cfgtechs
asked on
Win2012 R2 or 2016- i want to implement continuously available file shares, failover clustering hardware requirements
I have two identical win12 servers- the storage is via LSI mega Raid 5/Raid6 adapter. I have no network storage between the servers. Just the local network.
Will this work with just the two servers on the lan? CAFS will only be for file sharing availability not hyperV maintenance or SQL hosting.
I've read through this failover guide but it's unclear because it goes on, "if you use iscsi" so i'm taking it that an iscsi san between the two servers is not required but other blog posts say that it is.
https://technet.microsoft.com/en-us/library/jj612869.aspx
thanks and regards!
Will this work with just the two servers on the lan? CAFS will only be for file sharing availability not hyperV maintenance or SQL hosting.
I've read through this failover guide but it's unclear because it goes on, "if you use iscsi" so i'm taking it that an iscsi san between the two servers is not required but other blog posts say that it is.
https://technet.microsoft.com/en-us/library/jj612869.aspx
thanks and regards!
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
Lee,
it's only at proposal stage, i was thinking that CAFs would be closer towards a type of volume mirroring/disk duplexing albeit between two servers.
Thank you cliff, so there would be a SAN required.
The main goal was to have a file share readily available in case a physical machine drops. it looks like the SAN introduces a possible failure point, it looks like it'll have to be DFS?
peter
it's only at proposal stage, i was thinking that CAFs would be closer towards a type of volume mirroring/disk duplexing albeit between two servers.
Thank you cliff, so there would be a SAN required.
The main goal was to have a file share readily available in case a physical machine drops. it looks like the SAN introduces a possible failure point, it looks like it'll have to be DFS?
peter
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
With Server 2016 you can set up a 2 node Storage Spaces Direct (S2D) cluster.
A lot of folks balk at the Server Datacenter license requirement but we get around that by putting the two licenses on SPLA (Service Provider's License Agreement). It's a small monthly cost for a huge benefit.
I suggest picking up a pair of Mellanox ConnectX-3 (CX312B) NICs, direct connect them, and enable RoCE RDMA for the East-West (node to node) traffic requirement. They can be had for a song.
Note that the drives in each server must be JBOD not RAID. The current controller needs to support JBOD mode.
The following Microsoft article outlines a sub $5K two node cluster: Kepler-47.
A lot of folks balk at the Server Datacenter license requirement but we get around that by putting the two licenses on SPLA (Service Provider's License Agreement). It's a small monthly cost for a huge benefit.
I suggest picking up a pair of Mellanox ConnectX-3 (CX312B) NICs, direct connect them, and enable RoCE RDMA for the East-West (node to node) traffic requirement. They can be had for a song.
Note that the drives in each server must be JBOD not RAID. The current controller needs to support JBOD mode.
The following Microsoft article outlines a sub $5K two node cluster: Kepler-47.
The "if is only specifically calling out iSCSI because many people try to run general traffic over the same adapter as iSCSI traffic, which is a no-no. But with SAD or FC, that wouldn't even be a thing. So imolementing those two types of shared storage would not need to read that disclaimer. It does not.say or even properly imply tjatbshared storage isn't required.
2016 is a slightly different story as you can technically use direct attached storage to do storage spaces direct. But that is highly specialized and you'd not do it on the hardware you are currently discussing. S2D really needs to be purpose-built hardware full stop.