I need 2 SFP+ ports for a project (2 VM hosts, 10Gbps NICs) and am looking at the CRS210 as a possible option.
I’m looking at the block diagram here
http://i.mt.lv/routerboard/files/CRS210-150409135339.png
and it suggests that the SFP+ ports are on the same switch as the 8 1Gbps ports .
Question is, can I have the two SFP+ and one 1Gbps port in a switch group on this hardware without bridging?
Yeah, that should work without an issue!
But if you don’t mind me asking. If you are using the 2x10Gbps ports for VMware ESXi hosts, you are only left with 8x1Gbit. Which clients are going to use these 10Gbit connected servers? Because none of them have the bandwith to do so.
8x1gbps=8gbps is better than 1gbps
Also, all ports are in one switch. There are no 2sfp and one eth. All or bridges. I never used CRS so correct me if I’m wrong.
The 10Gbps ports are just for drbd sync.
The proxmox hosts will have multiple gigabit ports for virtual machines, but this is really just about having a very fast storage network.
I’m not quite sure what you are saying but you shouldn’t use a bridge on a CRS you need to use Master/Slave ports within the switch group. And yes, it has 8x1Gbit RJ45 and 2x SFP+?
Ah, I see, yes, as interlinked storage that could work but if it’s two boxes and they are close together, wouldn’t just linking them together be a much better option? That’s what I do in my home lab for my iSCSI storage. It also prevents any slow down a switch could cause.
If you intend to use the same 10Gbit NIC’s for Public and Sync, then I understand your design scenario and yes, that would work. Be mindful though that no CRS supports link-aggregation (or at least not LACP) and thus any client will indeed be limited to 1Gbit. Having the storage servers linked with 10Gbit would still be beneficial though. 
They are currently directly connected. The problem is that I can’t get quorum for a drbd cluster this way so I need a switch and this crs is by far the cheapest option for 2 sfp+ ports. A third cluster node would be on gigabit in the same switch group, but that would only have the cluster volume on it, not the production storage.
I see, yes, that should work just fine with a CRS.