Exchange Server Forums

Failover Question - CCR, 2 subnets, windows 2008

Users viewing this topic: none

Logged in as: Guest

Printable Version

All Forums >> [Microsoft Exchange 2007] >> High Availability >> Failover Question - CCR, 2 subnets, windows 2008

Page: [1]

Message

<< Older Topic Newer Topic >>

Limited time MSExchange.org offer! -- 1.Sep.2008 1:00:00 PM

TechGenix and SolarWinds have partnered to provide free copies of SolarWinds Exchange Monitor to all visitors who join the MSExchange.org Forums. SolarWinds Exchange Monitor is a handy desktop dashboard that continuously monitors Microsoft Exchange to deliver real-time insight into Exchange services, mail queue sizes, and host server health. Learn more about Exchange Monitor and the free offer!

Failover Question - CCR, 2 subnets, windows 2008 - 22.Feb.2008 9:17:17 AM

beowulfs

Posts: 7
Joined: 3.Jan.2008
Status: offline

Windows 2008 RTM, Exchange 2007 SP1.

So, everything is up. one node in is on subnet the other is in a different subnet. the private nics are connected with their own vpn tunnel and they have different subnets too. all network validation tests (except storage, which i thought doesn't matter) passed. replication is happening and shows healthy in the console, except that there is a 50000 copy queue length on the the first storage group. this is to be expected during initial seeding of the 74gb storage group across a 2mbps vpn tunnel. i know that part is not recommended, but it's working. here's my question:

when the link fails and the passive node becomes active (because when replication is finished i'm changing the witness to a server beside the passive node) how do clients that vpn into the failover site, which is on a different subnet, resolve the address of the passive node that is now active if there is only a dns entry for the active node in the active subnet? does the cluster automagically change the dns entry? will the cluster not play happy with a request for that ip address or does it deal with names only?

in cluster management, i created the active node first and not the active and passive together. installed exchange on active. had to wait a few days, added passive node to cluster, installed exchange on passive. i mention this because it's not the same as adding both nodes at once where you get an option to input two ip addresses for the cluster, one for each subnet (that never worked for me either though, one ip address always failed to come online). i also don't know where i would be the ip address for the passive subnet in exchange 2007 anyway.

that's question set one. what happens with name resolution during failover with different subnets?

question two is what does this mean run i run test-replicationhealth:

Network 'Cluster Netw
ork 3' used for internal
cluster communications is
up; however, node 'exc-a
ct' does not appear to ha
ve a configured network i
nterface card (NIC). Chec
k that a NIC is configure
d and enabled for this ne
twork on this node.
Network 'Cluster Netw
ork 4' used for client co
nnectivity is up but node
'exc-act' does not have
a Network Interface Card
configured on it. Check t
hat a NIC is configured f
or this network and is en
abled.
Network 'Cluster Netw
ork 2' used for client co
nnectivity is up but node
'exc-pas' does not have
a Network Interface Card
configured on it. Check
that a NIC is configured
for this network and is e
nabled.
Network 'Cluster Netw
ork 1' used for internal
cluster communications is
up; however, node 'exc-
pas' does not appear to h
ave a configured network
interface card (NIC). Che
ck that a NIC is configur
ed and enabled for this n
etwork on this node.

Thanks loads.

Post #: 1

Featured Links*

RE: Failover Question - CCR, 2 subnets, windows 2008 - 22.Feb.2008 11:59:34 AM

ScottSchnoll

Posts: 15
Joined: 26.Jan.2008
Status: offline

Hi Beowulfs,

There's actually no requirement that the active and passive be installed at the same time, so it's OK to install the active one day and then install the passive the next day. When you installed the active clustered mailbox role, it should have given you an option for both subnets. This option always appears during Setup on Windows 2008 (on Windows 2003, the second subnet option and advanced networking options such as IPv6 and DHCPv4 are unavailable).

Since it sounds like you've configured a single subnet cluster, you have two options to change to a multi-subnet cluster. One option is to remove the clustered mailbox server and reinstall from scratch. Another option is to simply create a new IP address resource in the CMS group, and then make that new address a dependency resource (using an OR clause) of the Network Name resource for the CMS.

When you have a multi-subnet Exchange failover cluster, network name resources for the Cluster and for the CMS have two IP address resource dependencies in which an OR clause is used to allow the NN resource to come online if any of the dependencies are online. In a multi-subnet environment, the Cluster and CMS network identity live at any given time on one node, and thus one subnet. As a result, of their two IP address dependencies, one will always be online and one will always be offline. In other words, we are not multi-homing the CMS so that it is always accessible through IPs at the same time; but rather, configuring it with the ability to change IP addresses when ownership of the CMS changes.

