pybind/mgr/cephadm: fix issue with multiple nfs clusters on the same port

Currently even if ingress and virtual_ip are used, the port_in_use
check in cephadm fails when the same port is used. This PR fixes
the issue by ensuring that the daemon spec includes the virtual
ip address for the endpoint to be checked is included so it
doesn't default to checking `0.0.0.0:port` only.

Fixes: https://tracker.ceph.com/issues/69744
Signed-off-by: Omid Yoosefi <omidyoosefi@ibm.com>

diff --git a/src/pybind/mgr/cephadm/services/nfs.py b/src/pybind/mgr/cephadm/services/nfs.py
index 74ecfb486bdaab79a9aa9c21738bc3ac7ec5579d..aa2a12fd6c2a5603d7932a8a779e9b57f94a6c23 100644 (file)
--- a/src/pybind/mgr/cephadm/services/nfs.py
+++ b/src/pybind/mgr/cephadm/services/nfs.py
@@ -97,11 +97,14 @@ class NFSService(CephService):
         # create the rados config object
         self.create_rados_config_obj(spec)
 
+        port = daemon_spec.ports[0] if daemon_spec.ports else 2049
+
         # create the RGW keyring
         rgw_user = f'{rados_user}-rgw'
         rgw_keyring = self.create_rgw_keyring(daemon_spec)
         if spec.virtual_ip:
             bind_addr = spec.virtual_ip
+            daemon_spec.port_ips = {str(port): spec.virtual_ip}
         else:
             bind_addr = daemon_spec.ip if daemon_spec.ip else ''
         if not bind_addr:
@@ -119,7 +122,7 @@ class NFSService(CephService):
                 "rgw_user": rgw_user,
                 "url": f'rados://{POOL_NAME}/{spec.service_id}/{spec.rados_config_name()}',
                 # fall back to default NFS port if not present in daemon_spec
-                "port": daemon_spec.ports[0] if daemon_spec.ports else 2049,
+                "port": port,
                 "monitoring_port": spec.monitoring_port if spec.monitoring_port else 9587,
                 "bind_addr": bind_addr,
                 "haproxy_hosts": [],