Tests: Fix test node_server.bind_same_p2p_port from randomly crashing
This commit is contained in:
parent
9a5579fc4f
commit
176cea0ec6
|
@ -293,6 +293,18 @@ TEST(node_server, bind_same_p2p_port)
|
|||
boost::program_options::variables_map vm;
|
||||
boost::program_options::store(boost::program_options::parse_command_line(1, argv, desc_options), vm);
|
||||
|
||||
/*
|
||||
Reason for choosing '127.0.0.2' as the IP:
|
||||
|
||||
A TCP local socket address that has been bound is unavailable for some time after closing, unless the SO_REUSEADDR flag has been set.
|
||||
That's why connections with automatically assigned source port 48080/58080 from previous test blocks the next to bind acceptor
|
||||
so solution is to either set reuse_addr option for each socket in all tests
|
||||
or use ip different from localhost for acceptors in order to not interfere with automatically assigned source endpoints
|
||||
|
||||
Relevant part about REUSEADDR from man:
|
||||
https://www.man7.org/linux/man-pages/man7/ip.7.html
|
||||
*/
|
||||
vm.find(nodetool::arg_p2p_bind_ip.name)->second = boost::program_options::variable_value(std::string("127.0.0.2"), false);
|
||||
vm.find(nodetool::arg_p2p_bind_port.name)->second = boost::program_options::variable_value(std::string(port), false);
|
||||
|
||||
boost::program_options::notify(vm);
|
||||
|
|
Loading…
Reference in New Issue