Delete completely the data

[zemich...@gmail.com]

Hi,

We have a old Isilon with various nodes.
And we would like to wipe data definitively, and completely.
So, right now, i've just do a "rm -rf /ifs/*".
And i'm going to do creat a big file with "dd if=/dev/urandom".

But according to you, wha is the best way to do wipe all data, please ?

Thank you in advance.

Regards,
Kanchana

[Jerry Uanino]

Something like http://dban.org/ is what I would usually use for this.... although I'm not sure if DBAN will boot on the modern isilon's.

Netapp used to have a command to do this, which was nice.

The other option is to boot linux from a usb stick and use the "shred" command to nuke each drive.

--
You received this message because you are subscribed to the Google Groups "Isilon Technical User Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to isilon-user-group+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Josh Hampton]

The command you're looking for is:

isi_reformat_node

This will reformat the node you are currently logged into.  You could easily pipe yes to the command and roll it into an isi_for_array as well if you wanted to do all of them at once.  Though likely you'd want to do this via serial connection so that you can watch it go.

[Matt Dey]

I had to do something similar once before. isi_reformat_node will not scrub the drives.  I suggest taking a look at the isi_ovt_check script and the tools in /usr/bin/isi_hwtools.

[Josh Hampton]

isi_reformat_node *will* reformat the disks.  If you need a multi-pass wipe, there are no built-in tools for this on Isilon.  You can however pay for Blanco...

[Dan Pritts]

Josh, when you say it will "reformat" the disks, is that a full pass with zeroes?  I would normally not expect a format to do that, only to recreate the data structures for a fresh filesystem.

Josh Hampton

January 28, 2017 at 10:26 PM

isi_reformat_node *will* reformat the disks.  If you need a multi-pass wipe, there are no built-in tools for this on Isilon.  You can however pay for Blanco...

Matt Dey

January 27, 2017 at 4:38 PM

I had to do something similar once before. isi_reformat_node will not scrub the drives.  I suggest taking a look at the isi_ovt_check script and the tools in /usr/bin/isi_hwtools.

--
Dan Pritts

ICPSR Computing & Network Services

University of Michigan 

[erik.j...@gmail.com]

Using isi_reformat_node doesn't wipe data from the disks by writing zeros to the drive. The main goal of this process is to get a node to a state to let it join a cluster again and perform essentially a quick format of the drives. 

--

Erik Weiman

Sent from my iPhone 6s

--
You received this message because you are subscribed to the Google Groups "Isilon Technical User Group" group.

To unsubscribe from this group and stop receiving emails from it, send an email to isilon-user-gr...@googlegroups.com.

[Michu Michu]

Hi !

So the best way is to boot from a liveCD and do a script to write zeros to the devices listed with "isi devices", isn't it ?

peace!

To unsubscribe from this group and stop receiving emails from it, send an email to isilon-user-group+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

You received this message because you are subscribed to the Google Groups "Isilon Technical User Group" group.

To unsubscribe from this group and stop receiving emails from it, send an email to isilon-user-group+unsubscribe@googlegroups.com.

[Alistair Stewart]

Be careful with this sort of thing. You don't want to erase the boot SSDs on a node or you'll find that you can't boot it up, even after re-imaging with a USB key. Stick to erasing the data drives only. 

Al... 

[Michu Michu]

Thank you for the advice.

I've just want to format the data.

I am not sur what is the best way to do this.

NB:
i) "mount" tell me that the system is on /dev/imdd device

ii) "isi devices" tell me that the nodes are /dev/daXY

Q1 : Is it good to delete just /dev/daXY ?

I deleted files and directories but Isilon told me that there were some taht can not be deleted.

So i deleted the quota on the directories, then it was able to delete the the directories.

Right now, i've still got some directories that i can't delete because there are error : Input/output error

Q2 : How could i delete these error's files, then the directories ?

Thank you in advance.

Peace!

[Josh Hampton]

It does do a full sweep of zero's on all drives.

