Hi,
I'm struggling to get NIC teaming (aggregating) between a ESXi 6.5 host and HPE switches.
Here are my VSwitch settings (interfaces inherit their settings from vswitch)
Link discovery, mode: Listen
Link discovery, protocol: Cisco discovery protocol
Security, Promiscuous mode: Reject
Security, MAC address changes: Reject
Security, Forged transmits: Reject
NIC teaming, Load balancing: Route based on IP hash
NIC teaming, Network failover detection: Link status only
NIC teaming, Notify switches: Yes
NIC teaming, Failback: Yes
NIC teaming, Failover order:
vmnic4 - 1000Mbps - Active
vmnic5 - 1000Mbps - Active
vmnic6 - 1000Mbps - Active
vmnic7 - 1000Mbps - Active
Traffic shaping: Disabled.
I think my vmware settings are more or less correct, or what?
However I need to know more about how to set up the HP switches, if LACP is used or not etc...
First HP switch: HPE 1920S-24G - JL381A
Menu item: Trunks / Configuration
Have configured three trunks with 4 ports each. All with a bunch of tagged VLANs.
Here are the default trunk settings:
Admin mode: Enabled
STP Mode: Disabled
Static Mode: Enabled (should it be dynamic to allow LACP?)
Load balance (5 choices, what do I select here??):
Source MAC, VLAN, Ethertype, Incoming port
Destination MAC, VLAN, Ethertype, Incoming port
Source+Destination MAC, VLAN, Ethertype, Incoming port
Source IP and Source TCP-UDP Port fields
Destination IP and Destination TCP-UDP Port fields
Source+Destination IP and Source+Destination TCP-UDP Port fields
Second switch: HPE V1910-24G
Menu item: Network / Link aggregation
Have configured three trunks with 4 ports each. All with a bunch of tagged VLANs.
The default setting for each trunk is:
Aggregation interface type: Static
(should I enable LACP for the ports here?)
If I should use LACP, it has its own menu item where I can select which switch ports to enable for LACP and the priority between the system and the LACP ports.
Menu item: VLAN / Modify Port:
Selected all trunks and changed all their ports 'Link type' from 'Access' to 'Trunk' type port.
I have played a lot with this and ended up with no traffic flow at all or only traffic via one of the interfaces...
Thanks a lot for comments on which settings are correct or if I need to do something else...
Tor
LACP is only possible when using a virtual distributed switch (vDS). That aside, unless you have a very specific requirement and also know in a detailed fashion your network traffic flows, you should use load-based teaming (LBT) instead with a vDS and forget about LAGs.
Please take a look at https://kb.vmware.com/s/article/2006129 for pros and cons as well as a link to a sample configuration with HPE switches.
Based on this you can decide whether it makes sense to use it, or to rather go for with a switch independent configuration, which provides redundancy in case of e.g. a switch failure/reboot.
André
Moderator: Moved to vSphere vNetwork
Thanks for the comments. However I am not in the position to upgrade so I get vDS now and I have understood that I cannot use LACP unless I use vDS.
I read the links to articles for HP and Cisco switches but they used LACP too...
So isn't there really ANY possibility to do NIC teaming / aggregating between my ESXi 6.5 vSwitch and the HPE switches with the properties I mention in my original post?
If NOT can you suggest other switch model(s) where this feature is possible..:?
Thanks again
Tor
ESXi has native teaming built-in and there is nothing else you need to do. Use of a LAG is not necessary if you want connection sharing and failover abilities.
It's basically not a limitation on the hardware switches, but in the software, so that other switches won't help.
Anyway, using multiple uplinks on a Standard vSwitch with the default policy does a round robin assignment for VMs once they are powered on.
André
Really, so I don't need to do anything on the switch side, just connect the ESXi teamed NICs to ordinary switch ports and there will be no looping problems or the like...?
That was really good news for me, - I have always thought that teaming required config on both sides of the link...
regards Tor
Yes, that's correct. ESXi will not create loops naturally regardless of the teaming configuration.
Great. Thanks for that confirmation.
However do I really get augmented bandwidth benefit with this solution (since it is 'active' only from one side of the link) or does this solution only give me failover benefit?
best regards
Tor
With the default configuration the VMs are distributed across all uplinks (i.e. no aggregation in any way). This means that multiple uplinks are active, and in case of an uplink failure, the VMs assigned to this uplink will transparently moved to another one. If needed, you can manage the behavior in the Teaming, and Failover Settings for the vSwitch and/or the Port Groups.
André