dear all,

we are planning configure DAG for our exchange 2013 infrastructure, the setup will be in following 

DC 1- one mailbox and one cas server

DC 2 - once mailbox and 1 cas server

Dc3 - file witness server

all the data center are in different network segment 

DC 1 and DC 2 have 10gbps fiber connectivity and to DC3 10 mpbs conenctvity

CAS server will be configured with load balcning using HLB and DAG for mail box servers.

I have question regarding the HA in this scenario,

1)if the connectivity between DC1(PAM) and DC2 is failed ,will the databases will automatically mount on DC1 if the PAM is in DC1?

2) if the connectivity between DC1(PAM) and DC3 is failed, but Dc2 and DC3 have connectivity, will the  databases mount automatically at DC2.

3) if DC3 connectivity is failed, DC1 (PAM) and DC2 have conenctivity, will the databases mount at DC1?

4) if the connectivity to DC2 and Dc3 failed, shall  we need to do a disaster recovery scenario? 

please help me

1) Not necessarily. The mailbox server that manages to keep the lock on the file witness server will keep running (and along with it the cluster), while the other mailbox node will shut down. If DC1 survives and also holds the PAM, then it will initiate mailbox database failover according to this Microsoft article, which - if the database copies are correctly set, will result in all the DBs being active on DC1. I don't know what happens if DC1 isn't the surviving node, but holds the PAM. Will the DC2 assume this role, going from SAM to PAM ? Would love if someone provides insight here.

2) If DC1 can communicate with DC2 then the cluster will keep running, regardless what happens with the witness

3) Nothing happens and the mailbox databases stay as they are. The witness going down will not trigger a failover action.

4) If in your scenario you consider DC2 and DC3 permanently lost as well, you can still survive by forcing the cluster to start on DC1

The server that is currently the PAM gets a 6 second headstart to lock the share on the witness. If it is unable to do so, the other servers try to get a lock. So if DC1 fails and had the PAM, DC2 will get a lock, the PAM will move, database copies in DC2 will mount, and any servers still with power in DC1 will dismount their databases and shut down the cluster service.