Assuming you plan to implement Microsoft Exchange 2010 and have some budget left, what do you do?
We decided to support the project with an external Exchange-specialized consultant. So far so good.
After going through the design-phase we worked out the following Exchange installation plan:
- 2 mailbox stores, with database availability group (DAG)
- 2 combined clientaccess (CAS) and hub/transport (HT) servers, with network load balancing (NLB)
The installation went more or less quick and smoothly. But afterwards..
I'm picking out one (hopefully interesting) issue, we sorted out a few days later.
During an internal network scan, we were surprised about the following finding:
The broadcast got an intermittent strong increase.
Futher analyzing showed up: multicast traffic from NLB was flooding all ports within the same vlan, instead only the designated ports.
How can this happen?
Cause #1: The Exchange-specialist knows how to click through the NLB setup, but had no in-depth knowledge of it.
Cause #2: The network switch, on which Exchange NLB was attached to, was an HP ProCurve.
Solution: A static MAC entry needs to be configured on the switch.
So we thought this could be fixed quite easy, but were brought down to earth very quickly..
After trying to set the required static entry, the following error appeared:
PROCURVE(config)# static 03bf0a-c8027e interface A2
Value static is invalid.
It turned out, that the used HP ProCurve switches were not accepting Microsoft NLB MAC addresses...
How to succeed?
Buy other switches, VLAN it off or check the NLB multicast IGMP option.
There is a sample configuration for Cisco switches available at vmware.com:
http://kb.vmware.com/kb/1006525
No comments:
Post a Comment