I have LACP'ed 100G ports for serving both cifs/nfs traffic and there is new requirement to configure the iscsi traffic . Is it ideal to use the same network ports, if not what is the best practice for doing this and appreciate for any documentation? I read somewhere that iscsi doesn't support the 100G ports it is for NVME but would like to confirm. #1062049107096633454 #1062049169520476220
#Can we use same network ports (100G) which is being used for cifs/nfs to configure the new iscsi.
1 messages · Page 1 of 1 (latest)
it's technical doable, I'd put the iscsi on it's own VLANs
And check the useage as well, could have some perf impact if it's already high.
@rocky haven - Is there any documenation which suggest to use the dedicated physical ports and is it required to use LACP on the iscsi?
Let me see the best doc for your question. but no, dedicated ports for iscsi are only required if you do a direct attachement connection config.
@rocky haven - We plan to use the network switch for this production traffic for redundancy and i feel direct attachment connection is little riskier provided we have vmhost in different subnets.
So with out direct attachment we need dedicated ports for iscsi?
That's fine. I only mentioned the direct connect because it's also in the docs. (direct is not really the norm config).
ideally for iscsi you also want a pair of switches,
there are some ilistrations here -
https://www.cisco.com/c/en/us/td/docs/unified_computing/ucs/UCS_CVDs/flexpod_esxi65u1_n9kiscsi.html
Cisco
Deployment Guide for FlexPod Datacenter with IP-Based Storage using VMware vSphere 6.5 U1, NetApp AFF A-Series, and Cisco UCS Manager 3.2
@inner bluff any idea of any better docs to show the config that's being asked about
When you say switches it is the network switch right? If so yes currently the cifs/nfs traffic is being routed with the network switch .
I am worried about is it advised to share the iscsi traffic with the current NAS traffic is that recommended or is it always better to use dedicated physical ports for iscsi.
correct.
back when i did PS, I did both configs for customers.
dedicated ports, because they plugged them directly into the vmware networks.
And shared using VLANs.
When you say dedicated ports does that mean direct attachment connection plugged in to vmware networks?
or Dedicated ports which on netapp physical ports but directly plugged in to vmware network with out the network switch? @rocky haven
Nope. just dedicated ports off the back of the netapp. so on a 4 port card for example. 2 were iSCSI and 2 were used for NAS traffic.
Got it. thank you for the details. Is there any documentation that points to use a dedicated physical ports ? I did search on netapp portal could not find definitive links pointing to that?
I couldn't find that specific requirment either (why i tagged TMAC). but I'd think that if the ports alrady have high utilization from existing NFS traffic or planning to have high utlization. then use seperate ports.
I could have used if i have dedicated ports for NFS traffic as per the below link, but i have same ports used for both cifs/nfs, on each node 2 ports LACP'ed as a0a and using for both the cifs and NFS traffic.
https://kb.netapp.com/on-prem/ontap/da/SAN/SAN-KBs/Is_it_recommended_to_have_ISCSI_and_NFS_configuration_on_same_link
I'd think with 100G you'd be fine with traffic honestly. (I just put in an ask to update the doc for 100G).
In nearly every case, I will ask the customer to create an LACP channel on the switch or ask them to let me do it. Then I TAG EVERYTHING. I set the base interface to an unused vlan ( the broadcast domain is affectionately called DoNotUse with a mtu of 9000).
Everything gets a vlan.
Iscsi
Nvme/tcp
nfs public
NFS private (VMware)
The only thing with iscsi that is NOT supported is having the host use iSCSI over a port channel. Windows, Linux and VMware all either discourage or don’t support it
I have had maybe two cases where the network guy was just simply afraid of doing port channels (past experience with improperly configured channels)