Exchange Server 2010 SP1 DAG
Hi All,
I have 3 node of Exchange Server 2010 SP1 running DAG on Windows Server 2008 R2. I want to testing HA.
1. Shoutdown one node: all Client still can access their mailbox.
2. Shoutdown two node: All Client cannot access their mailbox.
Whether the DAG technology need at lease 2 node is up? Could you guide me?
BR,
Khemarin Set
I have 3 node of Exchange Server 2010 SP1 running DAG on Windows Server 2008 R2. I want to testing HA.
1. Shoutdown one node: all Client still can access their mailbox.
2. Shoutdown two node: All Client cannot access their mailbox.
Whether the DAG technology need at lease 2 node is up? Could you guide me?
BR,
Khemarin Set
Radweld
Need further information on your exchange design. Were are the nodes geographically and did you configure DAC mode for the dag?
ASKER
3 node is the same subnet and one datacenter. I have 9 database for each server. I haven't configure DAC mode for DAG.
BR,
Khemarin Set
BR,
Khemarin Set
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Actually Akhater thats shouldnt be the case. When there are an even number of nodes in a cluster you need a file share witness in order to maintain quorum, if there is no standby FSW then quorum will lost and the cluster will no longer function.
http://technet.microsoft.com/en-us/library/dd298065.aspx
The only other Exchange setting that might inhibit failover is the MaximumActiveDatabases value you assign using the Set-MailboxServer -Server <servername> -MaximumActiveDatabases <numeric> to limit the number of databases that can mount on a single host.
However I suspect your issue is loss of quorum and subsequent failure of the cluster.
http://technet.microsoft.com/en-us/library/dd298065.aspx
The only other Exchange setting that might inhibit failover is the MaximumActiveDatabases value you assign using the Set-MailboxServer -Server <servername> -MaximumActiveDatabases <numeric> to limit the number of databases that can mount on a single host.
However I suspect your issue is loss of quorum and subsequent failure of the cluster.
it is the case Radweld, when you have an even number of nodes the share witness is to make it odd and it all goes back to what I said.
the share witness is a voter added to the even number of nodes to make it odd that's it
and the failure the OP is experiencing is totally normal
the share witness is a voter added to the even number of nodes to make it odd that's it
and the failure the OP is experiencing is totally normal
ASKER
The share witness map path my Active Directory. Do you have any more info?
What other info do you need? you have an odd number of nodes so you are not using the share witness to start with
ASKER
@Akhater,
I just want to verify odd or even, if the node is odd: the configure share witness, it will not working.
If my node is even, my share witness is working, right?
Lastly, if the node is odd, we no need to configure share witness, right? do you have any reference.
BR,
Khemarin
I just want to verify odd or even, if the node is odd: the configure share witness, it will not working.
If my node is even, my share witness is working, right?
Lastly, if the node is odd, we no need to configure share witness, right? do you have any reference.
BR,
Khemarin
Anytime you setup a DAG you have to configure the share witness exchange will automatically decide to use it or not
"if the node is odd: the configure share witness, it will not working."
True it is configured but not used
"If my node is even, my share witness is working, right"
right
"if the node is odd: the configure share witness, it will not working."
True it is configured but not used
"If my node is even, my share witness is working, right"
right
ASKER
This is the reason that, when I shoutdown 2 of 3 nodes. The client cannot connection to the mailbox, right? or any reason?
BR,
Khemarin Set
BR,
Khemarin Set
when you shutdown 2 nodes out of the 3 you have a loss of quorum and the remaining server will shut the databases down.
If you check the dbs on the remaining server they will be dismounted. this is a normal behavior
If you check the dbs on the remaining server they will be dismounted. this is a normal behavior
ASKER
Thank you for support.
My pleasure, thanks for the points
"when you shutdown 2 nodes out of the 3 you have a loss of quorum and the remaining server will shut the databases down.
If you check the dbs on the remaining server they will be dismounted. this is a normal behavior"
This is totally incorrect and only the case if you configure DAC mode to prevent split brain.
Normal operation
Fail 1st Server
Fail 2nd Server, all databases come active on 3rd Node
DAG2.JPG
If you check the dbs on the remaining server they will be dismounted. this is a normal behavior"
This is totally incorrect and only the case if you configure DAC mode to prevent split brain.
Normal operation
Fail 1st Server
Fail 2nd Server, all databases come active on 3rd Node
DAG2.JPG
Radweld no matter if the DAC is enabled or not Exchange 2010 uses Majority Node Set and it means the N/2+1 of the voters should be online to have quorum.
read http://technet.microsoft.com/en-us/library/dd979799.aspx
Database Availability Group Quorum Modes
read http://technet.microsoft.com/en-us/library/dd979799.aspx
Database Availability Group Quorum Modes