What is the going recommendation these days when setting up a new volume? Do people still recommend using NFS v3 over 4.1? And if so, how is load balancing handled between the vCenter datastore and the SPs?
#NFS 4.1 vs NFS 3 with vCenter
harsh tundra
feral pivot
We are still using v3 because VMware does not support SDRS and v4. Also with ONTAP 9.12 session trunking for the multipathing only works with LIFs on the same node
flat forum
Same for the same reasons. No supported failover is possible, it may works but that’s not good enough for me in an enterprise environment
prime aurora
" No supported failover is possible", is there any official statements from NetApp to state that? We have the same question/concern and there are a lot of saying differently.
sweet latch
<ping> @dawn ember
solar tiger
Same for the same reasons. No supported failover is possible, it may works but t...
It's supported. ONTAP and your ESXi-Hosts need to be on certain versions but it's definitely supported. If it's stable enough to use v4.1 vs. v3 which is working perfectly fine... I have my doubts.
Check HWU and these KBs:
https://kb.netapp.com/Advice_and_Troubleshooting/Data_Storage_Software/ONTAP_OS/VMWARE_Virtual_Machines__went_offline_during_upgrade
https://kb.netapp.com/Advice_and_Troubleshooting/Data_Storage_Software/ONTAP_OS/VMware_NFSv4.1_datastores_see_disruption_during_failover_events_for_ONTAP_9
harsh tundra
It's supported. ONTAP and your ESXi-Hosts need to be on certain versions but it'...
Okay, so if I have HA and use NFS v3, is the best route for load balancing between the SPs using DNS load balancing?
solar tiger
What do you mean with SPs?
harsh tundra
right now I have mounts pointing to x.x.x.20 thats the NFS share on SP1. but not to SP2 on x.x.x.21 Is there a way to get them to share the load? It seems one is idle the other is doing all the work.
flat forum
It's supported. ONTAP and your ESXi-Hosts need to be on certain versions but it'...
Of course it is supported, but there is no session trunking or pNFS or any other failover thing implemented in nfs4.x which is supported by NetApp and ESXi.
It will mostly work since ESXi removed that long delay for rebuilding the session in ESXi 6.0 or 6.5.
But it is a session based protocol and there is no session handling between the ONTAP nodes.
harsh tundra
or is it because the data partition is owned by a single sp
solar tiger
You are mixing things up. What I think you mean by "SP" is a node or controller.
harsh tundra
service processor this is what im seeing where one is taking all the load https://i.imgur.com/lZ24URW.png
solar tiger
right now I have mounts pointing to x.x.x.20 thats the NFS share on SP1. but not...
If you for example have 8x volumes which you want to use as NFS-datastores. They all reside on one aggr which belongs to nodeA. For 4x volumes you could use the NFS-LIF from nodeA to mount the datastores. For the other 4x simply use the NFS-LIF from nodeB.
Now you would at least share some of the load between the nodes. But half of your traffic would need to traverse the Cluster-LIFs to get to the other node which owns the aggr. (For current models the latency introduced by the cluster-backend is negligible.)
If you really want to use the compute-power of nodeB you need to create an aggr on nodeB. And then move volumes to it.
harsh tundra
Okay, that makes sense. Right now I configured one volume to be used. So I should create a second for load balancing then. This is a pretty simple configuration with one aggr for esxi
solar tiger
service processor this is what im seeing where one is taking all the load https...
The Service Processor is not involved in client-/frontend-traffic. It's simply an out-of-band processor included on the motherboards.
Think iLO, CIMC, iDRAC, etc.
https://kb.netapp.com/Advice_and_Troubleshooting/Data_Storage_Systems/FAS_Systems/What_is_a_Service_Processor_and_how_do_I_use_it
solar tiger
Okay, that makes sense. Right now I configured one volume to be used. So I shoul...
How many aggregates do you have?
harsh tundra
One aggregate, a 2 svms and a few volumes
solar tiger
Then creating a new volume and mounting it via a NFS-LIF from the other node is the only way to at least share some load with the other node. But this will be minimal.
Do you have performance-issues currently? It doesn't look like your node which owns the aggregate is getting limited by the CPU with that low utilization of ~8%.
Create a performance case with NetApp if you are not seeing the performance you are expecting.
harsh tundra
No, this is a new device. Migrating from a FAS2552 to this hybrid array