R25.10 hotfix client data nic (#874)

simplyblock_cli/cli-reference.yaml

@@ -923,6 +923,11 @@ commands:
                 min: 0
                 max: 128
             default: 3
+          - name: "--client-data-nic"
+            help: >
+              Network interface name from client to use for LVol connection.
+            dest: client_data_nic
+            type: str
       - name: add
         help: "Adds a new cluster"
         arguments:
@@ -1050,6 +1055,11 @@ commands:
             aliases:
               - "-n"
             type: str
+          - name: "--client-data-nic"
+            help: >
+              Network interface name from client to use for LVol connection.
+            dest: client_data_nic
+            type: str
       - name: activate
         help: >
           Activates a cluster.

simplyblock_cli/cli.py

@@ -386,6 +386,7 @@ def init_cluster__create(self, subparser):
         argument = subcommand.add_argument('--name', '-n', help='Assigns a name to the newly created cluster.', type=str, dest='name')
         argument = subcommand.add_argument('--qpair-count', help='NVMe/TCP transport qpair count per logical volume', type=range_type(0, 128), default=32, dest='qpair_count')
         argument = subcommand.add_argument('--client-qpair-count', help='default NVMe/TCP transport qpair count per logical volume for client', type=range_type(0, 128), default=3, dest='client_qpair_count')
+        argument = subcommand.add_argument('--client-data-nic', help='Network interface name from client to use for LVol connection.', type=str, dest='client_data_nic')
 
     def init_cluster__add(self, subparser):
         subcommand = self.add_sub_command(subparser, 'add', 'Adds a new cluster')
@@ -413,6 +414,7 @@ def init_cluster__add(self, subparser):
             argument = subcommand.add_argument('--inflight-io-threshold', help='The number of inflight IOs allowed before the IO queuing starts', type=int, default=4, dest='inflight_io_threshold')
         argument = subcommand.add_argument('--strict-node-anti-affinity', help='Enable strict node anti affinity for storage nodes. Never more than one chunk is placed on a node. This requires a minimum of _data-chunks-in-stripe + parity-chunks-in-stripe + 1_ nodes in the cluster."', dest='strict_node_anti_affinity', action='store_true')
         argument = subcommand.add_argument('--name', '-n', help='Assigns a name to the newly created cluster.', type=str, dest='name')
+        argument = subcommand.add_argument('--client-data-nic', help='Network interface name from client to use for LVol connection.', type=str, dest='client_data_nic')
 
     def init_cluster__activate(self, subparser):
         subcommand = self.add_sub_command(subparser, 'activate', 'Activates a cluster.')

simplyblock_cli/clibase.py

@@ -701,11 +701,13 @@ def cluster_add(self, args):
         inflight_io_threshold = args.inflight_io_threshold
         strict_node_anti_affinity = args.strict_node_anti_affinity
         is_single_node = args.is_single_node
+        client_data_nic = args.client_data_nic
 
         return cluster_ops.add_cluster(
             blk_size, page_size_in_blocks, cap_warn, cap_crit, prov_cap_warn, prov_cap_crit,
             distr_ndcs, distr_npcs, distr_bs, distr_chunk_bs, ha_type, enable_node_affinity,
-            qpair_count, max_queue_size, inflight_io_threshold, strict_node_anti_affinity, is_single_node, name, fabric)
+            qpair_count, max_queue_size, inflight_io_threshold, strict_node_anti_affinity, is_single_node, name, fabric,
+            client_data_nic)
 
     def cluster_create(self, args):
         page_size_in_blocks = args.page_size
@@ -740,14 +742,15 @@ def cluster_create(self, args):
         dns_name = args.dns_name
         is_single_node = args.is_single_node
         fabric = args.fabric
+        client_data_nic = args.client_data_nic
 
         return cluster_ops.create_cluster(
             blk_size, page_size_in_blocks,
             CLI_PASS, cap_warn, cap_crit, prov_cap_warn, prov_cap_crit,
             ifname, mgmt_ip, log_del_interval, metrics_retention_period, contact_point, grafana_endpoint,
             distr_ndcs, distr_npcs, distr_bs, distr_chunk_bs, ha_type, mode, enable_node_affinity,
             qpair_count, client_qpair_count, max_queue_size, inflight_io_threshold, disable_monitoring,
-            strict_node_anti_affinity, name, tls_secret, ingress_host_source, dns_name, fabric, is_single_node)
+            strict_node_anti_affinity, name, tls_secret, ingress_host_source, dns_name, fabric, is_single_node, client_data_nic)
 
     def query_yes_no(self, question, default="yes"):
         """Ask a yes/no question via raw_input() and return their answer.

