Hi there, form time to time I do a lot of decommissioning of older ONTAP systems... this involves deleting all configuration (SnapMirror, LIFs, igroups volumes, vservers, aggregates etc.) then "node run..." and disk sanitize....
It all takes awhile 😉 I was just wondering if anyone have found a quick way to do this? Main issue it that we need to have spare disks in order to do the disk sanitize and there are almost always old configuration and volumes that needs to be removed before the aggregates can be deleted... I guess what I am looking for is a "priv set diag;nuke everything" command 😉
I have already written my own little perl script that can give me the disk sanitize start command with all the disk devices from a "aggr status -s" output... There is ofcause a limit to how many disks you can sanitize at one time, which varies from system to system... After all this, there is a lot of booting into maint mode to destroy the boot aggregate, unassign all disks and then boot into menu to init a new system... i typically use the "boot_ontap maint" and "boot_ontap menu" instead of hoping to catch the boot banner and hitting Ctrl-C.. 🙂 (is there a way to do a "halt & boot into menu"? (would save a bit of time) 🙂 I guess what I am looking for is the most optimal way to sanitize a configured system... maybe there are some useful hidden commands that I don't know yet..
#Faster ways to decommission and sanitize a system?
tribal tide
raw gazelle
you can use the bootmenu option 9a to quickly remove all volumes, vservers, lifs and other configuration. it will re-initialize all disks as spares. follow that up with a wipeconfig from boot menu to get rid of everything that might still be stored on the NVRAM card, and a set-defaults afterwards to clear all bootargs.
Then you can start the sanitize process
tribal tide
you can use the bootmenu option 9a to quickly remove all volumes, vservers, lifs...
OK, but you still have to do all the sanitize stuff pretty much manually?
coarse shard
Yes, but in current ONTAP versions you can go into maint mode, assign all disks to one node and sanitize all disks. Or keep them split and have both nodes work. I don’t think it’s any faster as it’s a function of the media not how fast the system is
raw gazelle
if you use on-disk encryption I think you can just reset the encryption key to wipe all the data but I haven't actually tried that yet (and I don't know if it is in any way certified, in case you need that... there were some bugs where not the whole disk was encrypted for example, which I assume would not be good if you have sensitive data on the drives )
sleek anchor
For systems using NSE or NAE then you can just destroy the keys and you're good to go, along with a wipeconfig and set-defaults.
But for older systems, I wouldnt bother with the deletion of LIFs, peering, igroups, etc ... not at all. Run a wipeconfig from the boot menu, then Option 9a as Darkstar says, then assign disks in Maintenance Mode and sanitize.
A node will only sanitize 100 disks at a time, so I typically assign disks 50/50 and get both controllers sanitising at the same time to speed it up. For spinning disks it still takes days, especially with the drives only getting larger.
I wish there was a way to halt a node directly to the boot menu, that'd be a massive timesaver. And also a way to get back to the boot menu from maintenance mode without having to halt the node again. That is frustrating when sitting on the hard floor in a datacentre.
Developers of software never have to deploy it in the real-world on a customer site. We can't always do it from the comfort of our desks 🙂
coarse shard
Why spend the extra time with a wipeconfig if you’re going to use 9a anyway? If you do the process , after issuing 9b and letting it reboot, it will automatically do a wipeconfig anyway