[Raion]

Exercise never worked for me to zero out any data on the disk and takes way too long to actually do anything and in general I think it actually hurts the longevity of our drives. 

Instead, once you boot into the miniroot, go do admin, mkfs. Run mkfs. That will erase everything in the file system. The issue is if you run FX on a disk, I think it just modifies the partition table and doesn't actually format the underlying partitions so what happens is the miniroot sees that there's a valid file system and just uses it. 

Not sure if this works for older systems though.

[weblacky]

I'm going to play devil's advocate and say the "exercise wiping method" is the BEST method I was ever told and works every time for me. Now granted, I'm using like 4-6GB (MAX) for most SGI systems. It has always been FASTER then a low-level SCSI format. I've found no other way to erase a disklabel without it (since the disklabel is throughout the drive...not at the start of the volume, like MBR or GPT).

This ensures a wipe on the disklabel (disk-wide), the miniroot, and everything. This works great because you don't get miniroot merges/collisions as well.

I do the same thing as this: https://forums.irixnet.org/thread-2027-p...l#pid13397

If you were doing a 18GB or larger disk, then I'd agree with you. For a 4GB disk..it's like 25 mins...works every time.

Doing MKFS on admin on inst never resulted in a problem-free install more than a blank disk, for me.

[jan-jaap]

Same here. In addition to the post weblacky linked, I've also been bitten at least twice over the years trying to do a fresh installation on a disk that was developing media errors. Once you've wasted an evening chasing mysterious hangs in a newly installed system, you will make sure the hardware is sane *before* you get started. I also have a habit of running diagnostics tests every time an unknown system lands in my lap or I piece something together from parts.

So for me this is not just about erasing information, it's about establishing trust in the hardware before I invest my time.

NB: there are different methods of exercising a  disk. I imagine that a butterfly test (random I/O) stresses the hardware, but seriously, a disk that doesn't survive a sequential wipe should be binned.

[weblacky]

Same here. In addition to the post weblacky linked, I've also been bitten at least twice over the years trying to do a fresh installation on a disk that was developing media errors. Once you've wasted an evening chasing mysterious hangs in a newly installed system, you will make sure the hardware is sane *before* you get started. I also have a habit of running diagnostics tests every time an unknown system lands in my lap or I piece something together from parts.

So for me this is not just about erasing information, it's about establishing trust in the hardware before I invest my time.

NB: there are different methods of exercising a  disk. I imagine that a butterfly test (random I/O) stresses the hardware, but seriously, a disk that doesn't survive a sequential wipe should be binned.

I'm reusing a lot of my disks, but you bring up a great point, trust.  Doing the exercise (successfully) does give you confidence the disk is good and you're fine to load an OS on it.  Good point, that hadn't really occurred to me (in my situation).  If you are encountering this disk drive for a first time...you need to do some kind of realistic test on it.

Granted, old SCSI disks don't use SMART, so writing doesn't cause reallocation for bad sectors, but it's something.  I think for me, for first encounter with an unknown SCSI hard drive, I'd do a low-level format to update the defect list (but you can use a PC SCSI card util for that).  After that, I'm more of a wipe the drive guy so I know it's blank due to oddities I've seen in irix installations.

[Raion]

I'm using drives that are 70 plus gigs usually.

It takes forever even doing the sequential and I've never actually had it erase data. Maybe I'm not doing something right but yeah, how I would do it if I could fix the miniroot would be to have a /dev/zero type device that can just sequentially blast zeros and then mkfs.

[weblacky]

I'm using drives that are 70 plus gigs usually.

It takes forever even doing the sequential and I've never actually had it erase data. Maybe I'm not doing something right but yeah, how I would do it if I could fix the miniroot would be to have a /dev/zero type device that can just sequentially blast zeros and then mkfs.

 I'm mostly using older SGIs that you cannot really place a SCA 80 adapter in...so 50-pin, so small.  I agree, I'd not want to wait for a 70GB+ drive to wipe.

[chulofiasco]

It took me 2.5 hours to sequentially write a 300gb disk on an Octane.

[weblacky]

I'm using drives that are 70 plus gigs usually.

It takes forever even doing the sequential and I've never actually had it erase data. Maybe I'm not doing something right but yeah, how I would do it if I could fix the miniroot would be to have a /dev/zero type device that can just sequentially blast zeros and then mkfs.

I will say that I assume you know you can control where the exercise happens.  So you could read the section boundaries that the miniroot occupies and exercise the sequenced write in just that range.  As long you're not merging or hitting old data, you don't NEED to remove the disklabel.  Wiping the miniroot AND mkfs'ing the old XFS partition should be enough...I've just had miniroot issues without wiping...I think...never really looked into it once wiping worked.

[jan-jaap]

It took me 2.5 hours to sequentially write a 300gb disk on an Octane.

Sounds about right. The SCSI bus on an Octane has a theoretical max of 40MB/s but you never reach that; ~ 35MB/s is to be expected (with a disk capable of that). 35MB/s equals roughly 2GB/min, so 150minutes for a 300GB disk. All as expected, it seems.

I'm using drives that are 70 plus gigs usually.

It takes forever even doing the sequential and I've never actually had it erase data. Maybe I'm not doing something right but yeah, how I would do it if I could fix the miniroot would be to have a /dev/zero type device that can just sequentially blast zeros and then mkfs.

What you want to do (from the 'fx' prompt) is: 'exe'rcise > 'seq'uential > wr-o. Then select 0 as the start and <whatever big number fx suggests> as the end address, and you will overwrite the entire disk. Exactly what /dev/zero would do and only limited by the hardware. Stay away from random or butterfly tests. If you do a write-compare test, it will write a block, then read it back to verify. That means a reasonable amount of head movement as well.

Before you leave 'fx' you will have to create a new disk label on your freshly wiped disk.

Low-level formatting should not be necessary unless maybe if you suspect hardware failure. Never to get rid of data. It takes a long time and if for whatever reason it's interrupted the disk is usually bricked.

[Raion]

I never suggested low-level formatting. I said mkfs, which takes a minute at most. Exercise sequential still would take far too long and IME, if it's not completed it doesn't do anything. Maybe I did it wrong, but I wouldn't recommend it.