My collaborator is splitting from his institute and wants help moving his application and data off the institute's network. And of course things are moving quickly.
He has 12 TB of image data on a 400 TB SAN, which is addressed over iSCSI.
The application server currently runs in a separate 1U Dell pizza box, runs on Windows Server 2003 for now (plan is to upgrade to 2012 when the app company certifies their product build) and the app officially only supports iSCSI; although a long-time tech rep at the company says he's seen it work with NFS. He says he only needs to run an SQL script to update the paths.
Here's my proposed workflow:
- Setup a zpool (
tank) on FreeBSD 10.2 on a Dell Precision T7400 with 5 6TB disks in either a RAIDZ1 or 4 disks in two mirror vdevs. Let's call this machinebohr - Set up an iSCSI targets with LUNs that point to volumes (
imagesandimages2) - Plug
bohrinto the institute's network (they're fine with that) - Mount the institute's SAN, as an initiator or via SMB
- copy the data to the LUNs
- point the application server at the new LUNs
- Charles (the app company's rep) runs his SQL
- Take
bohrand the app server and set them up at the new site.
The only alternative I thought of was to make the app server do the transfer. It seems like it might actually slow things down, but would also avoid me having to set up bohr as an initiator.
My questions: am I making sense? Does the plan sound like it would work? Am I missing any steps? Does it make sense to also set up bohr to serve the same data over NFS simultaneously (for pulling data for other research uses) or should I require iscsi access for all clients?
This is the safest method.