2013-12-19 04:32 AM
Here is the situation:
We have 6 blade servers (HP BL420c), two interconnect switches HP 6120XG and NetApp FAS2240-2 with dual controllers and two 10GbE ports per controller.
VMware vSphrere 5.1 Standard will be used on the servers. We have one 10GbE dual-port NIC per server. Protocol of choice is iSCSI.
To save storage space I want to create one aggregate and 3 volumes on it. Two of them will be presented to VMware hosts.
I'm not sure how to make interconnect links between all this stuff.
First option is to link two ports of the first controller to first switch (using VIF) and two port ports of the second controller to second switch (using VIF). In this case I'm not sure how the second controller would handle the iSCSI traffic. Will it pass the traffic to first controller through interconnect? Or the second controller will not be used until failover?
Second option is to connect first port to first swich, second port to second switch etc.
Also, VMware recommends to add 1 VMkernel per 1 physical port for iSCSI traffic. I'm going to use 1 standard vswitch with VLANs, so traffic should be separated. What load balance technology should i choose on host side?
Please help to choose the right scheme for this situation.
Message was edited by: Denis Mukhamedov
2013-12-19 05:13 AM
Seems you have misinterpreted the function of the Cluster Interconnect - No host will access data over the interconnect .This cable if for communication between the cluster partners who replicate each other's NVRAM contents before being flushed to disk.
In you case - I think you can connect each 10 GB port on one filer head (say filer A) to both the switches (for redundancy and load balancing).I am sure you will connect the other filer heads network ports also to these two switches (for vfilers to be created in filer B).Include the partner statement in the RC file and whenver a failover happens - the traffic will be passed over from the partner filer's interface as the resources and vfilers will failover from filer A to filer B?
Make sure that whichever VLAN's are present on Filer A (which is the primary location of the ESX datastores) - the same are created in the Filer B as well as the second switch to ensure that the hosts can still reach the vfiler IP;s in that particular VLAN.
Let me know if this helps or not
2013-12-19 05:15 AM
Note that I am assuming that all of your data is being accessed (i.e the aggregate and volumes are created as well as the VM hosts are accessing some vfiler ) in Filer A only and Filer B is used for other purposes/applications/DB's etc.
2013-12-19 05:34 AM
Thank you for response.
Yes, all data is being accessed in Filer A only. So if I will connect port 1 of Filer A to switch A and port 2 of Filer A to switch B, I will have connect speed at 10Gb.
And if I use trunking (on case of connecting two port to switch A) I can get 20Gb link.
I just checked, HP 6120XG switches have 2 10Gb interconnect ports, so I have redundancy in network.