Spdk/test/vhost/migration/migration-tc2.sh
Michal Berger 5e03beae70 test/vhost: Don't use nvmftestinit() in vhost tests
For now this function is not working very well in case it finds E810
cards (we do have some installed across CI pool) as we don't have
proper irdma driver in place.

Instead, and since it shouldn't be relevant for vhost tests, enable
soft-RoCE devices and use them (or any other rdma-capable NIC that
ip is going to be allocated on) during the vhost migration tests.

Signed-off-by: Michal Berger <michalx.berger@intel.com>
Change-Id: I4e8cc7b8eea3ce91684d8a68d0a0e11b8d1e4b9b
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8750
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
2021-07-20 07:59:31 +00:00

168 lines
5.1 KiB
Bash

source $rootdir/test/nvmf/common.sh
function migration_tc2_cleanup_nvmf_tgt() {
process_shm --id $NVMF_APP_SHM_ID
nvmftestfini
}
function migration_tc2_cleanup_vhost_config() {
timing_enter migration_tc2_cleanup_vhost_config
notice "Shutting down all VMs"
vm_shutdown_all
notice "Removing vhost devices & controllers via RPC ..."
# Delete bdev first to remove all LUNs and SCSI targets
$rpc_0 bdev_nvme_detach_controller Nvme0
$rpc_0 vhost_delete_controller $incoming_vm_ctrlr
$rpc_1 delete_nvme_controller Nvme0
$rpc_1 vhost_delete_controller $target_vm_ctrlr
notice "killing vhost app"
vhost_kill 0
vhost_kill 1
unset -v incoming_vm target_vm incoming_vm_ctrlr target_vm_ctrlr
unset -v rpc_0 rpc_1
migration_tc2_cleanup_nvmf_tgt
timing_exit migration_tc2_cleanup_vhost_config
}
function migration_tc2_configure_vhost() {
timing_enter migration_tc2_configure_vhost
# Just in case we don't have any rdma-capable NICs on board - using hw or
# soft-RoCE during vhost tests should not matter.
detect_soft_roce_nics
allocate_nic_ips
# Those are global intentionally - they will be unset in cleanup handler
incoming_vm=1
target_vm=2
incoming_vm_ctrlr=naa.VhostScsi0.$incoming_vm
target_vm_ctrlr=naa.VhostScsi0.$target_vm
rpc_0="$rootdir/scripts/rpc.py -s $(get_vhost_dir 0)/rpc.sock"
rpc_1="$rootdir/scripts/rpc.py -s $(get_vhost_dir 1)/rpc.sock"
# Run nvmf_tgt and two vhost instances:
# nvmf_tgt uses core id 2 (-m 0x4)
# First uses core id 0
# Second uses core id 1
# This force to use VM 1 and 2.
timing_enter start_nvmf_tgt
notice "Running nvmf_tgt..."
nvmfappstart -s 512 -m 0x4 --wait-for-rpc
# Override the trap set in place via nvmfappstart()
trap 'migration_tc2_error_cleanup; error_exit "${FUNCNAME}" "${LINENO}"' INT ERR EXIT
rpc_cmd framework_start_init
rpc_cmd nvmf_create_transport -t rdma -u 8192
mapfile -t json < <("$rootdir/scripts/gen_nvme.sh")
rpc_cmd load_subsystem_config -j "'${json[*]}'"
timing_exit start_nvmf_tgt
vhost_run -n 0 -a "-m 0x1 -s 512 -u"
vhost_run -n 1 -a "-m 0x2 -s 512 -u"
local rdma_ip_list
local nvmf_target_ip
rdma_ip_list=$(get_available_rdma_ips)
nvmf_target_ip=$(echo "$rdma_ip_list" | head -n 1)
if [[ -z "$nvmf_target_ip" ]]; then
fail "no NIC for nvmf target"
fi
notice "Configuring nvmf_tgt, vhost devices & controllers via RPC ..."
# Construct shared bdevs and controllers
rpc_cmd nvmf_create_subsystem nqn.2016-06.io.spdk:cnode1 -a -s SPDK00000000000001
rpc_cmd nvmf_subsystem_add_ns nqn.2016-06.io.spdk:cnode1 Nvme0n1
rpc_cmd nvmf_subsystem_add_listener nqn.2016-06.io.spdk:cnode1 -t rdma -a $nvmf_target_ip -s 4420
$rpc_0 bdev_nvme_attach_controller -b Nvme0 -t rdma -f ipv4 -a $nvmf_target_ip -s 4420 -n "nqn.2016-06.io.spdk:cnode1"
$rpc_0 vhost_create_scsi_controller $incoming_vm_ctrlr
$rpc_0 vhost_scsi_controller_add_target $incoming_vm_ctrlr 0 Nvme0n1
$rpc_1 bdev_nvme_attach_controller -b Nvme0 -t rdma -f ipv4 -a $nvmf_target_ip -s 4420 -n "nqn.2016-06.io.spdk:cnode1"
$rpc_1 vhost_create_scsi_controller $target_vm_ctrlr
$rpc_1 vhost_scsi_controller_add_target $target_vm_ctrlr 0 Nvme0n1
notice "Setting up VMs"
vm_setup --os="$os_image" --force=$incoming_vm --disk-type=spdk_vhost_scsi --disks=VhostScsi0 \
--migrate-to=$target_vm --memory=1024 --vhost-name=0
vm_setup --force=$target_vm --disk-type=spdk_vhost_scsi --disks=VhostScsi0 --incoming=$incoming_vm --memory=1024 \
--vhost-name=1
# Run everything
vm_run $incoming_vm $target_vm
# Wait only for incoming VM, as target is waiting for migration
vm_wait_for_boot 300 $incoming_vm
notice "Configuration done"
timing_exit migration_tc2_configure_vhost
}
function migration_tc2_error_cleanup() {
vm_kill_all
migration_tc2_cleanup_nvmf_tgt
migration_tc2_cleanup_vhost_config
notice "Migration TC2 FAILED"
}
function migration_tc2() {
# Use 2 VMs:
# incoming VM - the one we want to migrate
# targe VM - the one which will accept migration
local job_file="$testdir/migration-tc2.job"
local log_file
log_file="/root/$(basename ${job_file%%.*}).log"
migration_tc2_configure_vhost
# Run fio before migration
notice "Starting FIO"
vm_check_scsi_location $incoming_vm
run_fio $fio_bin --job-file="$job_file" --no-wait-for-fio --local --vm="${incoming_vm}$(printf ':/dev/%s' $SCSI_DISK)"
# Wait a while to let the FIO time to issue some IO
sleep 5
# Check if fio is still running before migration
if ! is_fio_running $incoming_vm; then
vm_exec $incoming_vm "cat $log_file"
error "FIO is not running before migration: process crashed or finished too early"
fi
vm_migrate $incoming_vm
sleep 3
# Check if fio is still running after migration
if ! is_fio_running $target_vm; then
vm_exec $target_vm "cat $log_file"
error "FIO is not running after migration: process crashed or finished too early"
fi
notice "Waiting for fio to finish"
local timeout=20
while is_fio_running $target_vm; do
sleep 1
echo -n "."
if ((timeout-- == 0)); then
error "timeout while waiting for FIO!"
fi
done
notice "Fio result is:"
vm_exec $target_vm "cat $log_file"
migration_tc2_cleanup_vhost_config
notice "Migration TC2 SUCCESS"
}