LanMan resource fails and then faults during failover. Several error messages appear in VERITAS Cluster Server Logdesk.
| Article:TECH15055 | | | Created: 2001-01-26 | | | Updated: 2004-01-21 | | | Article URL http://www.symantec.com/docs/TECH15055 |
Problem
LanMan resource fails and then faults during failover. Several error messages appear in VERITAS Cluster Server Logdesk.
Error
TAG_B 2001/04/29 12:13:09 VCS:LANMAN Agent:180004:online:Exch_Lanman:Adding virtual server name failed. (Service:00h) Windows NT Error#234 : More data is available.
Solution
The following messages may appear in the VERITAS
Cluster Server (VCS) Logdesk, when the LanMan resource fails and then faults
during failover:
TAG_B 2001/04/29 12:13:09 VCS:LANMAN
Agent:180004:online:Exch_Lanman:Adding virtual server name failed. (Service:00h)
Windows NT Error#234 : More data is available.
TAG_B 2001/04/29 12:15:09 VCS:LANMAN
Agent:180003:clean:Exch_Lanman:Removing virtual server name failed.
(Service:20h)
The issue arises during a failover of a VCS Service
Group that contains a LanMan resource. When a node is hosting a Service Group,
the LanMan resources registers its virtual name. The Public Network switch
caches it along with the MAC address. The Browser service on the Domain
Controllers also have this information. If during a failover or onlining, the
ARP cache on the switch is not flushed, the failover node will not be able to
register its MAC address on it. This is because the network already has the name
registered to the first node. The second node is considered a duplicate and it
is ignored. The result is that the LanMan resource will not failover and will
eventually fault.
To resolve this problem:
1. Remove the WINS entry for the virtual
name.
2. Stop and Start the Browser Service for each of the
Domain Controllers. This will cause the Domain Controllers to create a new
Browse List.
3. Power Cycle the Public Network Switch. This will
cause the switch to recompile its list of MAC addresses.
To minimize the risk of reoccurrence:
1. Set the Public Switch's ARP Cache Flush Interval to
a lower setting.
2. Make sure the Public Switch's firmware is up to
date.
3. Enable Gratuitous Broadcasts on the Public Network
Switch. Refer to Microsoft TechNet Articles:
Q244331 - MAC Address Changes for Virtual Server During
a Failover with Clustering
Q199773 - Behavior of Gratuitous ARP in Windows NT 4.0
Thank you.