womax/nova
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge "Use 2nd RPC server in compute operations"

Browse Source
This commit is contained in:
Zuul
2026-02-26 17:44:59 +00:00
committed by Gerrit Code Review
parent fbfc44f73b d5ffb58a8d
commit 44a7c5c2b0
14 changed files with 510 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.zuul.yaml
+41
View File
@@ -138,6 +138,44 @@
block_migrate_cinder_iscsi: true
post-run: playbooks/nova-live-migration/post-run.yaml
- job:
name: nova-graceful-shutdown
parent: devstack-multinode
description: |
Run Nova graceful shutdown tests.
run: playbooks/nova-graceful-shutdown/run.yaml
timeout: 10800
vars:
devstack_services:
neutron-trunk: true
openstack-cli-server: true
s-account: false
s-container: false
s-object: false
s-proxy: false
c-bak: false
tempest_test_regex: ''
devstack_localrc:
<<: *uec_image_vars
SERVICE_GRACEFUL_SHUTDOWN_TIMEOUT: 180
NOVA_ALLOW_MOVE_TO_SAME_HOST: false
LIVE_MIGRATION_AVAILABLE: true
USE_BLOCK_MIGRATION_FOR_LIVE_MIGRATION: true
group-vars:
subnode:
devstack_services:
openstack-cli-server: true
s-account: false
s-container: false
s-object: false
s-proxy: false
c-bak: false
devstack_localrc:
SERVICE_GRACEFUL_SHUTDOWN_TIMEOUT: 180
NOVA_ALLOW_MOVE_TO_SAME_HOST: false
LIVE_MIGRATION_AVAILABLE: true
USE_BLOCK_MIGRATION_FOR_LIVE_MIGRATION: true
- job:
name: nova-alt-configurations
parent: tempest-multinode-full-py3
@@ -829,6 +867,9 @@
- ^nova/network/.*$
- nova/virt/libvirt/vif.py
- nova-live-migration
# NOTE(gmaan): We will be running the graceful shutdown testing in
# check pipeline only and not required to test in gate as such.
- nova-graceful-shutdown
- nova-live-migration-ceph
- nova-lvm
- nova-multi-cell
nova/compute/manager.py
+1 -1
View File
@@ -642,7 +642,7 @@ class ComputeVirtAPI(virtapi.VirtAPI):
class ComputeManager(manager.Manager):
"""Manages the running instances from creation to destruction."""
target = messaging.Target(version='6.4')
target = messaging.Target(version='6.5')
def __init__(self, compute_driver=None, *args, **kwargs):
"""Load configuration options and connect to the hypervisor."""
nova/compute/rpcapi.py
+92 -13
View File
@@ -414,6 +414,7 @@ class ComputeAPI(object):
* 6.2 - Add target_state parameter to rebuild_instance()
* 6.3 - Add delete_attachment parameter to remove_volume_connection
* 6.4 - Add allow_share() and deny_share()
* 6.5 - Add 2nd RPC server with new topic 'compute-alt'
'''
VERSION_ALIASES = {
@@ -572,6 +573,33 @@ class ComputeAPI(object):
serializer=serializer,
call_monitor_timeout=cmt)
def prepare_for_alt_rpcserver(
self, client, server, version, **kwargs):
