Hi,We have a few pools floating around where the workers are on a private subnet and the machines hosting the schedds are dual-homed. Sensibly, the system admins have been setting NETWORK_INTERFACE to the private subnet so that the schedd automatically binds to the private subnet.
What was surprising was that NETWORK_INTERFACE constraints not only services using bind() but also outgoing network connections. In other words, a user running condor_status to a foreign pool cannot succeed, as it attempts to use the private subnet.
I know the workaround is to enable BIND_ALL_INTERFACES -- but is the above behavior intentional or just accidental?
- B