Re-formatting nodes

Hello tazatemc ,

Thanks for your question! A reformat of your Isilon nodes will wipe the hard disks and return you to a cluster join prompt with the existing install of OneFS. A reimage of your nodes will do a complete wipe of the drives and reinstall OneFS. A reformat will be the quicker option here if there is no pressing reason for reimage. In most cases, a reimage is only required if:

1. Reformat is failing (which is usually attributed to reason 2, but not noticeable until the command is run)
2. The OneFS install has become corrupted
3. Your node does not boot or is in a boot loop
4. In VERY rare cases, if there is a dual boot drive failure.

When looking at doing a reformat or a reimage, there are a couple of things to consider:

1. Do you need the data on the nodes?
   1. I assume an answer of no to this question since you are looking to reformat/reimage and wipe the nodes.
2. Do you need to wipe and update a test environment?
   1. If this is the case, I would reimage ONE node (your first cluster node/intended node 1) and reformat the rest. This will allow you to build your new cluster on a new revision of OneFS and the other nodes will update to join.

If you have any other questions about reformat and reimage, please let us know!

First, isi_reformat_node requires the OneFS tarball for the version you want to reset to.  If the node is still in the cluster, you can copy this to the local file system of the node before you remove it.  You can then run isi_reformat_node.

This procedure does not securely wipe the drives so it depends on what you want done with the node after you wipe it.  If you're going to add it to another cluster, my suggestion would be get the tarball for the version that corresponds to the destination cluster, run isi_reformat_node, and when it's done add the node to the cluster.  Always try to add new nodes in multiples of your protection scheme (eg, 3 nodes).

If you want to sell or trade in the node you're wiping, this is NOT good enough - the data is recoverable on the drives, although it is not trivial to do so and unless you sell or trade in multiple nodes at once, getting entire pieces of files will be hard or impossible. 

My personal policy is to copy the tarball to the node and then remove it from the cluster.  I then do an isi_reformat_node. Then if the node is leaving my company, I pull all of the cables from the node, attach a serial console, and boot the node to the join prompt.  I abort that, set a dummy hostname (don't forget this!), and then run /usr/bin/isi_hwtools/drive_scrub (two passes with a log).   EMC also offers professional services to securely wipe the drives using a commercial 3rd party tool.

I've never needed to use a USB stick to wipe my nodes.