Recovery backup strategy

Topic

New Reply

I have written a generic Disaster Recovery Plan for our Company Servers (2 X Server 2008R2). This involves a Bare Metal Restore to dissimilar hardware (Recovery Server) after a complete loss of a Server (Hardware fault etc.), after having tested this all seems OK. Now stuck on how to recover back to original (better hardware).
Currently Servers have a scheduled Full +BMR +System State back up every night using Windows Server Backup.

1^st thought was to do a full backup and then BMR back onto original hardware, but this would take somewhere in the region of 24hrs with no data – Reverse Recovery Plan!
2^nd thought was to mirror data from Recovery Server back to original hardware and then turn off Recovery Server – Which software to use?
3^rd thought was to do a full back up on Recovery Server after BMR followed by Differential backups thereafter thus I would only need two backups to restore when original server repaired.

[*]
:huh:

Any thoughts?

Kev

Reply | Quote

Replies

The way I do server config to allow easy recovery.
1. Use RAID disks. These can be removed from dead hardware and installed as-is in new hardware from the same manufacturer.
2. Create a RAID 1 array for the OS and programs.
3. Use the rest as RAID 5 for data.

With this plan you can do a BMR in a couple of hours, then plug the data disks in and you’re away.

cheers, Paul

Reply | Quote

The way I do server config to allow easy recovery.
1. Use RAID disks. These can be removed from dead hardware and installed as-is in new hardware from the same manufacturer.
2. Create a RAID 1 array for the OS and programs.
3. Use the rest as RAID 5 for data.

With this plan you can do a BMR in a couple of hours, then plug the data disks in and you’re away.

cheers, Paul

Thanks Paul

1. Already using RAID disks.
2. OS and Programs on RAID 1 array.
3. Data on RAID 5 array.

Unfortunately we cannot afford to have additional new hardware to same spec as original due to financial constraints. Recovery Server is a medium spec core i7 machine (16Gb RAM) with 2 x 2TB HDDs 1 for OS and another for data, which is ok on a temporary basis.

Kev

Reply | Quote

Ouch!
Set up the recovery server as a copy of the original and mirror the data, preferably overnight, using Robocopy on the recovery server.
To use the recovery server, rename it as the original server so users don’t have to re-map – don’t forget to turn off Robocopy.

cheers, Paul

Reply | Quote

Ouch!
Set up the recovery server as a copy of the original and mirror the data, preferably overnight, using Robocopy on the recovery server.
To use the recovery server, rename it as the original server so users don’t have to re-map – don’t forget to turn off Robocopy.

cheers, Paul

Thanks Paul,

I found the link to Robocopy quite enlightening, couple of questions though (Never used Robocopy before!).
1. I assume that REM before each line stops that command from running?
2. In the setlocal area what would need to be changed for a different Server: the portion after the equals sign and what format would it need to be in?
3. Could I run the Alternate command first, and then when finished REM that out and then use step 4 to complete the data Migration?
4. How could I tell if the server has any offline files?
5. If I have, how do they get copied across to new Server.
Sorry for all the questions but trying to expand my knowledge and understand the processes involved.
Kev

Reply | Quote

1. REM = remark, or human readable bit that the machine ignores.

2. Yes, after the equal sign. You can use any valid path, e.g. \server1share1folder1subfolder1

3. Migration is really a 2 step process. 1. copy all the data, which takes a long time. 2. copy files that have been updated post copy. The other steps, like security, are to allow you to fix all the broken permissions you will encounter during the migration – I’ve never had a clean migration and I’ve done lots.
Perform the migration steps one at a time and peruse the error log after each step. Fix any errors and run that step again until all is well.

4. If you don’t know, you don’t have any – this is for archive programs like Symantec Enterprise Vault.

5. Restore from the archive program to the new server.

cheers, Paul

Reply | Quote

1. REM = remark, or human readable bit that the machine ignores.

2. Yes, after the equal sign. You can use any valid path, e.g. \server1share1folder1subfolder1

3. Migration is really a 2 step process. 1. copy all the data, which takes a long time. 2. copy files that have been updated post copy. The other steps, like security, are to allow you to fix all the broken permissions you will encounter during the migration – I’ve never had a clean migration and I’ve done lots.
Perform the migration steps one at a time and peruse the error log after each step. Fix any errors and run that step again until all is well.

4. If you don’t know, you don’t have any – this is for archive programs like Symantec Enterprise Vault.

5. Restore from the archive program to the new server.

cheers, Paul

Thanks Paul

I have set up a small test LAN and used the Alternate Command to copy across the data initially from one server to another, only problem is the date modified for all Folders are not same as original but the date on which they were created? All files have the correct date modified!

Will running step 4 sort that out or have I missed something?

Kev

Reply | Quote

Folder creation is one of those tricky things that Robocopy doesn’t do well, but shouldn’t matter as you only access the files.

Did you check the error log? Did everything copy successfully?

cheers, Paul

Reply | Quote

This is a slightly different strategy that you may not have considered, or you may have already rejected. If you install Hyper-V (or an alternative hypervisor) on your recovery hardware you could do the Bare Metal Restores into virtual machines. It should then be relatively simple to transfer the recovered VMs to new hardware when it becomes available. Hyper-V in Windows Server 2012 allows live migrations without any additional software.

Reply | Quote