# NOTE(gmaan): By override the 'topic' in prepare() method, we make
# this rpc client to send the message to the different RPC server,
# which listen to RPC_TOPIC_ALT (the RPC server which is active during
# compute service graceful shutdown).
topic = RPC_TOPIC_ALT
msg = _("RPC: Sending the message to topic: %s") % topic
# NOTE(gmann): The old compute will not have the new 2nd RPC server
# so we need to handle it with RPC versioning. For the old compute,
# it will fallback to send the message to the original RPC server,
# which listen to RPC_TOPIC.
if not client.can_send_version('6.5'):
topic = RPC_TOPIC
msg = _("Fallback to send the message to original topic: %s as "
"RPC version is too old.") % topic
LOG.debug(msg)
params = {
'server': server,
'version': version,
'topic': topic}
params.update(kwargs)
return client.prepare(**params)
def add_fixed_ip_to_instance(self, ctxt, instance, network_id):
version = self._ver(ctxt, '5.0')
cctxt = self.router.client(ctxt).prepare(
@@ -612,6 +640,12 @@ class ComputeAPI(object):
kwargs.pop('migration')
kwargs.pop('limits')
version = '5.0'
# NOTE(gmaan): Most of the live migration RPC methods use the
# 'compute-alt' topic, but this RPC method should use the 'compute'
# topic. If a shutdown is initiated on the destination compute, the
# RPC server for the 'compute' topic will be stopped. If a live
# migration request arrives after that, the destination compute node
# should not take it.
cctxt = client.prepare(server=destination, version=version,
call_monitor_timeout=CONF.rpc_response_timeout,
timeout=CONF.long_rpc_timeout)
@@ -621,6 +655,10 @@ class ComputeAPI(object):
version = self._ver(ctxt, '5.0')
client = self.router.client(ctxt)
source = _compute_host(None, instance)
# NOTE(gmaan): Like check_can_live_migrate_destination, this RPC
# method should use topic 'compute'. If a shutdown is initiated
# on the source compute and, after that, a live migration request
# arrives, the source compute should not take it.
cctxt = client.prepare(server=source, version=version)
return cctxt.call(ctxt, 'check_can_live_migrate_source',
instance=instance,
@@ -867,8 +905,14 @@ class ComputeAPI(object):
def validate_console_port(self, ctxt, instance, port, console_type):
version = self._ver(ctxt, '5.0')
cctxt = self.router.client(ctxt).prepare(
server=_compute_host(None, instance), version=version)
client = self.router.client(ctxt)
# NOTE(gmaan): Send this RPC request to 'compute-alt' topic. This is
# called when the console is already requested. If shutdown is
# requested after that, compute should finish the port validation
# so that users can get their requested console.
cctxt = self.prepare_for_alt_rpcserver(
client,
server=_compute_host(None, instance), version=version)
return cctxt.call(ctxt, 'validate_console_port',
instance=instance, port=port,
console_type=console_type)
@@ -904,7 +948,13 @@ class ComputeAPI(object):
migration, migrate_data=None):
version = self._ver(ctxt, '5.0')
client = self.router.client(ctxt)
cctxt = client.prepare(server=host, version=version)
# NOTE(gmaan): Send this RPC request to 'compute-alt' topic. At this
# stage, both the source and destination compute have already confirmed
# that live migration can proceed. If the shutdown is initiated after
# that, the compute should finish the live migration using the
# 'compute-alt' RPC server.
cctxt = self.prepare_for_alt_rpcserver(
client, server=host, version=version)
cctxt.cast(ctxt, 'live_migration', instance=instance,
dest=dest, block_migration=block_migration,
migrate_data=migrate_data, migration=migration)
@@ -933,7 +983,12 @@ class ComputeAPI(object):
def post_live_migration_at_destination(self, ctxt, instance,
block_migration, host):
version = self._ver(ctxt, '5.0')
cctxt = self.router.client(ctxt).prepare(
client = self.router.client(ctxt)
# NOTE(gmaan): Send this RPC request to 'compute-alt' topic. If the
# shutdown is initiated during live migration, the compute should
# finish the live migration using the 'compute-alt' RPC server.
cctxt = self.prepare_for_alt_rpcserver(
client,
server=host, version=version,
call_monitor_timeout=CONF.rpc_response_timeout,
timeout=CONF.long_rpc_timeout)
@@ -951,9 +1006,14 @@ class ComputeAPI(object):
version = '5.0'
# We just need to honor the argument in the v5.0 RPC API method
msg_args['block_migration'] = None
cctxt = client.prepare(server=host, version=version,
timeout=CONF.long_rpc_timeout,
call_monitor_timeout=CONF.rpc_response_timeout)
