Notes from the MMMUG August 2009

Wow not many spare chairs, so this was an Exchange 2010 HA session with Ewan Tsang-MacKellar.

Session started with the HA options Exchange 2007 gave us.  Then Ewan introduced us to Exchange 2010 DAG’s (Database Availability Groups).

• DAG
  • Config information is stored in the AD and the Windows Failover cluster (yeah it’s still used). 
    • It uses the cluster database, heartbeats and the quorum for a failover cluster
  • The DAG is the boundary of replication and failover.
  • Each DAG contains a number of mailbox role servers.  You can have up to 16 servers per DAG and 100 databases.
  • DAGs use Log shipping to move data to other replicas of the database with  1mb Log files. 
    • Replay process similar to Exchange 2010. 
    • Now use TCP sockets NOT SMB for log shipping.
    • Push based log shipping.
    • Can choose the TCP port now too.
    • Supports encryption and compression
  • If a database fails, Exchange will determine the best copy to use.  The CAS will then reconnect outlook automatically.  Failover times are less than 30 seconds (which seem a bit more than Exchange 2007 CCR)
  • Can spread a DAG over Active Directory Sites ;-) Can make it like an SCR target in Exchange 2007.
  • Maximum number of passive copies = # servers in DAG –1 active
  • You can co-locate other roles on a mailbox server that is in a DAG.
  • Database names are unique across an forest
  • Still uses a FSW
  • Incremental Reseeding
  • Continuous Circular logging
• Active Manager
  • HA’s brain
  • Primary (holds changable state information of each database) and secondary
    • PAM
  • One server in DAG is selected
  • Secondary
    • SAM
    • Helps CAS find the active copy of the database
    • All other servers in DAG
  • On failure of a database will select best copy
    • Copy Status [Good]
    • Catalog (healthy of crawling) [Healthy]
    • Copy Queue length [<10)]
    • Replace queue length
  • If all are okay can you set a preference
• Backups
  • Uses VSS
  • Can backup active or passive copies or select a server to backup
  • Can use a database copy as a “backup”
    • 7 – 14 day lag copy

Transport HA

• Shadow Messaging (Redundancy)
  • Hub Transport keeps a copy of a message until it gets confirmation that the message was sent.
  • Uses MsgID to identify each message.  In the event of a failure there is a possibility for duplicate messages to be sent, but the two messages will share a MsgID and “most” email systems will notice this
  • Exchange 2010 keeps a table of MsgID’s for 7 days
  • Size of transport dumpster depends on the time stamp of the worst  log copied.

Other Stuff

• Autodiscover DNS TTL lower to 5 mins