Exchange 2003 Disaster Recovery
CSP
Is there no way to get an Information Store up and running on a new
piece of hardware / in a new geographic location without doing a
/disasteryrecovery installation of Exchange?
The Recovery Storage Group is handly for recovering pieces of mail /
mailboxes by extracting to a PST, but will only merge that mail back
into the original mailbox on the original server, which in the case of
a disaster does no good as it doesn't exist. It would be incredibly
handy to be able to move the RSG into a production Information Store
somehow, but that does not seem to be supported.
Short of buying a replication product like DoubleTake, what do people
have in their recovery plan for a speedy recovery from a machine /
site failure?
I feel I must be missing something basic here, or is it acceptable to
have to build a new Exchange server using /disasterrecovery before you
can get your users up and running again?
Jamestechman
software your choices are:
1. Hosted providers that offer business continuity
http://www.microsoft.com/exchange/services/continuity.mspx
2. Create a standby server and when servers crashes; re-home exchange
attributes to the new server which will take some time and is not very
seamless. I did some testing with this a while back and provided how
to and caveats.
Exchange: BCP Planning Exchange 2003
http://msexchangetips.blogspot.com/2007/05/exchange-bcp-planning.html
3. Upgrade to Exchange 2007 and use CCR/SCR
James Chong (MVP)
MCSE | M+, S+, MCTS, Security+
msexchangetips.blogspot.com
Andy David {MVP}
>
>Is there no way to get an Information Store up and running on a new
>piece of hardware / in a new geographic location without doing a
>/disasteryrecovery installation of Exchange?
the /disasterrecovery switch assumes the original AD is intact.
If the new location is not the same AD forest, it wont work.
>
>The Recovery Storage Group is handly for recovering pieces of mail /
>mailboxes by extracting to a PST, but will only merge that mail back
>into the original mailbox on the original server, which in the case of
>a disaster does no good as it doesn't exist. It would be incredibly
>handy to be able to move the RSG into a production Information Store
>somehow, but that does not seem to be supported.
>
>Short of buying a replication product like DoubleTake, what do people
>have in their recovery plan for a speedy recovery from a machine /
>site failure?
UPgraded to 2007 and CCR :)
>
>I feel I must be missing something basic here, or is it acceptable to
>have to build a new Exchange server using /disasterrecovery before you
>can get your users up and running again?
If the server is completely hosed, its your best option and quite easy
actually.
TheITDude
recoveries to totally new hardware at a cold site during DR testing. I used
the technique here:
http://safari.oreilly.com/0596007175/exchangeckbk-CHP-11-SECT-9
TheITDude
cookbook:
Recipe 11.9. Performing Disaster Recovery of a Cluster Node to a
Nonclustered Server
Problem
Your clustered Exchange server is down. While you restore it, you want to
offer service using its databases on a regular, unclustered node.
Solution
Using a graphical user interface
1.
Use ADSI Edit to remove the cluster's AD object:
1.
Launch ADSIEdit.msc.
2.
Expand the forest configuration container.
3.
Expand the container for the administrative group that contains
the clustered server (it'll be located under CN=Services, CN=Microsoft
Exchange, CN=Administrative Groups).
4.
Expand the Servers container in the AG.
5.
Right-click on the target server and use the Delete command.
When ADSI Edit prompts you to confirm the deletion, click Yes.
2.
Remove mailbox attributes from the user accounts whose mailboxes were
on the former server:
1.
On the Exchange server or on a computer with the Exchange
Management tools installed, launch the ADUC snap-in.
2.
Make sure that the advanced features mode is on (use the
Viewrightwards double arrow Advanced Features command to toggle it if
necessary).
3.
Select the users whose mailbox attributes you want to remove.
You can select them one at a time or by using the standard Windows selection
modifier keys.
4.
Right-click the user and use the Exchange Tasks command to start
the Exchange Task Wizard.
5.
In the Available Tasks window, select Remove Exchange Attributes
and click Next.
6.
In the Remove Exchange Attributes window, click Next to indicate
your understanding of the tool's dire warnings.
7.
Click Finish.
3.
Build the new server, using the same name as the cluster's virtual
server and the same version (and service pack level) of Windows.