# NOTE(gmaan): Send this RPC request to 'compute-alt' topic. If the
# shutdown is initiated during live migration, the compute should
# finish the live migration using the 'compute-alt' RPC server.
cctxt = self.prepare_for_alt_rpcserver(
client,
server=host, version=version,
timeout=CONF.long_rpc_timeout,
call_monitor_timeout=CONF.rpc_response_timeout)
return cctxt.call(ctxt, 'pre_live_migration',
instance=instance,
disk=disk, migrate_data=migrate_data,
@@ -1161,8 +1221,12 @@ class ComputeAPI(object):
if not client.can_send_version(version):
kwargs.pop('delete_attachment')
version = self._ver(ctxt, '5.0')
cctxt = client.prepare(server=host, version=version)
# NOTE(gmaan): This is called during live migration rollback. Send
# this RPC request to 'compute-alt' topic. If the shutdown is initiated
# during live migration rollback, the compute should finish the it
# using the 'compute-alt' RPC server.
cctxt = self.prepare_for_alt_rpcserver(
client, server=host, version=version)
return cctxt.call(ctxt, 'remove_volume_connection', **kwargs)
def rescue_instance(self, ctxt, instance, rescue_password,
@@ -1262,7 +1326,12 @@ class ComputeAPI(object):
migrate_data):
version = self._ver(ctxt, '5.0')
client = self.router.client(ctxt)
cctxt = client.prepare(server=host, version=version)
# NOTE(gmaan): This is called during live migration rollback. Send
# this RPC request to 'compute-alt' topic. If the shutdown is initiated
# during live migration rollback, the compute should finish it using
# the 'compute-alt' RPC server.
cctxt = self.prepare_for_alt_rpcserver(
client, server=host, version=version)
cctxt.cast(ctxt, 'rollback_live_migration_at_destination',
instance=instance, destroy_disks=destroy_disks,
migrate_data=migrate_data)
@@ -1286,7 +1355,12 @@ class ComputeAPI(object):
"""
version = self._ver(ctxt, '5.3')
client = self.router.client(ctxt)
cctxt = client.prepare(server=host, version=version)
# NOTE(gmaan): This is called during live migration rollback. Send
# this RPC request to 'compute-alt' topic. If the shutdown is initiated
# during live migration rollback, the compute should finish it using
# the 'compute-alt' RPC server.
cctxt = self.prepare_for_alt_rpcserver(
client, server=host, version=version)
cctxt.call(ctxt, 'drop_move_claim_at_destination', instance=instance)
def set_admin_password(self, ctxt, instance, new_pass):
@@ -1523,8 +1597,13 @@ class ComputeAPI(object):
def external_instance_event(self, ctxt, instances, events, host=None):
instance = instances[0]
version = self._ver(ctxt, '5.0')
cctxt = self.router.client(ctxt).prepare(
server=_compute_host(host, instance),
client = self.router.client(ctxt)
# NOTE(gmaan): This is initiated by the external services (for
# example, neutron send event for network change) and let's not block
# them during shutdown. Make this RPC request to 'compute-alt' topic.
cctxt = self.prepare_for_alt_rpcserver(
client,
_compute_host(host, instance),
version=version)
cctxt.cast(ctxt, 'external_instance_event', instances=instances,
events=events)
nova/objects/service.py
+4 -1
View File
@@ -37,7 +37,7 @@ __all__ = [
# NOTE(danms): This is the global service version counter
SERVICE_VERSION = 70
SERVICE_VERSION = 71
# NOTE(danms): This is our SERVICE_VERSION history. The idea is that any
@@ -249,6 +249,9 @@ SERVICE_VERSION_HISTORY = (
# Version 70: Compute RPC v6.4:
# Compute manager supports USB controller model traits
{'compute_rpc': '6.4'},
# Version 71: Compute RPC v6.5:
# Add 2nd RPC server for compute service
{'compute_rpc': '6.5'},
)
# This is the version after which we can rely on having a persistent
nova/test.py
+6
View File
@@ -461,6 +461,12 @@ class TestCase(base.BaseTestCase):
if host is not None:
# Make sure that CONF.host is relevant to the right hostname
self.useFixture(nova_fixtures.ConfPatcher(host=host))