For example, say I have Subnet A (10.0.0.x) and Subnet B (192.168.1.x). NodeA is on Subnet A and has an address of 10.0.0.5 and Node B is on subnet 2 with an address of 192.168.1.5. (Note these are just example addresses). The Cluster network name will have two IP address dependencies that correspond to IP address resources in the cluster group (for example, 10.0.0.4 and 192.168.1.4) and the CMS network name will also have two IP address dependencies that correspond to the IP address resources in the CMS group (for example 10.0.0.8 and 192.168.1.8). When the CMS is owned by NodeA, because that node is associated with SubnetA, the CMS will use 10.0.0.8. When ownership of the CMS is transferred to NodeB (through a move or a failover), the CMS will use 192.168.1.8. Windows Clustering in Windows 2008 knows which network cards are associated with which subnets and IP addresses, and it will bring the correct IP address online, and take the IP address for the other subnet offline.

Name resolution will obviously change when the CMS or cluster changes ownership. Whenever you'll be in a situation where the Exchange server's IP address is going to change (e.g., such as moving the CMS in multi-subnet CCR environment), we recommend configuring a DNS TTL of 5 minutes for the CMS Network Name. Also, if you have any management tools that connect to the cluster itself by it's name (again, in a multi-subnet environment) we recommend a DNS TTL of 5 minutes for the Cluster Network Name, as well. In Windows 2008, this is a quick cluster.exe command:

cluster.exe res <CMSNetworkNameResource> /priv HostRecordTTL=300

As for the Test-ReplicationHealth failures, those indicate that your cluster networking is not configured correctly, or that there are network interfaces that the cluster can see, but not use or manage. Or vice versa.

Have a look at http://technet.microsoft.com/en-us/library/bb629714(EXCHG.80).aspx for more information on CCR on Windows 2008.

Hope this helps.

_____________________________

Regards,

Scott Schnoll
Principal Technical Writer - Exchange Team
Microsoft Corporation
This posting is provided "AS IS" with no warranties, and confers no
rights.

(in reply to beowulfs)

Post #: 2

RE: Failover Question - CCR, 2 subnets, windows 2008 - 22.Feb.2008 12:16:56 PM

beowulfs

Posts: 7
Joined: 3.Jan.2008
Status: offline

ok, so i added the ip address, and it shows up under Other Resources for the exchange cluster name. i couldn't get the OR clause though. that little part is blank.

also, when i run the validation wizard in fcm, everything comes up fine, so how do i fix those exchange errors? i mean replication is growing. i can watch the database size grow on the passive node. the public folder storage group has already finished replicating.

thanks.

(in reply to ScottSchnoll)

Post #: 3

RE: Failover Question - CCR, 2 subnets, windows 2008 - 22.Feb.2008 12:19:06 PM

beowulfs

Posts: 7
Joined: 3.Jan.2008
Status: offline

what about possible owners? it should only be the passive node, correct? the original active ip address has no dependencies. do i add any? its only possible owner is the passive node.

(in reply to beowulfs)

Post #: 4

RE: Failover Question - CCR, 2 subnets, windows 2008 - 22.Feb.2008 12:30:32 PM

ScottSchnoll

Posts: 15
Joined: 26.Jan.2008
Status: offline

To configure the proper dependencies, you go into the Properties of the Network Name resource. Then, select the Dependencies tab, and click Insert to insert a new dependency line. Be sure to choose OR and not AND.

Then go into the Properties for each IP Address resource, click Advanced Policies, and make the appropriate node a Possible Owner of that resource. Make only one node a possible owner of one IP address resource, and the other node the only possible owner of the other IP address resource.

To fix the Exchange errors, you need to fix your networking. You might have a look at the cluster properties using Cluster.exe to see if anything stands out. You might also need to configure your cluster networks for cluster access, client access, or both if you want log shipping to happen over them.

_____________________________

Regards,

Scott Schnoll
Principal Technical Writer - Exchange Team
Microsoft Corporation
This posting is provided "AS IS" with no warranties, and confers no
rights.

(in reply to beowulfs)

Post #: 5

RE: Failover Question - CCR, 2 subnets, windows 2008 - 26.Feb.2008 1:22:20 PM

beowulfs

Posts: 7
Joined: 3.Jan.2008
Status: offline

still haven't fixed the exchange errors, but i tested failover and it tentatively works. i didn't test mail flow in and out.

quick question:

how do i add a second ip address to the main cluster just like i did for the exchange cluster?

thanks

(in reply to ScottSchnoll)

Post #: 6