Skip to content

Read RAID logical drive created with HP Smart Array Controller

License

Notifications You must be signed in to change notification settings

ScuroGuardiano/SmartArrayReader

Repository files navigation

HP SMART ARRAY RAID READER

Cover image Wait, since when does P420 have a fan? Since I used a superglue!

I hope you and I will never have to use this program.

Everything was implemented and tested using HP Smart Array P420 with 8.00-0 firmware version. I don't know if it will work correctly for arrays operated with different controller or firmware version.

You had very nice Hardware RAID 6 array operated with HP Smart Array Controller but the controller have died? Now you have to buy new controller and wait 'til it arrives and pray for it to work in order to recover your precious data? NOT ANYMORE. Here it is brand new tool to read smart drive arrays on Linux!

Compiling

You need to install nbd client, at least BUSE documentation says you have to. On Arch Linux it's pacman -S nbd.

Sorry, I can't into make, so to compile you just do

./build.sh

and wait for ages for it to compile.

Usage

First load nbd kernel module:

sudo modprobe nbd

Then use packard-tell to identify logical drives on the array:

./packard-tell /dev/sdb /dev/sdc /dev/sdd /dev/sde

If some of your drives failed you don't have to provide them, packard-tell will figure out which drives are missing. Example:

Packard tell result

Note: If some property returned from packard-tell is followed by ? for example Serial? then that means I am not sure if it's 100% correct. For my case it was correct ^^

After you found logical drive that you want to mount use command provided by packard-tell
It will assemble the array and expose it on /dev/nbd0. This is readonly. Use lsblk to identify if there are any partitions. Now you can mount with mount -o ro, for example:

mount -o ro /dev/nbd0p1 /mnt/d1 --mkdir

Remember to include -o ro, otherwise you will get an error as hewlett-read exposes raid array as readonly.

If you want to use other nbd target then you can use --out option on hewlett-read command, for example --out /dev/nbd1.

If your metadata are destroyed you can use hewlett-read directly, if you don't provide size it will use maximum size possible. It won't read last 32MB from drives tho, coz metadata lays there and I just ignore that section. Example:

./hewlett-read --raid=5 /dev/sdc /dev/sdd /dev/sdf

If one drive has failed replace it's path with X:

./hewlett-read --raid=5 /dev/sdc X /dev/sdf

Of course you have to remember, RAID 0 can't have failed drives, RAID 5 only one, RAID 6 only two*

* For RAID 6 only 1 missing drive recovery works at this moment. - see Raid 6 problem

Features

  • Reading and parsing controller metadata
  • Assembling RAID 0, 1, 5, 6, 10, 50, 60
  • Supports multiple logical drives on single array
  • Supports recovery in case of failed drive in RAID 1, 5 and 6
  • Supports recovery in case of failed drive for each mirror group in RAID 10
  • Supports recovery in case of failed drive for each pairty group in RAID 50 and 60

Caveats

  • Sometimes it doesn't read correctly very end of drive, last full stripe to be precise. I don't really know why, sorry. This shouldn't be a problem until you filled up your RAID array to the very last megabyte.

TODO

Libraries used

BUSE - 💜

Legal

This Project is in no way affiliated with, authorized, maintained, sponsored or endorsed by HP or Hewlett Packard Enterprise or any of its affiliates or subsidiaries.

License

GPL 3.0