im unsure about the setup of our new mailserver and the connection to the luns on the netapp.
at the moment we have 1 mailserver with connection to several luns on our fas3020.
what if we have 2 servers in a cluster in the future, how is it possible that they both have access to the storage lun like ESX shares its storage ?
Are you talking specifically about Exchange 2007 or 2010? The difference is quite substantial as the former supports many different ways of clustering, whilst the latter is pretty much limited to DAG - see this for handy explanation & comparison: http://www.exchange-genie.com/2009/04/database-availability-group-dag-exchange-2010/
In particular 2010 doesn't allow so called Single Copy Cluster (SCC) where two mail servers access the same LUN with the same mail database (well, not at the same time, but you know what I mean).
Additional reading for 2007 version:
that already helps me a lot. its about 2010.
so if i can ask further, im gonna make a new LUN, install exchange 2010 and use DAG. if i add a 2nd node to DAG it will automatically get the storage from DAG/Exchange without me having to worry about mapping that lun to the 2nd vm ?
My understanding of DAG is as follows:
- as CCR or SCR this is a form of host-based data replication
- DAG is 1-server-to-1-LUN type of scenario - multiplied by a number of DAG instances
- witness node can be any file share (UNC path), so shared LUN is not required for quorum.
So, simply speaking, it will be LUN_1 mapped to MailboxServer_1 & LUN_2 mapped to MailboxServer_2 - the data transfer between LUNs will be handled by Exchange.
In a DAG setup, each node will have a dedicated LUN of its own. So you'll have to make sure that the new node that is added has LUNs connected to it as dedicated disks. Note that DAG is somewhat similar to a CCR cluster setup so there is no need for shared disks.
Additionally, if you plan to use SME (SnapManager for Exchange) then the same drive letters/mount points should be used on each member server of the DAG.
Technical Marketing Engineer
SMAI , NetApp
thanks a lot. gonna try this setup then.
1 additional question, wouldnt this setup mean to have all database luns twice ? since every node has its own lun, its double the space its consuming right now, correct ?
It's 100% correct.
That's why ideally I'd like Exchange virtualised, because in that case I can rely on following solution stack, not touching Microsoft clustering:
- SMVI for OS / binaries backup
- SME for Exchange mailstore backup
- VMware HA for unplanned hardware failures
- VMware VMotion for planned hardware maintenance
And all that with just single Exchange licence (for mailbox role), one OS to manage & patch and one copy of data!
When we throw SnapMirror into the mix (possibly with VMware SRM), we have site-level resiliency, still maintaining a single copy of Exchange binaries - fantastic simplicity comparing to native Exchange clustering!