simplyblock_core/cluster_ops.py

@@ -223,7 +223,7 @@ def create_cluster(blk_size, page_size_in_blocks, cli_pass,
                    cap_warn, cap_crit, prov_cap_warn, prov_cap_crit, ifname, mgmt_ip, log_del_interval, metrics_retention_period,
                    contact_point, grafana_endpoint, distr_ndcs, distr_npcs, distr_bs, distr_chunk_bs, ha_type, mode,
                    enable_node_affinity, qpair_count, client_qpair_count, max_queue_size, inflight_io_threshold, disable_monitoring, strict_node_anti_affinity, name,
-                   tls_secret, ingress_host_source, dns_name, fabric, is_single_node) -> str:
+                   tls_secret, ingress_host_source, dns_name, fabric, is_single_node, client_data_nic) -> str:
 
     if distr_ndcs == 0 and distr_npcs == 0:
         raise ValueError("both distr_ndcs and distr_npcs cannot be 0")
@@ -335,6 +335,7 @@ def create_cluster(blk_size, page_size_in_blocks, cli_pass,
     cluster.disable_monitoring = disable_monitoring
     cluster.mode = mode
     cluster.full_page_unmap = False
+    cluster.client_data_nic = client_data_nic or ""
 
     if mode == "docker":
         if not disable_monitoring:
@@ -441,7 +442,7 @@ def _run_fio(mount_point) -> None:
 def add_cluster(blk_size, page_size_in_blocks, cap_warn, cap_crit, prov_cap_warn, prov_cap_crit,
                 distr_ndcs, distr_npcs, distr_bs, distr_chunk_bs, ha_type, enable_node_affinity, qpair_count,
                 max_queue_size, inflight_io_threshold, strict_node_anti_affinity, is_single_node, name, cr_name=None,
-                cr_namespace=None, cr_plural=None, fabric="tcp", cluster_ip=None, grafana_secret=None) -> str:
+                cr_namespace=None, cr_plural=None, fabric="tcp", cluster_ip=None, grafana_secret=None, client_data_nic="") -> str:
 
 
     default_cluster = None
@@ -527,6 +528,7 @@ def add_cluster(blk_size, page_size_in_blocks, cap_warn, cap_crit, prov_cap_warn
     cluster.fabric_tcp = protocols["tcp"]
     cluster.fabric_rdma = protocols["rdma"]
     cluster.full_page_unmap = False
+    cluster.client_data_nic = client_data_nic or ""
 
     cluster.status = Cluster.STATUS_UNREADY
     cluster.create_dt = str(datetime.datetime.now())

simplyblock_core/controllers/lvol_controller.py

@@ -1302,6 +1302,9 @@ def connect_lvol(uuid, ctrl_loss_tmo=constants.LVOL_NVME_CONNECT_CTRL_LOSS_TMO):
             else:
                 keep_alive_to = constants.LVOL_NVME_KEEP_ALIVE_TO
 
+            client_data_nic_str = ""
+            if  cluster.client_data_nic:
+                client_data_nic_str = f"--host-iface={cluster.client_data_nic}"
             out.append({
                 "ns_id": lvol.ns_id,
                 "transport": transport,
@@ -1312,11 +1315,13 @@ def connect_lvol(uuid, ctrl_loss_tmo=constants.LVOL_NVME_CONNECT_CTRL_LOSS_TMO):
                 "ctrl-loss-tmo": ctrl_loss_tmo,
                 "nr-io-queues": cluster.client_qpair_count,
                 "keep-alive-tmo": keep_alive_to,
+                "host-iface": cluster.client_data_nic,
                 "connect": f"sudo nvme connect --reconnect-delay={constants.LVOL_NVME_CONNECT_RECONNECT_DELAY} "
                            f"--ctrl-loss-tmo={ctrl_loss_tmo} "
                            f"--nr-io-queues={cluster.client_qpair_count} "
                            f"--keep-alive-tmo={keep_alive_to} "
-                           f"--transport={transport} --traddr={ip} --trsvcid={port} --nqn={lvol.nqn}",
+                           f"--transport={transport} --traddr={ip} --trsvcid={port} --nqn={lvol.nqn} "
+                           f"{client_data_nic_str}",
             })
     return out
 

