The Datacenter move is COMPLETE

Now I just have to go in and pull about a thousand cables out of the old datacenter and tidy up a bit.  We were there until about 4am.  My goal from 9pm to 3am was to shutdown and move the Clariion to it’s new home in the primary data center.  I got it done at minutes before midnight, however was plagued with Veritas issues for the rest of the evening.

When we made the move, we moved to a new, more robust network.  This allowed us to do away with the dedicated gig network we were using with Veritas, because between the more robust production network and the fact that most of the database backups will now be done with TimeFinder, the need for the extra network is nil.

The issue is that Microsoft doesn’t like dual-networks – getting it configured for DNS and a dual-homed network was hard enough – getting the dual-homed config removed was even harder. 

8 comments

Skip to comment form

    • on January 22, 2007 at 4:10 pm
    • Reply

    Dunno how extensive your *dual networks* needed to be. I’ve found it to be pretty darn easy myself though.

    Say for instance you have:
    10.10.10.1 >> going to your backend storage
    10.10.20.1 >> going to the rest of your network/external network/www (so you want this as your *primary* connectoin.

    You simply set a gateway for the 10.10.20.1 so all traffic by default flows down it, and don’t set one for 10.10.10.1. It will still find it’s way out the 10.10.10.1 for traffic that needs to head that way, but all traffic by default will attempt to go down the route that has a gateway set. Make sense?

    Of course, it might not even apply to your situation :>

  1. Been there, done that, got the t-shirt.

    The problem wasn’t with the routing, though our network god (he should get his own site) says he saw a lot of bogus traffic on the production and backup network from a few hosts that were under the mistaken impression that they were routers.

    The main problem was in Windows’ DNS server. Because the host with two IP’s was set to auto-register in DNS, it would register both IP’s in the ‘example.com’ domain. Since the 10.0.x.x backup network was not reachable from the 192.168.1.x production network, a host doing a lookup on ‘bob.example.com’ was actually getting the 10.0.x.x address and trying to contact the server through it, obviously failing.

    I hate windows…. This works so much better in *nix.

    • on January 24, 2007 at 12:49 am
    • Reply

    Obvioulsy congrats on a successful job and providing us with some great insights.

    But what are you going to do with your time now? And waht am I going to read over lunch?

  2. LOL – next step is implementing TimeFinder/SNAP – I got the VDEV devices carved up today, and got the first snap run, now it’s a lot of scripting.

    Is there any suggestions on how to synchronize windows batch files on two different servers?

    With Unix I’d use rcp or something along those lines, but don’t think anything exists like that for Windoze.

    Don’t worry, my life/work never get boring.

    • on January 24, 2007 at 7:59 am
    • Reply

    SFU and rcp?

  3. I didn’t think MS still made Services For Unix.

    Might be – i’m going to investigate using Veritas to script it. That way I can generate the snap on the production host using TSIM and then just kick off the backup policy when that’s done. The preaction script on the media server can mount the drives before the backup starts, and the postaction can unmount them and terminate the snap session.

    • on January 24, 2007 at 8:45 am
    • Reply

    You sure do trust veritas an awful lot :>

  4. I almost had a use for Veritas BMR tonight, (read my latest post. http://www.sangod.com/?p=76

    I bet a spare server will manifest itself this week for Backup/Restore/BMR testing…. Given how close we actually came to having to use it. 🙂

Leave a Reply

Your email address will not be published.