I've done many of these in the past.  Hundreds even.  If you don't need a secure wipe, this is a reasonably quick way of deleting all data on the cluster.

[Michu Michu]

To resume, writing zero to /dev/daXY devices is the best manner to secure wipe.

The aim is to give Isilon back to Dell (right now) : so it is important to delete all data.

Since i'm new with Isilon storage, i'd like to need some advices.

Does it mean that i must connect to each nodes and umount the NFS share : /ifs, before writing zero to these devices ?

If the NFS is not umount first, you can't write to the device, right ?

Peace

[Chris Pepper]

FWIW, I used dd if=/dev/zero of=/ifs/1T with a count equal to 1tibibyte, then made copies of that file until I ran out of space. That way I could do it all online without breaking the cluster.

Chris

> To unsubscribe from this group and stop receiving emails from it, send an email to isilon-user-gr...@googlegroups.com.

> For more options, visit https://groups.google.com/d/optout.
>
>

> --
> You received this message because you are subscribed to the Google Groups "Isilon Technical User Group" group.

> To unsubscribe from this group and stop receiving emails from it, send an email to isilon-user-gr...@googlegroups.com.

[John Beranek - PA]

If you want truly secure (multi-pass) deletion you'll need to use something like 'shred'

https://www.gnu.org/software/coreutils/manual/html_node/shred-invocation.html

Cheers,

John

[Michu Michu]

I know shred.

But the data are already clean up...

But we still want to do a pass to clean up once time :

- by writing with /dev/zero into a big file

- by writing with /dev/zero to the devices of the node

- by take off the disks from the nodes
- ...

Then, the best pratices ? :-)

--

You received this message because you are subscribed to the Google Groups "Isilon Technical User Group" group.

To unsubscribe from this group and stop receiving emails from it, send an email to isilon-user-group+unsubscribe@googlegroups.com.

[Matt Dey]

I know this is a bit of an old thread and I no longer admin Isilon so this might be out of date but I came across some old notes I had on a process I had to wipe the data of an Isilon node after having evicted it from a cluster.   Here are my notes in case anybody else finds them useful.

After the node was evicted from the cluster  and rebooted back to the wizard I hit Ctrl-C to break out.

 

I then started a screen since I only had console access.  This way once the process was running I could do other things if needed.  I then change directories to /var/crash and made sure it was clean.  I did this to prevent the log directory from getting created somewhere else and filling up / or some other file system while the process was running.  I then ran the test command.

cd /var/crash

/usr/bin/isi_hwtools/./drive_scrub -v -X –W

 

This come back with a message  about messing things up and unmounting /ifs and then killing a bunch of processes which I would then kill.  I would then kick it off again and the scrub would start.  It would create a directory to log progress to and seems like it was writing data to all the drives. 

 

At the end of a pass in the log file for each drive was this summary.

 

#exit: end da2s1e: 38629.1 secs, 24.6652 MB/s

#exit:                    preread         write          read       compare

#exit: Total Bytes:             0  999074796032             0             0

#exit: Scrub Bytes:             0             0             0             0

#exit:               preread    write     read  compare      ecc

#exit:  Total Errs:        0        0        0        0        0

#exit:  Scrub Errs:        0        0        0        0        0

#stat: summary: pass

 

Given that the –X option runs the test in destructive mode, the –W tells it to do writes only, and the summary indicated it didn’t do any pre-reads, I believe I can safely assume that this process wrote garbage to the entire partition which contained IFS thus the drive has been wiped.  The default options also appear to do 2 passes.

To unsubscribe from this group and stop receiving emails from it, send an email to isilon-user-gr...@googlegroups.com.

[Paul Carrington]

I just use DBAN to erase a node fully after eviction. If you want a compliance cert then I think BLANCO should work as its the commercial version of DBAN.

To view this discussion on the web visit https://groups.google.com/d/msgid/isilon-user-group/9d0fa065-1dde-468b-8669-7417d639a518n%40googlegroups.com.