simplyblock_core/models/cluster.py

@@ -72,6 +72,7 @@ class Cluster(BaseModel):
     is_re_balancing: bool = False
     full_page_unmap: bool = True
     is_single_node: bool = False
+    client_data_nic: str = ""
 
     def get_status_code(self):
         if self.status in self.STATUS_CODE_MAP:

simplyblock_core/rpc_client.py

@@ -236,12 +236,14 @@ def transport_create(self, trtype, qpair_count=6, shared_bufs=24576):
             params.update({"c2h_success": True,"sock_priority": 0})
         return self._request("nvmf_create_transport", params)
 
-    def sock_impl_set_options(self):
-        method = "sock_impl_set_options"
-        params = {"impl_name": "posix", "enable_quickack": True,
-                  "enable_zerocopy_send_server": True,
-                  "enable_zerocopy_send_client": True}
-        return self._request(method, params)
+    def sock_impl_set_options(self, bind_to_device=None):
+        params = {
+            "impl_name": "posix", "enable_quickack": True,
+            "enable_zerocopy_send_server": True,
+            "enable_zerocopy_send_client": True}
+        if bind_to_device:
+            params["bind_to_device"] = bind_to_device
+        return self._request("sock_impl_set_options", params)
 
     def transport_create_caching(self, trtype):
         params = {

simplyblock_core/storage_node_ops.py

@@ -1274,7 +1274,10 @@ def add_node(cluster_id, node_addr, iface_name, data_nics_list,
             logger.warning(f"Failed to set nvmf max subsystems {constants.NVMF_MAX_SUBSYSTEMS}")
 
         # 2- set socket implementation options
-        ret = rpc_client.sock_impl_set_options()
+        bind_to_device = None
+        if snode.data_nics and len(snode.data_nics) == 1:
+            bind_to_device = snode.data_nics[0].if_name
+        ret = rpc_client.sock_impl_set_options(bind_to_device)
         if not ret:
             logger.error("Failed to set optimized socket options")
             return False
@@ -1838,7 +1841,10 @@ def restart_storage_node(
     rpc_client.accel_set_options()
 
     # 2- set socket implementation options
-    ret = rpc_client.sock_impl_set_options()
+    bind_to_device = None
+    if snode.data_nics and len(snode.data_nics) == 1:
+        bind_to_device = snode.data_nics[0].if_name
+    ret = rpc_client.sock_impl_set_options(bind_to_device)
     if not ret:
         logger.error("Failed socket implement set options")
         return False

simplyblock_web/api/v1/cluster.py

@@ -55,12 +55,14 @@ def add_cluster():
     inflight_io_threshold = cl_data.get('inflight_io_threshold', 4)
     strict_node_anti_affinity = cl_data.get('strict_node_anti_affinity', False)
     is_single_node = cl_data.get('is_single_node', False)
+    client_data_nic = cl_data.get('client_data_nic', "")
 
     return utils.get_response(cluster_ops.add_cluster(
         blk_size, page_size_in_blocks, cap_warn, cap_crit, prov_cap_warn, prov_cap_crit,
         distr_ndcs, distr_npcs, distr_bs, distr_chunk_bs, ha_type, enable_node_affinity,
-        qpair_count, max_queue_size, inflight_io_threshold, strict_node_anti_affinity, is_single_node, name, 
-        cr_name, cr_namespace, cr_plural, fabric
+        qpair_count, max_queue_size, inflight_io_threshold, strict_node_anti_affinity, is_single_node, name,
+        cr_name, cr_namespace, cr_plural, fabric,
+        client_data_nic
     ))
 
 

simplyblock_web/api/v2/cluster.py

@@ -48,6 +48,8 @@ class ClusterParams(BaseModel):
     cr_plural: str = ""
     cluster_ip: str = ""
     grafana_secret: str = ""
+    client_data_nic: str = ""
+
 
 @api.get('/', name='clusters:list')
 def list() -> List[ClusterDTO]: