RabbitMQ loses everything on any custom config

asked 2016-05-20 08:47:41 -0500

al_indigo gravatar image

updated 2016-05-20 09:40:17 -0500

I have working liberty openstack cluster. Initially we have set up everything with rabbitmq default config (there was no /etc/rabbitmq/rabbitmq.conf at all) and everything worked.

Now I want to make some custom changes and after /etc/rabbitmq/rabbitmq.conf creation, the whole cluster can not even connect to rabbitmq (but it works and listens its interfaces). From client side (nova-compute for example) it looks like thousands of errors like this one:

2016-05-20 09:17:10.998 21198 ERROR oslo.messaging._drivers.impl_rabbit [req-425b4cc8-e530-4f25-94d0-bff6452f2e19 - - - - -] AMQP server 10.10.21.31:5672 closed the connection. Check login credentials: Socket closed
2016-05-20 09:17:11.544 21198 ERROR oslo.messaging._drivers.impl_rabbit [-] AMQP server 10.10.21.31:5672 closed the connection. Check login credentials: Socket closed

The config I want to apply is rather simple:

[
 {rabbit,
  [%%
    {tcp_listeners, [{"10.10.21.31", 5672},
                      {"127.0.0.1",   5672}]},
    {num_tcp_acceptors, 64},
    {reverse_dns_lookups, false},
    {tcp_listen_options, [{backlog,       8192},
                          {nodelay,       true},
                          {sndbuf,    196608},
                          {recbuf,    196608},
                          {exit_on_close, false}]}
  ]},
 {kernel,
  [%% Sets the net_kernel tick time.
  ]},
 {rabbitmq_management,
  [%% Pre-Load schema definitions from the following JSON file. See
    {listener, [{port,     15672},
                {ip,       "10.10.15.31"} ]}
  ]},
 {rabbitmq_shovel,
  [{shovels,
    [%% A named shovel worker.
    ]}
  ]},
 {rabbitmq_stomp,
  [%% Network Configuration - the format is generally the same as for the broker
  ]},
 {rabbitmq_mqtt,
  [%% Set the default user name and password. Will be used as the default login
  ]},
 {rabbitmq_amqp1_0,
  [%% Connections that are not authenticated with SASL will connect as this
  ]},
 {rabbitmq_auth_backend_ldap,
  [%%
  ]}
].

After restart rabbit with this config all the queues, messages etc disappear. If I delete the config and restart rabbit without it, everything returns to its place (and the cloud becomes operational again).

It looks like I'm missing something stupid about it because I cannot google anything related. Have anyone seen something like this before?

edit retag flag offensive close merge delete

Comments

I've isolated the piece of config that causes error: if I use any tcp_listen_options (even default ones) explicitly, the whole openstack can't talk with RabbitMQ. But I don't have a clue how to explain it

al_indigo gravatar imageal_indigo ( 2016-05-20 11:20:53 -0500 )edit