Networking with Dell Switches: VLANs and NICs


Userlevel 2
Badge +3
Hi All

Is anybody else just going round and round in circles with Nutanix and the network it resides on? I am getting soooo peeeeeee'd off!! The man at the show said easy peasy, 15 minutes to deploy, which 15 minutes? The last 15 in 2019?

Dell supplied, but don't really know the product so they pass calls off to Nutanix who do their bit (very well, no complaint there) but then hand the things about Dell back to Dell, so I am not getting a joined up answer. So the crux is that I don't have a stable network and I am ready to rip it all out and go back to VMware.

Unfortunately my budget has gone and I have to rely on Communities for answers - not a lot forthcoming.

Is anyone else having these sort of problems or is it just me?

7 replies

Userlevel 7
Badge +26
Are you asking in the Install and Config forum? Community Edition is a bit of a different beast from commercial and uses Nutanix's KVM fork on commodity hardware. This sounds like Dell XC which should have support and Nutanix to back it up when they reach their limit.

So whats the issue with network you are having specifically?

@aluciani doesn't seem like this case is getting moved from Dell over to Nutanix for escalation for some reason. Not sure if the Nutanix support folks can help with XC?
Userlevel 7
Badge +35
Thanks for flagging @jrack

Let me see what I can do here @Eric-The_Viking
Userlevel 2
Badge +3
Hi ,
The current issue is:

I have 2 x Dell X4012 switches with 10gb SPF+ going to my Nutanix Hosts, the X4012's are connected to a N3024 stack. I had to change the Nutanix vlan and subnet after install and add some other vlans for production servers. I had some problems with dynamic arp and Dell fixed that, however in the fixing that problem another has come from the changes. I can ping all the Nutanix Hosts and CVMs, but I am unable to launch the Prism Central Web Console - I have done some further investigation and find that the old vlan is still required in the VM Nic 2 (See pictures)



So Nic1 has the correct Vlan but no IP and Nic2 has the correct IP but wrong Vlan as 0 was a temporary fix while the host IPs were changed.

I also have an oddity in the Network view in Prism - I have 2 + 1 switch in the network but the Network view is showing 5 switches - 3 now showing as NONE but showing connections that should be within the switches that I add snmp for.



Thirdly - I have lost routing from the Nutanix environment to the outside - NTP, DNS, updates etc.

I currently have call open with Dell to clear random entries on the 2 x X4012 switches (engineer will remote in on Tuesday) - it seems the GUI has a mind of its own when adding entries into the Running Config. The original call was to fix the problem with a non contactable CVM - that was passed to Nutanix who determined the Nutanix environment was ok but the X4012 was not dynamically picking up the ARP addresses from the CVM - he entered a static address and ping started, thus it was passed back to Dell to fix.

I just feel I am getting nowhere and I can't get joined up answers, some of the problem could be that my past knowledge has been with vSphere and I am struggling to map what I did for that to what I need to do now in Nutanix.

Regards

E.
Userlevel 7
Badge +26
So sounds like you are using AHV and the underlying OVS for your deployment and not on vSphere on the XC. Yeah this seems to be noise from the integration of the upstream L2 switches w AOS which honestly I have never used. I just setup the VLANs on my L2s and configure the parts as appropriate, but I get that is a feature you paid for w Dell.

And there is no security features on those switches to filter off ARPs as potential spoofing? The CVM and Central VMs are on an OVS so maybe their inspection is just confused?

Do the subnet/gateway look good on your interfaces? Assume the VLANs on the switches are configured on an upstream router and gateway looks right? Just thinking basic ifconfigs in each of the shells to match what you expect to see on the switch and routers.

Any idea on the "static address" that was added as that seems like a kludge and the next switch recycle may drop that?

