Lost RAID configuration

Servers

Servers
Information and ideas on Dell PowerEdge rack, tower and blade server solutions.

Lost RAID configuration

This question is answered

Recently got a call whereby a company went to access a (rarely used) server and couldn't access it.  The server was offline and wouldn't boot.  They replaced both power supplies and got it past post but it's not recognizing the RAID configuration.  Here's the details:

  • Dell 2850
  • has add-on RAID card with nothing connected so it must be using on-board RAID (LSI?)
  • This server was rarely used and any IT staff responsible for it have left years ago - so they have no documentation and no idea what the original configuration of the server was
  • Their last good backup is over 2yrs old
  • 6 drives - all showing a healthy status; the last user to be on the server can't remember "for certain" but believes there was only a single drive letter showing (indicating a single RAID, which I can only presume was RAID 5 given there were 6 drives)
  • the CTL+M option is not showing - I check bios and it's not set to RAID - if I change it to RAID then it gives me a big warning on reboot about destroying all data on the drives
  • before contacting me they tried to boot from CD using XP lite - from there they could see all of the drives, but attempts to access any of them resulted in a "drive not formatted" prompt; they assured me that they did not format any of the drives
  • client indicates that they did NOT change any BIOS settings

Given that they know nothing about what the configuration is supposed to be I think the best place to start is to try and determine what the configuration was.  Is there a tool that anyone is aware of that can help analyze this based on reading the drive headers and/or information from the on-board RAID?

If I switch the BIOS to RAID mode will it really destroy data on the drives or will it use the existing header data to recognize the existing RAID configuration?

Any other ideas/input?

Thanks,

Chris

 

 

 

Verified Answer
  • The RAID configuration information is stored on the controller and on the drives.  Any time the configurations do not match, the controller wants you to decide which one it should use.  You always want to use the configuration on the drives in this case.  

    (The RAID battery keeps cache data alive on the RAID memory - it does not store configuration data for the array; configuration data is stored in non-volatile RAM on the controller itself - not the cache DIMM.)  

    Whether you put the drives into this system or another, the process will be the same - you will need to view the config first (View/Add - if there are multiple configurations (NVRAM / Disk), it will ask which you want to see).  Once you like what you see, you then save the config on exit.  Either way, it won't affect anything on the drives.  All we are dealing with here is telling the controller which config to use.  

    It is possible that once you switch to RAID  and accept the scary message, that the controller will recognize and accept the configuration automatically and everything will be fine.  BUT if you see a message about a mismatch, you will need to save the Disk config in View/Add.

All Replies
  • At the point you are at now, you have to presume data loss has occurred and you are at the point of attempted salvage/recovery. There is no guaranteed way to get things back.

    When the integrated RAID controller is turned off and then back on again, it does give that scary warning that data could be lost etc. I have not seen that specific task actually erase data though. And there is no other way of turning the RAID back on without seeing that message.

    Once turned back on you will then get the prompt during POST to enter the PERC 4 BIOS. And there is a chance that the controller will recognize the drive configuration information on the hard drives and you can import that configuration, but don’t hold your breath on that one.

    Once you are in the RAID BIOS you should be able to see all the hard drives, and you can then attempt to retag the array. This is done by building the array, how you presume it was built in the past. So if you think it is a RAID 5, then make a RAID 5, but DO NOT INITIALIZE it.

    Here is a video on  how a RAID 5 is made in the BIOS, again do not initialize the array.

    supportapj.dell.com/.../raid5%20perc4%20bios

    Once the array is made you should see a healthy, online , RAID 5. At that point you want to exit the RAID BIOS and let the server try to boot. See what happens and go from there.

    The additional RAID card you see probably has external ports, for a backup unit or external storage.

    Let us know what you find.

    Thanks.

    Download the Dell Quick Resource Locator app today to access PowerEdge support content on your mobile device! (iOS, Android, Windows)

  • You can safely ignore the scary "data loss" message ... up to a point.  If the drives were set up in SCSI mode (no RAID), and set up in an OS-managed RAID (so there was only one drive letter), then switching to RAID mode in the BIOS would only lead to data loss IF you wrote anything to the drives (all of them).

    If it was originally in RAID mode, then you must switch it to RAID mode.  When you do, the controller will see the configuration on the drives ... you must go to CTRL-M, Configure, View/Add, and choose/save Disk View.  This will import the configuration for use on the controller.

  • Thanks for the prompt replies.  If we move the drives to another 2850 server (there's a second one available in the server room) that has already been set to RAID (vs. SCSI) in the BIOS, will it pick up the configuration from the drives?  Will I need to wipe out the RAID configuration on the 2nd server first (obviously making an updated backup firrst).

    I was also curious if there is some utility/tool that could be run from another boot drive (like XP lite) that may be able to read the drive headers to determine what the old configuration was.

    Also, any recommended tool that may be able to backup each physical drive prior to any of the above steps so that recovery can be done back to the current (albeit broken) state.  Standard backup software is really designed to backup logical drives so I was just curious if anyone knew of a tool that could see the physical drive.  I'm also aware of some RAID recovery tools, but you really need to know the current RAID configuration for those to work (which we don't have).

  • Moving the drives to another computer will make no difference - it will be exactly the same as doing it on this box - you will still need to accept the new/changing config.  Don't afraid of the data loss message - changing between modes will not harm anything.

    CTRL-M, Configure, View/Add is the utility used to read the headers from the drives.  The only other way is to get a hex editor and a JBOD enclosure to inspect the individual disks, and unless you know what you are looking for, this will not mean anything to you (I.e. it will not say "disk 2 of 4 in RAID 5").

    RAID Reconstructor is one such tool (will read and interpret RAID/drive metadata to piece things together) used to recover RAID arrays.

  • Thanks - will try this tomorrow.  I had read a recommendation on another site that suggested removing the drives, switching the mode from SCSI to RAID, and viewing the expected configuration PRIOR to putting the physical drives back in.  As the array information should be stored in the RAID controller's memory (and provided that memory/battery has not failed or been reset) then it should work without the risk of affecting anything on the existing physical drives, correct?

  • The RAID configuration information is stored on the controller and on the drives.  Any time the configurations do not match, the controller wants you to decide which one it should use.  You always want to use the configuration on the drives in this case.  

    (The RAID battery keeps cache data alive on the RAID memory - it does not store configuration data for the array; configuration data is stored in non-volatile RAM on the controller itself - not the cache DIMM.)  

    Whether you put the drives into this system or another, the process will be the same - you will need to view the config first (View/Add - if there are multiple configurations (NVRAM / Disk), it will ask which you want to see).  Once you like what you see, you then save the config on exit.  Either way, it won't affect anything on the drives.  All we are dealing with here is telling the controller which config to use.  

    It is possible that once you switch to RAID  and accept the scary message, that the controller will recognize and accept the configuration automatically and everything will be fine.  BUT if you see a message about a mismatch, you will need to save the Disk config in View/Add.

  • Great.  Thanks again for the very prompt replies.  Your comments fit 100% with what I would expect - just work with HP so wanted to confirm expected behavior on Dell.  I'm still going to see if I can find a bit-level backup utility that can grab physical images of the disks (if possible) before I do anything.  May be paranoid, but better safe than sorry.  Thanks so much.

  • Also you can take out all disks, then you can swtich from SCSI to RAID with no fear. Then power off and re install all disks and power on. With this procedure  no data will be lost.

    Let us know!

  • There's already nothing to fear ... the data loss message can only cause data loss if you perform certain disk operations (like writing or initializing) in a different mode other than the one used to install the OS/data.