# By default, service creates a RPC server for auto populated
# 'topic' from service binary name. For compute service, we need
# to create the 2nd RPC server which will be done by pass the
# 'topic_alt' explicitly.
if name == 'compute' and 'topic_alt' not in kwargs:
kwargs['topic_alt'] = compute_rpcapi.RPC_TOPIC_ALT
if name == 'compute' and self.USES_DB:
# NOTE(danms): We need to create the HostMapping first, because
nova/tests/unit/compute/test_rpcapi.py
+67 -7
View File
@@ -131,10 +131,13 @@ class ComputeRpcAPITestCase(test.NoDBTestCase):
prepare_extra_kwargs = {}
cm_timeout = kwargs.pop('call_monitor_timeout', None)
timeout = kwargs.pop('timeout', None)
topic_alt = kwargs.pop('topic_alt', None)
if cm_timeout:
prepare_extra_kwargs['call_monitor_timeout'] = cm_timeout
if timeout:
prepare_extra_kwargs['timeout'] = timeout
if topic_alt:
prepare_extra_kwargs['topic'] = topic_alt
# NOTE(sbauza): If expected args are provided, we need to use them
# for the expected kwargs and just add the needed _return_value that
@@ -368,7 +371,8 @@ class ComputeRpcAPITestCase(test.NoDBTestCase):
def test_validate_console_port(self):
self._test_compute_api('validate_console_port', 'call',
instance=self.fake_instance_obj, port="5900",
console_type="novnc", version='6.0')
console_type="novnc", version='6.0',
topic_alt=compute_rpcapi.RPC_TOPIC_ALT)
def test_host_maintenance_mode(self):
self._test_compute_api('host_maintenance_mode', 'call',
@@ -387,7 +391,8 @@ class ComputeRpcAPITestCase(test.NoDBTestCase):
instance=self.fake_instance_obj, dest='dest',
block_migration='blockity_block', host='tsoh',
migration='migration',
migrate_data={}, version='6.0')
migrate_data={}, version='6.0',
topic_alt=compute_rpcapi.RPC_TOPIC_ALT)
def test_live_migration_force_complete(self):
migration = migration_obj.Migration()
@@ -420,7 +425,8 @@ class ComputeRpcAPITestCase(test.NoDBTestCase):
self._test_compute_api('post_live_migration_at_destination', 'call',
instance=self.fake_instance_obj,
block_migration='block_migration', host='host', version='6.0',
timeout=1234, call_monitor_timeout=60)
timeout=1234, call_monitor_timeout=60,
topic_alt=compute_rpcapi.RPC_TOPIC_ALT)
def test_pause_instance(self):
self._test_compute_api('pause_instance', 'cast',
@@ -448,7 +454,8 @@ class ComputeRpcAPITestCase(test.NoDBTestCase):
expected_args, instance=self.fake_instance_obj,
block_migration='block_migration', disk='disk', host='host',
migrate_data=None, version='6.0',
call_monitor_timeout=60, timeout=1234)
call_monitor_timeout=60, timeout=1234,
topic_alt=compute_rpcapi.RPC_TOPIC_ALT)
def test_supports_numa_live_migration(self):
mock_client = mock.MagicMock()
@@ -506,10 +513,19 @@ class ComputeRpcAPITestCase(test.NoDBTestCase):
instance=self.fake_instance_obj, block_migration=False,
disk_over_commit=False)
def test_rollback_live_migration_at_destination(self):
self._test_compute_api('rollback_live_migration_at_destination',
'cast', instance=self.fake_instance_obj,
host='host', destroy_disks=True,
migrate_data=None, version='6.0',
_return_value=None,
topic_alt=compute_rpcapi.RPC_TOPIC_ALT)
def test_drop_move_claim_at_destination(self):
self._test_compute_api('drop_move_claim_at_destination', 'call',
instance=self.fake_instance_obj, host='host',
version='6.0', _return_value=None)
version='6.0', _return_value=None,
topic_alt=compute_rpcapi.RPC_TOPIC_ALT)
def test_prep_resize(self):
self._test_compute_api('prep_resize', 'cast',
@@ -965,7 +981,8 @@ class ComputeRpcAPITestCase(test.NoDBTestCase):
def test_remove_volume_connection(self):
self._test_compute_api('remove_volume_connection', 'call',
