[openstack-dev] [oslo] using transport_url disables connection pooling?

Gordon Sim gsim at redhat.com
Thu Apr 10 20:02:46 UTC 2014


This may be a known and accepted fact, but it was a surprise to me so 
thought it was worth a mention on this list.

If you use the transport_url config option (or pass a url directly to 
get_transport()), then the amqp connection pooling appears to be 
disabled[1]. That means a new connection is created for every request 
send as well as for every response.

So I guess rpc_backend is the recommended approach to selecting the 
transport at present(?).

[1] I think this is because if a transport url is specified then the url 
will always have at least the hosts set, so the AMQPDriverBase will 
always have a non-empty map for server_params, which overrides any use 
of pooling.



More information about the OpenStack-dev mailing list