test/nvmf_multipath: Retry checking if ANA state is changed with timeout

The latest Linux NVMe host (e.g. Linux kernel 5.12) have improved
the quality of multipath but still has an issue such that ANA state
is observed to be invalid even after connection is successfully
created.

By code inspection and system test, initializing ANA state may be
delayed even after connection is successfully created.

We can insert a short sleep but it will be better to retry checking
with timeout. The created function check_ana_state() is applied to
the remaining cases.

This fixes the part of the issue #1819.

Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ifc4eaef7c702d961caeee5bcdc9e6eab4704fc5d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8475
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: <dongx.yi@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This commit is contained in:
Shuhei Matsumoto 2021-06-22 19:00:39 +09:00 committed by Jim Harris
parent b503ef4fa0
commit 3aeeb4b68f

View File

@ -10,6 +10,21 @@ MALLOC_BLOCK_SIZE=512
rpc_py="$rootdir/scripts/rpc.py"
function check_ana_state() {
local subsys_id=$1
local ctrl_id=$2
local ana_state=$3
local timeout=3
while [ $(cat /sys/block/nvme"$subsys_id"c"$ctrl_id"n1/ana_state) != "$ana_state" ]; do
sleep 1
if ((timeout-- == 0)); then
echo "timeout before ANA state (nvme$subsys_id c$ctrl_id) becomes $ana_state"
return 1
fi
done
}
nvmftestinit
if [ -z $NVMF_SECOND_TARGET_IP ]; then
@ -44,8 +59,8 @@ subsys_id=$(nvme list-subsys | sed -n 's/nqn.2016-06.io.spdk:cnode1//p' | sed 's
ctrl1_id=$(nvme list-subsys | sed -n "s/traddr=$NVMF_FIRST_TARGET_IP trsvcid=$NVMF_PORT//p" | sed 's/[^0-9]*//g')
ctrl2_id=$(nvme list-subsys | sed -n "s/traddr=$NVMF_SECOND_TARGET_IP trsvcid=$NVMF_PORT//p" | sed 's/[^0-9]*//g')
[ $(cat /sys/block/nvme"$subsys_id"c"$ctrl1_id"n1/ana_state) == "optimized" ]
[ $(cat /sys/block/nvme"$subsys_id"c"$ctrl2_id"n1/ana_state) == "optimized" ]
check_ana_state "$subsys_id" "$ctrl1_id" "optimized"
check_ana_state "$subsys_id" "$ctrl2_id" "optimized"
# Set IO policy to numa
echo numa > /sys/class/nvme-subsystem/nvme-subsys$subsys_id/iopolicy
@ -58,28 +73,22 @@ sleep 1
$rpc_py nvmf_subsystem_listener_set_ana_state nqn.2016-06.io.spdk:cnode1 -t $TEST_TRANSPORT -a "$NVMF_FIRST_TARGET_IP" -s "$NVMF_PORT" -n inaccessible
$rpc_py nvmf_subsystem_listener_set_ana_state nqn.2016-06.io.spdk:cnode1 -t $TEST_TRANSPORT -a "$NVMF_SECOND_TARGET_IP" -s "$NVMF_PORT" -n non_optimized
sleep 1
[ $(cat /sys/block/nvme"$subsys_id"c"$ctrl1_id"n1/ana_state) == "inaccessible" ]
[ $(cat /sys/block/nvme"$subsys_id"c"$ctrl2_id"n1/ana_state) == "non-optimized" ]
check_ana_state "$subsys_id" "$ctrl1_id" "inaccessible"
check_ana_state "$subsys_id" "$ctrl2_id" "non-optimized"
$rpc_py nvmf_subsystem_listener_set_ana_state nqn.2016-06.io.spdk:cnode1 -t $TEST_TRANSPORT -a "$NVMF_FIRST_TARGET_IP" -s "$NVMF_PORT" -n non_optimized
$rpc_py nvmf_subsystem_listener_set_ana_state nqn.2016-06.io.spdk:cnode1 -t $TEST_TRANSPORT -a "$NVMF_SECOND_TARGET_IP" -s "$NVMF_PORT" -n inaccessible
sleep 1
[ $(cat /sys/block/nvme"$subsys_id"c"$ctrl1_id"n1/ana_state) == "non-optimized" ]
[ $(cat /sys/block/nvme"$subsys_id"c"$ctrl2_id"n1/ana_state) == "inaccessible" ]
check_ana_state "$subsys_id" "$ctrl1_id" "non-optimized"
check_ana_state "$subsys_id" "$ctrl2_id" "inaccessible"
wait $fio_pid
$rpc_py nvmf_subsystem_listener_set_ana_state nqn.2016-06.io.spdk:cnode1 -t $TEST_TRANSPORT -a "$NVMF_FIRST_TARGET_IP" -s "$NVMF_PORT" -n optimized
$rpc_py nvmf_subsystem_listener_set_ana_state nqn.2016-06.io.spdk:cnode1 -t $TEST_TRANSPORT -a "$NVMF_SECOND_TARGET_IP" -s "$NVMF_PORT" -n optimized
sleep 1
[ $(cat /sys/block/nvme"$subsys_id"c"$ctrl1_id"n1/ana_state) == "optimized" ]
[ $(cat /sys/block/nvme"$subsys_id"c"$ctrl2_id"n1/ana_state) == "optimized" ]
check_ana_state "$subsys_id" "$ctrl1_id" "optimized"
check_ana_state "$subsys_id" "$ctrl2_id" "optimized"
# Set IO policy to round-robin
echo round-robin > /sys/class/nvme-subsystem/nvme-subsys$subsys_id/iopolicy
@ -92,18 +101,14 @@ sleep 1
$rpc_py nvmf_subsystem_listener_set_ana_state nqn.2016-06.io.spdk:cnode1 -t $TEST_TRANSPORT -a "$NVMF_FIRST_TARGET_IP" -s "$NVMF_PORT" -n inaccessible
$rpc_py nvmf_subsystem_listener_set_ana_state nqn.2016-06.io.spdk:cnode1 -t $TEST_TRANSPORT -a "$NVMF_SECOND_TARGET_IP" -s "$NVMF_PORT" -n non_optimized
sleep 1
[ $(cat /sys/block/nvme"$subsys_id"c"$ctrl1_id"n1/ana_state) == "inaccessible" ]
[ $(cat /sys/block/nvme"$subsys_id"c"$ctrl2_id"n1/ana_state) == "non-optimized" ]
check_ana_state "$subsys_id" "$ctrl1_id" "inaccessible"
check_ana_state "$subsys_id" "$ctrl2_id" "non-optimized"
$rpc_py nvmf_subsystem_listener_set_ana_state nqn.2016-06.io.spdk:cnode1 -t $TEST_TRANSPORT -a "$NVMF_FIRST_TARGET_IP" -s "$NVMF_PORT" -n non_optimized
$rpc_py nvmf_subsystem_listener_set_ana_state nqn.2016-06.io.spdk:cnode1 -t $TEST_TRANSPORT -a "$NVMF_SECOND_TARGET_IP" -s "$NVMF_PORT" -n inaccessible
sleep 1
[ $(cat /sys/block/nvme"$subsys_id"c"$ctrl1_id"n1/ana_state) == "non-optimized" ]
[ $(cat /sys/block/nvme"$subsys_id"c"$ctrl2_id"n1/ana_state) == "inaccessible" ]
check_ana_state "$subsys_id" "$ctrl1_id" "non-optimized"
check_ana_state "$subsys_id" "$ctrl2_id" "inaccessible"
wait $fio_pid