4.
Install Exchange on the new server, but don't use the
/disasterrecovery switch.
5.
After Exchange is installed, create storage groups and databases whose
names and locations match what was on the original server.
6.
Restore your databases from a backup (or copying them from the
original server or SAN).
7.
Mount the databases and check the event log to verify that there were
no errors.
8.
Reconnect the restored mailboxes to their users using the Mbconn
utility
CSP
odd to me that the fastest way to get a dead server up and running
would be a fresh Exchange Install. Especially with how fast an easy a
RSG restore is.
I suppose thinking about it the install will probably only add 10 or
15 minutes onto the restoration time as long as I have the OS up and
running. I imagine I can have a server standing by and rename it to
the name of the dead machine, correct?
I can't think of a safe way to test this process other than creating
an entirely new AD and Exchange Forest, is that right?
CSP
I had seen some mentions of option 2, but it didn't seem like a really
recommended way to go. Editing with ADSI in the middle of an outage
isn't really a situation I'd like to put myself in.
It looks like I'll try to test and document the /disasteryrecovery
method and try to push through a proposal for 2007, but I know where
that is going to end up.
Thanks again.
Andy David {MVP}
>Thanks Andy. I was afraid that might be the answer. It still seems
>odd to me that the fastest way to get a dead server up and running
>would be a fresh Exchange Install. Especially with how fast an easy a
>RSG restore is.
>
>I suppose thinking about it the install will probably only add 10 or
>15 minutes onto the restoration time as long as I have the OS up and
>running. I imagine I can have a server standing by and rename it to
>the name of the dead machine, correct?
Yep. Keep the O/S patched the same as production, then install
Exchange with disasterrecovery swtich and service pack and hotfix to
the same level as the production then do a restore, reboot and you are
ready to go. (You'll have to remove or disable the old computer
account first). This also works when you want to move to a new server
with new hardware and dont feel like moving mailboxes or public
folders etc...
>
>I can't think of a safe way to test this process other than creating
>an entirely new AD and Exchange Forest, is that right?
Yea, labs are good!
Holz
> I had seen some mentions of option 2, but it didn't seem like a really
> recommended way to go. Editing with ADSI in the middle of an outage
> isn't really a situation I'd like to put myself in.
Just an add-on, I have implemented a stand by mail server at work based
on open source. It is not a real time solution, however it allows you to
continue send/receive inside and outside e-mails without interruptions,
and when you Exchange is back on line, retrieving your "offline" e-mails
into outlook is as easy as drag and drop. The stand by can run on a
Pentium 4 with 526 MB Ram, depending of course on the number of you users
and store size you have. Setting up the server including Linux
installation takes about an hour. I am in the process of writing the
manual and guidelines (free of charge), should be done in a week or so.
If you need some heads up, come back here and post.
--
:-)
daveberm
> piece of hardware / in a new geographic location without doing a
> /disasteryrecovery installation of Exchange?
>
> The Recovery Storage Group is handly for recovering pieces of mail /
> mailboxes by extracting to a PST, but will only merge that mail back
> into the original mailbox on the original server, which in the case of
> a disaster does no good as it doesn't exist. It would be incredibly
> handy to be able to move the RSG into a production Information Store
> somehow, but that does not seem to be supported.
>
> have in their recovery plan for a speedy recovery from a machine /
> site failure?
>
> I feel I must be missing something basic here, or is it acceptable to
> have to build a new Exchange server using /disasterrecovery before you
> can get your users up and running again?
I would investigate the advanced recovery stragetegy of re-homing a
user's mailbox described here:
http://technet.microsoft.com/en-us/library/bb125016.aspx
and database portability as described here:
http://technet.microsoft.com/ en-us/library/aa997006.aspx
The advantage of this recovery strategy is that it is much easier to
test this DR scenario without causing downtime in your production
environment since you can re-home a single user as a test. I would
recommend you set this up in a lab and have a go at it before you
implement in production.
Sincerely,
David A. Bermingham, MCSE, MCSA:Messaging
Director of Product Management
http://www.steeleye.com
"HA and DR for Windows and Linux"