Help with IPMultiNICPlus

[Julio_Pacheco]

Hello, On a test cluster (Windows 2003 Std w/SP2, SFW-HA 5.0 RP2) i've been trying to add an IPMultiNICPlus resource.

The network configuration of the nodes is as follows:

* Node LABVCS-NODE1
1st NIC MAC Address: 00-0C-29-87-23-75
1st NIC IP address: 192.168.0.11 (static, configured via Windows Control Panel)
2nd NIC MAC Address: 00-0C-29-87-23-93
2nd NIC IP Address: 192.168.0.12 (static, configured via Windows Control Panel)

* Node LABVCS-NODE2
1st NIC MAC Address: 00-0C-29-F0-59-91
1st NIC IP address: 192.168.0.13 (static, configured via Windows Control Panel)
2nd NIC MAC Address: 00-0C-29-F0-59-AF
2nd NIC IP Address: 192.168.0.14 (static, configured via Windows Control Panel)

The configuration for the IPMultiNICPlus resource is:

    IPMultiNicPlus ip-FO (
        AdapterList @LABVCS-NODE2 = { "00-0C-29-F0-59-91", "00-0C-29-F0-59-AF" }
        AdapterList @LABVCS-NODE1 = { "00-0C-29-87-23-75", "00-0C-29-87-23-93" }
        AdminIPAddr @LABVCS-NODE2 = "192.168.0.16"
        AdminIPAddr @LABVCS-NODE1 = "192.168.0.15"
        AdminSubnetMask = "255.255.255.0"
        Address = "192.168.0.17"
        SubNetMask = "255.255.255.0"
        )
 

All NICs mantain connection status, and there are no duplicate IP addresses on the segment.

The documentation states that If the AdminIPAddr attribute is configured incorrectly, the system will not be accessible in the network until the IPMultiNicPlus resource is brought online on the system.

As I understand, the configuration shown should work, however, as soon as the resource is probed, all the static IP addresses dissapear, being shown as 0.0.0.0.

After onlining the resource, the only addresses configured are the admin IP of the active node, and the service IP  (192.168.0.17). The inactive NICs remain at 0.0.0.0

Any ideas?

Thanks,

Julio

[RiaanBadenhorst]

Hi

The guides states this

If the active NIC of the IPMultiNicPlus resource maintains its connection status, the agent reads the status variable of the active NIC to determine the status of the resource. If the resource does not maintain its connection status, the agent verifies that the active NIC has an IP address assigned, and sends a ping to all the hosts listed in the attribute PingHostList.

I would therefore deduct that if your card maintain their connection status VCS will not require to configure the IP address on the in-active NIC. It trust the status from the NIC. If it didn't, then it would online the IP so it can ping the address listed in the ping hosts lists.

You can test this theory by configuring the "UseConnectionStatus = False" for each system and the PingHostList attribute.

Then recheck the IP's. I think it will show both 16 and 165 online on their respective systems.

R

[Wally_Heim]

Hi Julio,

The IPMultiNicPlus agent is not used much in Windows.  From your description and my past experience with the agent it sounds like this is working exactly like it should.

The AdminIP is the IP address that you should use to connect to the node and not the static IP addresses there were there on each NIC.  As Riaan pointed out the NIC is maintaining the connection status of the physical NICs.

The more common configuration on Windows would be to use hardware NIC Teaming with the regular NIC and IP resources.  With NIC teaming you can do either load balancing or failover with 2 or more NICs on the server.  The load balancing mode allows for much higher throughput than what the IPMultiNicPlus agent allows while still maintaining the same fault tolerance that IPMultiNicPlus has.  You might want to consider switching to NIC Teaming on Windows.

FYI - Support for IPMultiNicPlus is being limited in the 5.1 release and is not supported on Windows 2008 servers at this time.

Thanks,

Wally

[Julio_Pacheco]

Hi Wally,

As you correctly noted, the  agent is actually working, since the admin and service IP addresses got assigned to the active
NIC on the online node, and they migrate as expected upon disconnection of the NIC.

What's puzzling me is that according to the agents guide, "If this attribute (AdminIPAddr) is configured incorrectly, the
system will not be accessible in the network until you bring the IPMultiNicPlus resource online on the system".

As i understand this statement, on a cluster with a correct configuration you shoud have network access to any node using the
static IP addresses assigned to each NIC, however, in the configuration described on my previous post, the static IP addresses on
both nodes get removed as soon as the IPMultiNicPlus resource is probed, effectively disabling network access until the resource is online.

By the way, thanks for the tip about Windows 2008. I'll forward it to our customers.

Regards,

Julio

[Wally_Heim · Accepted solution]

Hi Julio,

The IPMultiNicPlus agent removes all static IPs from the NICs that it is configured for.  The AdminIP is the IP address that will enable you to have access to the node for administrative purposes.  If the IPMultiNicPlus agent is misconfigured then it might not put a valid AdminIP on the node making it unreachable from the network.  Assuming that the Address attribute is set correctly then the node would be accessible via that IP address once the resource is brought online. 

I agree that the statement is confusing and should be reworded. 

Thanks,

Wally

[Julio_Pacheco]

Wally , thanks for the information.

Now it's clear that the agent is working as intended.

Regards,

Julio