instance=self.fake_instance_obj, volume_id='id', host='host',
delete_attachment=True, version='6.3')
delete_attachment=True, version='6.3',
topic_alt=compute_rpcapi.RPC_TOPIC_ALT)
def test_rescue_instance(self):
self._test_compute_api('rescue_instance', 'cast',
@@ -1218,7 +1235,8 @@ class ComputeRpcAPITestCase(test.NoDBTestCase):
self._test_compute_api('external_instance_event', 'cast',
instances=[self.fake_instance_obj],
events=['event'],
version='6.0')
version='6.0',
topic_alt=compute_rpcapi.RPC_TOPIC_ALT)
def test_build_and_run_instance(self):
# With rpcapi 5.11, when a list of accel_uuids is passed as a param,
@@ -1354,3 +1372,45 @@ class ComputeRpcAPITestCase(test.NoDBTestCase):
instance=self.fake_instance_obj,
share_mapping=self.get_fake_share_mapping(),
version='6.4')
def test_prepare_for_alt_rpcserver_select_topic_alt(self):
rpcapi = compute_rpcapi.ComputeAPI()
mock_client = mock.MagicMock()
mock_client.can_send_version.return_value = True
rpcapi.prepare_for_alt_rpcserver(
mock_client, server='fake_host', version='6.5')
mock_client.can_send_version.assert_called_once_with('6.5')
mock_client.prepare.assert_called_once_with(
server='fake_host',
version='6.5',
topic=compute_rpcapi.RPC_TOPIC_ALT)
def test_prepare_for_alt_rpcserver_fallback_topic_for_old_compute(self):
rpcapi = compute_rpcapi.ComputeAPI()
mock_client = mock.MagicMock()
mock_client.can_send_version.return_value = False
rpcapi.prepare_for_alt_rpcserver(
mock_client, server='fake_host', version='6.0')
mock_client.can_send_version.assert_called_once_with('6.5')
mock_client.prepare.assert_called_once_with(
server='fake_host',
version='6.0',
topic=compute_rpcapi.RPC_TOPIC)
def test_prepare_for_alt_rpcserver_with_extra_kwargs(self):
rpcapi = compute_rpcapi.ComputeAPI()
mock_client = mock.MagicMock()
mock_client.can_send_version.return_value = True
rpcapi.prepare_for_alt_rpcserver(
mock_client,
server='fake_host',
version='6.5',
call_monitor_timeout=60,
timeout=120)
mock_client.can_send_version.assert_called_once_with('6.5')
mock_client.prepare.assert_called_once_with(
server='fake_host',
version='6.5',
topic=compute_rpcapi.RPC_TOPIC_ALT,
call_monitor_timeout=60,
timeout=120)
playbooks/nova-graceful-shutdown/run.yaml
+7
View File
@@ -0,0 +1,7 @@
---
- hosts: all
roles:
- orchestrate-devstack
- hosts: controller
roles:
- run-graceful-shutdown-tests
releasenotes/notes/nova-services-graceful-shutdown-564a321e2769152d.yaml
+55
View File
@@ -0,0 +1,55 @@
---
features:
- |
Nova services now support graceful shutdown on ``SIGTERM``. When a service
receives ``SIGTERM``, it will stop accepting new RPC requests and wait for
in-progress tasks to reach a safe termination point.
The compute service creates a second RPC server on an ``compute-alt`` topic
which remains active during graceful shutdown, allowing compute service to
finish the in-progress tasks.
Currently below operations are using second RPC server:
* Live migration
* Server external Event
* Get Console output
Nova added two new configuration options which will control this behavior:
* ``[DEFAULT]/graceful_shutdown_timeout`` - The overall time the service
waits before forcefully exit. This is defaults to 180 seconds for each
Nova services.
* ``[DEFAULT]/manager_shutdown_timeout`` - The time the service manager
waits for in-progress tasks to complete during graceful shutdown. This
is defaults to 160 seconds for each service manager. This must be less
than ``graceful_shutdown_timeout``.
You can increase these timeouts based on the traffic and how long the
long-running (e.g. live migrations) tasks take in your deployment.
