Wednesday, June 26, 2013

Microsoft Network Load Balancing, and how to make it work with a Cisco Nexus

Microsoft Network Load Balancing has one or two “interesting” issues due to the way it does networking.  More specifically, it utilises a Multicast MAC address with a Unicast IP address.  Someone much smarter than me explains the ins and outs of the issue here

This issue manifests itself in that clients outside of the cluster’s LAN cannot connect to the IP.

Now that we understand the issue,  how do we resolve it?  The fixes listed in the article are all valid, however we can get this working by tweaking the config on our Nexus.

interface Vlan10
  no ip redirects
  ip address
  ip arp 03BF.0A03.00CF
  description Server Farm
  no shutdown

Your clients should now be able connect to the cluster IP address without issue.

Updating the firmware (DDOS) on a DataDomain

EMC has done a great job with DDOS 5.2, as such I’ve had to do a couple of upgrades recently.  The upgrade manual is quite a tome, so this is my notes, distilled.  I’ve added a couple of extra steps just to make sure there are no error conditions that can hamper the upgrade or the operation of the DataDomain.

Pre-Upgrade checks

Review the output of the below commands for any errors and warnings.

  1. Reboot the DataDomain (I do it just because)
  2. ssh into the DataDomain and issue the commands listed below
  3. enclosure show powersupply 1
  4. alert show current
  5. reg show (confirm the following settings)
    1. noauto.enabled
    2. noauto.speed
    3. noauto.full_duplex
  6. ifgroup
  7. disk show state
  8. disk show reliability-data
  9. enclosure show all
  10. enclosure show topology
  11. system show hardware
  12. system show ports
  13. filesys status
  14. log view debug/platform/

Upgrade Steps

Well after all that the steps to actually upgrade the DDOS is a bit of a let-down.

  1. Copy your DDOS .rpm to /ddr/var/releases
    1. You can do this via a NFS mount or you can upload via the GUI
  2. ssh into the DataDomain and issue the commands listed below
  3. system upgrade
  4. Monitor progress with “system upgrade status” or “log watch debug/platform/infra.log

Post-Upgrade Steps

I guess strictly speaking these steps are not necessary, as with the pre-upgrade steps.  I do it just to ensure I have a nice neat base to work from

  1. ssh into the DataDomain and issue the commands listed below
  2. ost disable
  3. ost enable
  4. nfs disable
  5. nfs enable

And that is all there is to it, mostly.