[HTCondor-users] prefer ipv4 without disabling ipv6

[Michael Thomas <wart@xxxxxxxxxxx>]

We're starting to test ipv6 at our datacenter.  We recently upgraded to 
condor-23.4.0 and added ipv6 addresses to the public interfaces on all 
access points.  After doing so, I found that one of our monitoring 
servers could no longer run 'condor_q -name xxx -pool xxx' (command 
hangs and eventually times out).

On the access points we set:

NETWORK_INTERFACE=<host private IP address>

On the misbehaving monitoring server, we have:

NETWORK_INTERFACE=<host private IP address>

On both systems we have a firewall that blocks any incoming connections 
over the public interface on ports 9618, 9620, among other things.

If I disable the public ipv6 address on the monitoring server (or set 
ENABLE_IPV6=False), the condor_q command works as expected.

Using strace, I found that if ipv6 is enabled, the 'condor_q' command 
tries to connect to the access point using the public ipv6 address:

2893155 connect(3, {sa_family=AF_INET6, sin6_port=htons(9618), 
sin6_flowinfo=htonl(0), inet_pton(AF_INET6, "2607:f390:3ff0:2016::79", 
&sin6_addr), sin6_scope_id=0}, 28) = -1 EINPROGRESS (Operation now in 
progress)

...which will certainly not work (as expected and intended) due to the 
firewall.

I've tried various combinations of BIND_ALL_INTERFACES and setting 
NETWORK_INTERFACE=<host private interface>, but that doesn't seem to 
change the fact that the offending client tries to talk to the access 
point over the public ipv6 address instead of the preferred private 
internal ipv4 address.  Is it possible to get the condor_q command to 
prefer the internal ipv4 address without setting ENABLE_IPV6=False?

--Mike