And can you ping the IP either configured or assigned to Central from an AHV node? Sounded like the CVMs were reachable from members of your subnet. Just trying to figure out if there are L2 issues or just L3 issues. I mean Central is just another VM so not sure why you could reach the CVM clusterIP, but not the Central IP. And also inbound (your Prism UI) seems to work, but outbound doesn't?
Userlevel 7
Badge +35
Hi @Eric-The_Viking

I just emailed you - lets see if we can get you some help. 👍
Userlevel 2
Badge +3
jrack wrote:

So sounds like you are using AHV and the underlying OVS for your deployment and not on vSphere on the XC. Yeah this seems to be noise from the integration of the upstream L2 switches w AOS which honestly I have never used. I just setup the VLANs on my L2s and configure the parts as appropriate, but I get that is a feature you paid for w Dell.


I think some of the problems come from changing the VLANs and subnet for the AHV and CVM management, following the Nutanix best practices.

And there is no security features on those switches to filter off ARPs as potential spoofing? The CVM and Central VMs are on an OVS so maybe their inspection is just confused? No security features that I can see, and if that was the case the hosts and CVMs would not ping either - very confusing.
Do the subnet/gateway look good on your interfaces? Assume the VLANs on the switches are configured on an upstream router and gateway looks right? Just thinking basic ifconfigs in each of the shells to match what you expect to see on the switch and routers.I checked the gateways and they all appear to be correct.
Any idea on the "static address" that was added as that seems like a kludge and the next switch recycle may drop that?I thought that too, so I cleared the ARP tables on all the physical switches - the MAC address returns to the ARP tables but no ping, very odd!
And can you ping the IP either configured or assigned to Central from an AHV node? Sounded like the CVMs were reachable from members of your subnet. Just trying to figure out if there are L2 issues or just L3 issues. I mean Central is just another VM so not sure why you could reach the CVM clusterIP, but not the Central IP. And also inbound (your Prism UI) seems to work, but outbound doesn't?I have a call open with Nutanix via Dell now, hopefully this afternoon will reveal what is going awry.




Thank you for responding to my plight, I don't seem to have much luck on the forums regarding Nutanix.

cheers
E.
Userlevel 2
Badge +3
Eric-The_Viking wrote:


jrack wrote:

So sounds like you are using AHV and the underlying OVS for your deployment and not on vSphere on the XC. Yeah this seems to be noise from the integration of the upstream L2 switches w AOS which honestly I have never used. I just setup the VLANs on my L2s and configure the parts as appropriate, but I get that is a feature you paid for w Dell.

I think some of the problems come from changing the VLANs and subnet for the AHV and CVM management, following the Nutanix best practices.

And there is no security features on those switches to filter off ARPs as potential spoofing? The CVM and Central VMs are on an OVS so maybe their inspection is just confused? No security features that I can see, and if that was the case the hosts and CVMs would not ping either - very confusing.
Do the subnet/gateway look good on your interfaces? Assume the VLANs on the switches are configured on an upstream router and gateway looks right? Just thinking basic ifconfigs in each of the shells to match what you expect to see on the switch and routers.I checked the gateways and they all appear to be correct.
Any idea on the "static address" that was added as that seems like a kludge and the next switch recycle may drop that?I thought that too, so I cleared the ARP tables on all the physical switches - the MAC address returns to the ARP tables but no ping, very odd!
And can you ping the IP either configured or assigned to Central from an AHV node? Sounded like the CVMs were reachable from members of your subnet. Just trying to figure out if there are L2 issues or just L3 issues. I mean Central is just another VM so not sure why you could reach the CVM clusterIP, but not the Central IP. And also inbound (your Prism UI) seems to work, but outbound doesn't?I have a call open with Nutanix via Dell now, hopefully this afternoon will reveal what is going awry.


Thank you for responding to my plight, I don't seem to have much luck on the forums regarding Nutanix.

cheers
E.




As it turned out the PC VM was in the incorrect vlan - my worst nightmare was confirmed, the reconfiguring of the network left all the hosts and CVMs in the old VLAN but in the new subnet. The PC VM was put into the correct VLAN with the correct subnet but no route to get to vlan0 - what a mess. So the engineer did a sterling job and put the PC VM into the vlan0 and now it talks to the rest of the Nutanix environment but not the rest of the world to get updates, time and whatever. Back to the drawing board...

Reply