Make sure that the fabric's names in eSwitchd, Neutron agent and OpenFlow are the same.

The Neutron agent fails to start

Check if the eSwitchd deamon is running:

Run the command 'service eswitchd status'

.

if eswitchd is not running, start both the eswitchd and the Neutron agent using the commands 'service eswitchd start' :

#'service quantum-mlnx-agent start  // for Grizzly release

# service neutron-mlnx-agent start   // for Havana/Icehouse releases

Note: The Neutron agent and the eswitchd should run only on compute nodes. Neutron agent requires the eswitchd daemon to run.

Check that /etc/neutron/plugins/mlnx/mlnx_conf.ini file has the following entry:

daemon_endpoint = 'tcp://127.0.0.1:60001'

Note: The daemon endpoint is the address that the eswitch daemon is listening on.

60001 is the default port the eswitch daemon is listening on. If you change the value

socket_os_port in /etc/eswitchd/eswitchd.conf, you should also change the 'daemon_endpoint' variable in the /etc/neutron/plugins/mlnx/mlnx_conf.ini file.

There is a known bug(Havana)

Please confirm that /etc/neutron/plugins/mlnx/mlnx_conf.ini contains the entry:

rpc_support_old_agents = True

Make sure you are following the instructions for the DHCP server:

https://wiki.openstack.org/wiki/Mellanox-Neutron-Havana#DHCP_Server

Also, make sure you apply the patch for the Linux Bridge Agent on the Network node.

Make sure ipoibd is running.

If the eswitchd.log contains The following:

Stderr: 'sudo: sorry, you must have a tty to run sudo\n'

Make sure you have the latest version of eswitchd

In Red Hat Enterprise Linux OpenStack Platform -

Neutron server fails to start with error "neutron-db-check: ERROR: neutron database does not have version information. You must run the "neutron-db-manage stamp" command before starting neutron services"

In the file  /etc/init.d/neutron-server

comment the line "[ -x $dbcheck ] && $dbcheck || exit 6"

And run /etc/init.d/neutron-server start