diff --git a/test/nvmf/host/discovery.sh b/test/nvmf/host/discovery.sh
new file mode 100755
index 000000000..d504231fd
--- /dev/null
+++ b/test/nvmf/host/discovery.sh
@@ -0,0 +1,113 @@
+#!/usr/bin/env bash
+
+testdir=$(readlink -f $(dirname $0))
+rootdir=$(readlink -f $testdir/../../..)
+source $rootdir/test/common/autotest_common.sh
+source $rootdir/test/nvmf/common.sh
+
+rpc_py="$rootdir/scripts/rpc.py"
+
+if [ "$TEST_TRANSPORT" == "rdma" ]; then
+	echo "Skipping tests on RDMA because the rdma stack fails to configure the same IP for host and target."
+	exit 0
+fi
+
+DISCOVERY_PORT=8009
+DISCOVERY_NQN=nqn.2014-08.org.nvmexpress.discovery
+
+# NQN prefix to use for subsystem NQNs
+NQN=nqn.2016-06.io.spdk:cnode
+
+HOST_NQN=nqn.2021-12.io.spdk:test
+HOST_SOCK=/tmp/host.sock
+
+nvmftestinit
+
+# We will start the target as normal, emulating a storage cluster. We will simulate new paths
+# to the cluster via multiple listeners with different TCP ports.
+
+nvmfappstart -m 0x2
+
+$rpc_py nvmf_create_transport $NVMF_TRANSPORT_OPTS -u 8192
+$rpc_py nvmf_subsystem_add_listener $DISCOVERY_NQN -t $TEST_TRANSPORT -a $NVMF_FIRST_TARGET_IP \
+	-s $DISCOVERY_PORT
+$rpc_py bdev_null_create null0 1000 512
+$rpc_py bdev_null_create null1 1000 512
+$rpc_py bdev_wait_for_examine
+
+# Now start the host where the discovery service will run.  For our tests, we will send RPCs to
+# the "cluster" to create subsystems, add namespaces, add and remove listeners and add hosts,
+# and then check if the discovery service has detected the changes and constructed the correct
+# subsystem, ctrlr and bdev objects.
+
+$SPDK_BIN_DIR/nvmf_tgt -m 0x1 -r $HOST_SOCK &
+hostpid=$!
+waitforlisten $hostpid $HOST_SOCK
+
+trap 'process_shm --id $NVMF_APP_SHM_ID; kill $hostpid; nvmftestfini; exit 1' SIGINT SIGTERM EXIT
+
+$rpc_py -s $HOST_SOCK bdev_nvme_start_discovery -b nvme -t $TEST_TRANSPORT -a $NVMF_FIRST_TARGET_IP \
+	-s $DISCOVERY_PORT -f ipv4 -q $HOST_NQN
+
+function get_bdev_list() {
+	$rpc_py -s $HOST_SOCK bdev_get_bdevs | jq -r '.[].name' | sort | xargs
+}
+
+function get_subsystem_names() {
+	$rpc_py -s $HOST_SOCK bdev_nvme_get_controllers | jq -r '.[].name' | sort | xargs
+}
+
+function get_subsystem_paths() {
+	$rpc_py -s $HOST_SOCK bdev_nvme_get_controllers -n $1 | jq -r '.[].ctrlrs[].trid.trsvcid' | sort -n | xargs
+}
+
+[[ "$(get_subsystem_names)" == "" ]]
+[[ "$(get_bdev_list)" == "" ]]
+
+$rpc_py nvmf_create_subsystem ${NQN}0
+[[ "$(get_subsystem_names)" == "" ]]
+[[ "$(get_bdev_list)" == "" ]]
+
+$rpc_py nvmf_subsystem_add_ns ${NQN}0 null0
+[[ "$(get_subsystem_names)" == "" ]]
+[[ "$(get_bdev_list)" == "" ]]
+
+# Add listener for the subsystem.  But since the the subsystem was not created with -a option, the
+# discovery host will not be able to see the subsystem until its hostnqn is added.
+$rpc_py nvmf_subsystem_add_listener ${NQN}0 -t $TEST_TRANSPORT -a $NVMF_FIRST_TARGET_IP -s $NVMF_PORT
+[[ "$(get_subsystem_names)" == "" ]]
+[[ "$(get_bdev_list)" == "" ]]
+
+# Discovery hostnqn is added, so now the host should see the subsystem, with a single path for the
+# port of the single listener on the target.
+$rpc_py nvmf_subsystem_add_host ${NQN}0 $HOST_NQN
+[[ "$(get_subsystem_names)" == "nvme0" ]]
+[[ "$(get_bdev_list)" == "nvme0n1" ]]
+[[ "$(get_subsystem_paths nvme0)" == "$NVMF_PORT" ]]
+
+# Adding a namespace isn't a discovery function, but do it here anyways just to confirm we see a new bdev.
+$rpc_py nvmf_subsystem_add_ns ${NQN}0 null1
+[[ "$(get_bdev_list)" == "nvme0n1 nvme0n2" ]]
+
+# Add a second path to the same subsystem.  This shouldn't change the list of subsystems or bdevs, but
+# we should see a second path on the nvme0 subsystem now.
+$rpc_py nvmf_subsystem_add_listener ${NQN}0 -t $TEST_TRANSPORT -a $NVMF_FIRST_TARGET_IP -s $NVMF_SECOND_PORT
+[[ "$(get_subsystem_names)" == "nvme0" ]]
+[[ "$(get_bdev_list)" == "nvme0n1 nvme0n2" ]]
+[[ "$(get_subsystem_paths nvme0)" == "$NVMF_PORT $NVMF_SECOND_PORT" ]]
+
+# Remove the listener for the first port.  The subsystem and bdevs should stay, but we should see
+# the path to that first port disappear.
+$rpc_py nvmf_subsystem_remove_listener ${NQN}0 -t $TEST_TRANSPORT -a $NVMF_FIRST_TARGET_IP -s $NVMF_PORT
+[[ "$(get_subsystem_names)" == "nvme0" ]]
+[[ "$(get_bdev_list)" == "nvme0n1 nvme0n2" ]]
+[[ "$(get_subsystem_paths nvme0)" == "$NVMF_SECOND_PORT" ]]
+
+$rpc_py nvmf_delete_subsystem ${NQN}0
+[[ "$(get_subsystem_names)" == "" ]]
+[[ "$(get_bdev_list)" == "" ]]
+
+trap - SIGINT SIGTERM EXIT
+
+kill $hostpid
+nvmftestfini
diff --git a/test/nvmf/nvmf.sh b/test/nvmf/nvmf.sh
index bb8a16521..65b8d4e41 100755
--- a/test/nvmf/nvmf.sh
+++ b/test/nvmf/nvmf.sh
@@ -70,6 +70,7 @@ timing_enter host
 run_test "nvmf_identify" test/nvmf/host/identify.sh "${TEST_ARGS[@]}"
 run_test "nvmf_perf" test/nvmf/host/perf.sh "${TEST_ARGS[@]}"
 run_test "nvmf_multipath" test/nvmf/host/multipath.sh "${TEST_ARGS[@]}"
+run_test "nvmf_discovery" test/nvmf/host/discovery.sh "${TEST_ARGS[@]}"
 
 # TODO: disabled due to intermittent failures (RDMA_CM_EVENT_UNREACHABLE/ETIMEDOUT)
 #run_test test/nvmf/host/identify_kernel_nvmf.sh $TEST_ARGS