I solved this issue with some prompting from Allen. This is what I wanted NAS send acct req to radius R1 R1 replicate acct req to R2 R1 respond to NAS I simply could not get R1 to respond to the NAS when the replicate was in the configuration, so the NAS would continue to send acct req packets to R1 and after a few trys, mark R1 as down. It appears to be an issue with freeradius-2.1.12. Upgrading to 2.2.4 seems to have fixed the issue. Here are some details for those interested. I make no claims about this being the best way, it's just "a way". Building a 2.2.4 rpm for my version of centos was a bit of trouble as well. I followed this and got it done: http://adriansmith.blogspot.co.uk/2014/01/building-freeradius-rpm-for-el5.ht... Follow the instructions but change the version and release in the spec file to values that match your freeradius source rpm. I had to add one additional step which was to add this line to the spec file: %config(noreplace) /etc/raddb/panic.gdb To configure radius to do a replicate (fire and forget) to an accounting server: in sites-enabled/default in the accounting { } section, add update control and replicate items like this: # daily update control { Replicate-to-Realm='Acct2' } replicate # Update the wtmp file in proxy.conf, define a home_server, a pool, and a realm as follows (use your IP and password values) home_server_acct2 { type = acct ipaddr = 10.10.1.1 port = 1813 secret = testing123 # src_ipaddr = 127.0.0.1 require_message_authenticator = yes response_window = 20 # no_response_fail = no zombie_period = 40 revive_interval = 120 status_check = status-server check_interval = 30 num_answers_to_alive = 3 max_outstanding = 65536 coa { irt = 2 mrt = 16 mrc = 5 mrd = 30 } } home_server_pool acct2_pool { type = fail-over home_server = acct2 } realm Acct2 { acct_pool = acct2_pool }