Network connectivity in Sandbox moddeling labs broken

[e.nieuwstad]

I have been testing with the moddeling labs sandbox a couple of times but when I load my own lab the connectivity is often broken. All the connectivity is fine when I start the lab with the default topology but the moment I load my own connectivity via the unmanaged switch is not working or only for a few hosts after a long time. When I logon to the console I see that the devices don't have an arp entry for the default gateway of 10.10.20.254. Connectivity between nodes via the unmanaged switch is also intermittent at best. 
I really like the product but this makes it hard to use for me

[bigevilbeard]

This is a new one, what happens when/if you clear the arp on these ones?

[e.nieuwstad]

My lab has expired at the moment. But the weirdest thing is when I do a packet capture on the link to the unmanaged switch I do see the arp request from the device but no answer from the def gw. As also the connectivity between the various devices is a unreliable I suspect it is the unmanaged switch causing the issue. I will spin up a new lab and test some more.

[Jesus Illescas]

I had similar experiences, I have the impression somewhere outside the CML instance there is a L2 component that is not updating/forwarding the arp correctly.

There are several way I try to make arp work.

• Shut/unshut management interfaces connected to the unmmanged swtich.
• Ping the GW (10.10.20.254) from the mgmt interface of the devices.
• Ping/ssh from the DevBox VM towards the mgmt IP of the devices (10.10.20.x subnet)
• Ping from the mgmt IP of the devices towards the DevBox VM.
• Ping between the mgmt IP of the devices to each other.

Usually a combination of the actions above fixed the issue. In some cases, I reset the lab but is not that common.

[e.nieuwstad]

I have created a new and shiny modelling labs sandbox and all connectivity is working from the start then. I exported the default lab, stopped the default lab and imported it as a new one and it is broken. You can see the arp request for the gateway but no reply

[e.nieuwstad]

I just did some more testing. Adding a static arp entry on the virtual routers and switches did fix it somewhat but not reliable however it really seems related to some arp process or something related to forwarding broadcasts within the virtual network

[bigevilbeard]

@rkochery have you seen this issue before please?

[rkochery]

Hi @bigevilbeard , Checking in the Sandbox webex teams space, got pointer that vrf is used in sandbox, Can you try ping vrf Mgmt-intf 10.10.20.254

[e.nieuwstad]

Yes I used the correct vrf. Also the ping initiated from my workstation which has the vpn connection to the lab fails. The connectivity from my workstation to the modeling labs server is working fine.