VMware Cloud Community
iscsitarget
Contributor
Contributor
Jump to solution

Odd network behavior when router changes

Hi All,

On one of our subsites we've got an ESXi host connected to a central vCenter server. The location has a main router with a failover router in case the main connection fails.

Main Router: 192.168.66.254

Backup router: 192.168.66.253

Core switch: 192.168.66.250

On the ESXi host the default gateway is set to the core switch ip: 192.168.66.250 (just like every other device in the network)

Our main link failed in the night and the backup link took over, everything (pc's printers, virtual servers etc) was working as expected on the backup link, but the ESXi host was unreachable in vCenter. I could ping the ESXi server from my laptop, but could not do it from the console on the vCenter server.

I logged on with SSH to the ESXi host and ran:

esxcli network ip route ipv4 list

Network        Netmask          Gateway         Interface  Source

------------   -------------    ------------    ---------  ------

default        0.0.0.0          192.168.66.250  vmk0       MANUAL

192.168.27.12  255.255.255.255  192.168.66.254  vmk0       MANUAL

192.168.28.2   255.255.255.255  192.168.66.254  vmk0       MANUAL

192.168.17.154 255.255.255.255  192.168.66.253  vmk0       MANUAL

I've noticed the default gateway for the DNS server (192.168.27.12) and vCenter (192.168.28.2) had the main router set instead of the core switch, and also that my pc (192.168.17.154) was listed with the backup gateway. I ran the route remove command to remove the routes and the ESXi host became visible in vCenter again.

I double checked the settings, it all was fine, gateway was indeed set to the core switch.

Now the main link is back up so the failover link stops working, guess what, the ESXi host was again not connected into vCenter, tried to ping it from the vCenter and my laptop, no replies from them.

I logged on to another computer and I was able to ping and ssh into the ESXi host

esxcli network ip route ipv4 list

Network        Netmask          Gateway         Interface  Source

------------   -------------    ------------    ---------  ------

default        0.0.0.0          192.168.66.250  vmk0       MANUAL

192.168.27.12  255.255.255.255  192.168.66.253  vmk0       MANUAL

192.168.28.2   255.255.255.255  192.168.66.253  vmk0       MANUAL

192.168.17.154 255.255.255.255  192.168.66.253  vmk0       MANUAL

192.168.17.179 255.255.255.255  192.168.66.254  vmk0       MANUAL

So once again I manually removed the routes and everything was functioning again.

I simulated the failover in our test lab and it had the exact same behaviour, it looks like ESXi takes on gateway and then never releases that IP.

any thoughts on this odd behavior? I expected ESXi to failover just like any other pc/printer/server etc in our network would do, but clearly it doesn't

Thanks for your input!

0 Kudos
1 Solution

Accepted Solutions
Nick_Andreev
Expert
Expert
Jump to solution

What you're experiencing is called ICMP redirect. You can read more about it here: https://kb.vmware.com/kb/2081185

Disabling ICMP redirect on your switch should fix your issue.

---
If you found my answers helpful please consider marking them as helpful or correct.
VCIX-DCV, VCIX-NV, VCAP-CMA | vExpert '16, '17, '18
Blog: http://niktips.wordpress.com | Twitter: @nick_andreev_au

View solution in original post

2 Replies
iscsitarget
Contributor
Contributor
Jump to solution

bump Smiley Happy

0 Kudos
Nick_Andreev
Expert
Expert
Jump to solution

What you're experiencing is called ICMP redirect. You can read more about it here: https://kb.vmware.com/kb/2081185

Disabling ICMP redirect on your switch should fix your issue.

---
If you found my answers helpful please consider marking them as helpful or correct.
VCIX-DCV, VCIX-NV, VCAP-CMA | vExpert '16, '17, '18
Blog: http://niktips.wordpress.com | Twitter: @nick_andreev_au