to serve up thousands of WordPress blogs and other LAMP sites. We want to know that if we get a server load spike, that it can be distributed across the cloud. We want to know that if a node shows signs of flaking out, or a VM inside a node shows signs of flaking, that we can migrate the node to another server or the VM to another node. We want to avoid 100% data loss in the event of a node being powered off (as in a crashed server).
To that end, what is the recommended storage configuration? For instance, we have available to us a NAS, NFS, or SCSI attached disk in a RAID-10 arrangement and don't know which is best for what we want to do. Also, I need to know where this gets mounted such that the nodes and VMs can see it or any other suggestions you might have.
Hello,
Eucalyptus has 2 different storages: walrus, which is an s3-like storage, and storage controller (sc) which serves EBS volumes to a cluster (which will show up as availability zones).
In the current model, when an instance is terminated, you loose all your data in that instance: to keep data persistent you either put it into walrus or you create a volume with the sc and you work off there. The latter is probably more familiar since it shows up in the instance as a block device, while walrus follows a get-put model (that is you cannot mount it). Just noticed that a crash of the node, before unmounting an EBS volume, could (not very likely but it could happen) corrupt the file system on the volume (the same way that a hard crash on a node, can corrupt a file system).
Walrus and sc use whatever storage is given to them (you configured where the storage is going to be) so it's up to you to make that storage as reliable as possible (and as backup-ed up as you care about your data). The open source version uses AOE to make the volumes available to the node, but the physical data will still be sitting on the sc, so you need to make only the sc storage safe.
If I misunderstood or not answered your questions, feel free to post more questions. And keep us posted on your progress with Eucalyputs!
cheers
graziano