Comware

Hello!

We are using 5940 as agregation switch with many vlan interfaces such as:

interface Vlan-interface555
ip address 10.10.10.1 255.255.255.0
dhcp select relay
dhcp relay server-address algorithm master-backup
dhcp relay master-server switch-delay 3
dhcp relay server-address 10.10.20.4
dhcp relay server-address 10.10.20.5

Problem is - too much log messages about DHCP switchover between master and backup:

%Feb 21 12:06:44:895 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.4 because the current DHCP server did not respond.
%Feb 21 12:06:58:995 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.5 because the current DHCP server did not respond.
%Feb 21 12:07:05:295 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.4 because the current DHCP server did not respond.
%Feb 21 12:08:03:395 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.5 because the current DHCP server did not respond.
%Feb 21 12:08:13:295 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.4 because the current DHCP server did not respond.
%Feb 21 12:08:21:995 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.5 because the current DHCP server did not respond.
%Feb 21 12:08:22:595 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.5 because the current DHCP server did not respond.
%Feb 21 12:08:24:695 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.5 because the current DHCP server did not respond.
%Feb 21 12:08:46:295 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.5 because the current DHCP server did not respond.
%Feb 21 12:09:23:095 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.4 because the current DHCP server did not respond.
%Feb 21 12:09:37:595 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.4 because the current DHCP server did not respond.

There is no packet loss between switch and DHCP server:

PING/6/PING_STATISTICS: Ping statistics for 10.10.20.4: 10000 packet(s) transmitted, 10000 packet(s) received, 0.0% packet loss, round-trip min/avg/max/std-dev = 0.950/1.873/62.192/4.801 ms.
PING/6/PING_STATISTICS: Ping statistics for 10.10.20.5: 10000 packet(s) transmitted, 10000 packet(s) received, 0.0% packet loss, round-trip min/avg/max/std-dev = 0.725/0.877/9.558/0.271 ms.

Situation didnt change even using timeout and delay values more longer than default. Log messages appear but rarely, every 1-5 minutes depend of timeout or switch-delay values:

dhcp relay dhcp-server timeout 240
dhcp relay master-server switch-delay 10

%Feb 27 12:40:14:095 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.5 because the current DHCP server did not respond.
%Feb 27 12:40:30:395 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.4 because the current DHCP server did not respond.
%Feb 27 12:42:34:595 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.4 because the current DHCP server did not respond.
%Feb 27 12:45:04:395 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.4 because the current DHCP server did not respond.
%Feb 27 12:45:32:395 2018 switch1 DHCPR/3/DHCPR_SERVERCHANGE:
Switched to the DHCP server at 10.10.20.5 because the current DHCP server did not respond.

How switch determine DHCP master did not respond? Which protocol?

Thanks!

Your switch mode, "dhcp relay server-address algorithm master-backup" is probably not the best way of doing it if your Windows DHCP cluster is in active-active "load-balancing" mode.

If the DHCP servers are "active-active" then either one of them could respond, but your switch is only forwarding the request to one of them. If it forwards it to the wrong one, that DHCP server will calculate that its active partner should be responding and it will not respond. Because you haven't send the request to both servers, no response is sent.

Yes, actually we are using Windows DHCP cluster in active-active load-balancing mode.

I will try to configure DHCP relay with polling algoritm.

Thanks for your answear!

Polling algorithm works good! No messages, ip acquiring much faster than before.