Disaster Recovery Best Practices

[Ben]

I wanted to get some thoughts around disaster recovery. I've read the MS
provided document and accroding to that document I need to start marking
transaction logs across my databases.

Is this really necessary. I don't need a point of failure recovery. If the
project server fails there really isn't a problem in loosing some data in the
case I'm working on. Is there any problem with just using weekly full
back-ups of the database along with daily incrementals? Thanks.

 

[Rick Roszko]

Short answer

All you need is the make a full backup (+ incrementals) of the
"ProjectServer" (default) database and the SharePoint Services database
"STS_your-server-name_1" (default).

Long Answer

So, let’s say your data center is wiped out due to whatever reason. Lucky
you, the tapes are offsite. What to do?

(a) Install a fresh Win2003Server with IIS and MS SQL2000 with Analysis
Services and MS Project Server 2003 with SP1. (I don't like SP2a; but of
course if you are at SP2a already, make sure you go to SP2a...)

(b) Restore your “ProjectServer†as a different name to the SQL db

(c) Restore "STS_your-server-name_1†as a different name to the SQL db

Great, how the heck do you connect the old data to the new install?

(d) Search on “editsite†on www.microsoft.com and find the instructions and
the executable. Follow directions, and your “ProjectServer†data is easily
connected.

(e) SharePoint is more involved. Search on “905386†to get the
knowledgebase doc to connect your SharePoint site to your new install.

(f) Recovered from Disaster. Done.

Okay, there is a whole bunch of steps that need to be done correctly in (d)
and (e), so please try this out before disaster strikes! There will be
things you didn’t think about or something didn’t work exactly right. Time
to find out is now, not when disasterstrikes!

Hope this helped…

 

[Aaron Tamblyn]

One technique we use is to image the server and apply the image to Virtual
Machine. This validates that the image is healthy and the machine can be
brought quickly online as a temporary DR measure while the server is being
rebuilt. Food for thought...

 

[Rick Roszko]

Ah, good point. Another IT trick is that disks are so cheap now, instead of
RAID5 they are going RAID1 (i.e. mirrored disks). The easiest and fastest
way is to break the mirror and remove the entire hard drive (assuming server
class machines with hot swap capability of course) and save that off site.
Put in a fresh hard drive in and rebuild the mirror. (For each server, of
course.)

If a server fries, get the same model/configed server and just pop in the
off-site hard drive, and viola! Your're up by the time the server boots. (Of
course, if it's a complete disaster, the network Active Directory/domains
need to be up...)