test(bdd): make nvme controller usage more robust

Caters for when the device is /dev/nvmeX but X not the same as the controller!

Signed-off-by: Tiago Castro <[email protected]>

scripts/pytest-tests.sh

@@ -36,6 +36,15 @@ function clean_all()
   echo "Done"
 }
 
+function is_test()
+{
+  file="${1:-}"
+  if [ -d "$file" ] || [ -f "$file" ] || [ -f "${file%::*}" ]; then
+    return 0
+  fi
+  return 1
+}
+
 function run_tests()
 {
   while read name extra
@@ -95,10 +104,13 @@ while [ "$#" -gt 0 ]; do
     *)
       set +e
       real_1="$(realpath $1 2>/dev/null)"
+      real_2="$(realpath $SRCDIR/test/python/$1 2>/dev/null)"
       set -e
       param="$1"
-      if [ -d "$real_1" ] || [ -f "$real_1" ] || [ -f "${real_1%::*}" ]; then
+      if is_test "$real_1"; then
         param="$real_1"
+      elif is_test "$real_2"; then
+        param="$real_2"
       else
         TEST_ARGS="${TEST_ARGS:-}$1"
       fi

test/python/common/nvme.py

@@ -68,6 +68,7 @@ def nvme_connect(uri, delay=10, tmo=600):
     command = (
         f"nix-sudo nvme connect -t tcp -s {port} -a {host} -n {nqn} -c {delay} -l {tmo}"
     )
+    print(command)
     subprocess.run(command, check=True, shell=True, capture_output=False)
     time.sleep(1)
     command = "nix-sudo nvme list -v -o json"
@@ -97,6 +98,43 @@ def nvme_id_ctrl(device):
     return id_ctrl
 
 
+def match_host_port(addr, host, port):
+    traddr = f"traddr={host}"
+    trsvcid = f"trsvcid={port}"
+
+    return traddr in addr and trsvcid in addr
+
+
+def nvme_find_ctrl(uri):
+    """Find controller from the device uri."""
+    u = urlparse(uri)
+    port = u.port
+    host = u.hostname
+    nqn = u.path[1:]
+
+    command = "nix-sudo nvme list -v -o json"
+    discover = json.loads(
+        subprocess.run(
+            command, shell=True, check=True, text=True, capture_output=True
+        ).stdout
+    )
+
+    # Finds correct Device
+    devs = list(filter(lambda d: nqn in d.get("SubsystemNQN"), discover.get("Devices")))
+    assert len(devs) is 1, "Multiple devices with the same subnqn"
+
+    # Find correct Controller
+    ctrls = list(
+        filter(
+            lambda d: match_host_port(d.get("Address"), host, port),
+            devs[0].get("Controllers"),
+        )
+    )
+    assert len(ctrls) is 1, "Multiple controllers with the same address"
+
+    return ctrls[0].get("Controller")
+
+
 def nvme_resv_report(device):
     """Reservation report."""
     command = "nix-sudo nvme resv-report {0} -c 1 -o json".format(device)
@@ -129,18 +167,21 @@ def nvme_disconnect(uri):
     nqn = u.path[1:]
 
     command = "nix-sudo nvme disconnect -n {0}".format(nqn)
+    print(command)
     subprocess.run(command, check=True, shell=True, capture_output=True)
 
 
 def nvme_disconnect_controller(name):
     """Disconnect the given NVMe controller on this host."""
     command = "nix-sudo nvme disconnect -d {0}".format(name)
+    print(command)
     subprocess.run(command, check=True, shell=True, capture_output=True)
 
 
 def nvme_disconnect_all():
     """Disconnect from all connected nvme subsystems"""
     command = "nix-sudo nvme disconnect-all"
+    print(command)
     subprocess.run(command, check=True, shell=True, capture_output=True)
 
 

test/python/tests/nexus_fault/test_nexus_fault.py

@@ -16,7 +16,12 @@
 from common.command import run_cmd
 from common.fio import Fio
 from common.mayastor import container_mod, mayastor_mod
-from common.nvme import nvme_connect, nvme_disconnect, nvme_disconnect_controller
+from common.nvme import (
+    nvme_connect,
+    nvme_disconnect,
+    nvme_disconnect_controller,
+    nvme_find_ctrl,
+)
 
 import nexus_pb2 as pb
 
@@ -119,13 +124,12 @@ def _(recreate_pool, republish_nexus_ana):
 
 
 @when("the initiator swaps the nexuses")
-def _(recreate_pool, republish_nexus_ana):
+def _(publish_nexus, recreate_pool, republish_nexus_ana):
     """the initiator swaps the nexuses."""
     print(republish_nexus_ana)
-    device = nvme_connect(republish_nexus_ana)
-    # disconnect previous nexus: /dev/nvme*n*
-    split_dev = device.split("n")
-    nvme_disconnect_controller(f"{split_dev[0]}n{split_dev[1]}")
+    prev_ctrl = nvme_find_ctrl(publish_nexus)
+    nvme_connect(republish_nexus_ana)
+    nvme_disconnect_controller(f"/dev/{prev_ctrl}")
 
 
 @then("the fio workload should complete gracefully")