We plan to improve the graceful shutdown in future releases by task
tracking and transitioning resources to a recoverable state. Until then,
this feature is experimental.
upgrade:
- |
The default value of ``[DEFAULT]/graceful_shutdown_timeout`` has been
changed from 60 to 180 seconds for all Nova services. This means that
when a Nova service receives ``SIGTERM``, it will now wait up to 180
seconds for a graceful shutdown before being forcefully terminated.
Operators using external system (e.g. k8s, systemd) to manage the
Nova serviecs should ensure that their service stop timeouts are set
to at least ``graceful_shutdown_timeout`` to avoid forcefully killing
service before Nova finish its graceful shutdown. For example, the
systemd ``TimeoutStopSec`` should be set to at least 180 seconds (or
greater) for Nova services.
- |
A new configuration option ``[DEFAULT]/manager_shutdown_timeout`` has been
added with a default value of 160 seconds. This controls how long the
service manager waits for in-progress tasks to finish during graceful
shutdown. Operators may want to tune this value based on how long their
typical long-running operations (e.g. live migrations) take to complete.
- |
The compute service now creates a second RPC server on the ``compute-alt``
topic. This means each compute worker will create an additional RabbitMQ
queue.
roles/run-graceful-shutdown-tests/README.rst
+1
View File
@@ -0,0 +1 @@
Run Nova graceful shutdown tests and verify the operations.
roles/run-graceful-shutdown-tests/files/cleanup_test_servers.sh
Executable
+47
View File
@@ -0,0 +1,47 @@
#!/bin/bash
source /opt/stack/devstack/openrc admin
set -x
set -e
confirm_resize() {
local server=$1
echo "Confirming resize on ${server}"
openstack server resize confirm "${server}"
count=0
while true; do
status=$(openstack server show "${server}" -f value -c status 2>/dev/null || echo "NOT_FOUND")
if [ "${status}" == "ACTIVE" ] || [ "${status}" == "ERROR" ]; then
break
fi
sleep 5
count=$((count+1))
if [ ${count} -eq 10 ]; then
echo "Timed out waiting for ${server} to be ACTIVE or Error after confirm resize"
break
fi
done
}
cleanup_server() {
local server=$1
status=$(openstack server show "${server}" -f value -c status 2>/dev/null || echo "NOT_FOUND")
if [ "${status}" == "VERIFY_RESIZE" ]; then
confirm_resize "${server}"
fi
status=$(openstack server show "${server}" -f value -c status 2>/dev/null || echo "NOT_FOUND")
if [ "${status}" == "ACTIVE" ] || [ "${status}" == "ERROR" ]; then
echo "Deleting ${server} (status: ${status})"
openstack server delete --wait "${server}"
else
echo "Skipping ${server} deletion (status: ${status})"
fi
}
for server in "$@"; do
cleanup_server "${server}"
done
roles/run-graceful-shutdown-tests/files/start_and_verify_compute_service.sh
Executable
+39
View File
@@ -0,0 +1,39 @@
#!/bin/bash
set -x
set -e
COMPUTE_HOST=$1
EXPECTED_STATE=${2:-active}
get_service_status() {
local host=$1
local status
status=$(ssh "${host}" systemctl is-active devstack@n-cpu || true)
echo "${status}"
}
wait_for_service_state() {
local host=$1
local expected=$2
local timeout=${3:-30}
local count=0
local status
status=$(get_service_status "${host}")
while [ "${status}" != "${expected}" ]; do
sleep 5
count=$((count+1))
if [ ${count} -eq ${timeout} ]; then
echo "Timed out waiting for compute service on ${host} to be ${expected} (current: ${status})"
exit 5
fi
status=$(get_service_status "${host}")
done
echo "Compute service on ${host} is ${expected}"
}
if [ "${EXPECTED_STATE}" == "active" ] && [ "$(get_service_status "${COMPUTE_HOST}")" != "active" ]; then
ssh "${COMPUTE_HOST}" sudo systemctl start devstack@n-cpu
fi
wait_for_service_state "${COMPUTE_HOST}" "${EXPECTED_STATE}"
roles/run-graceful-shutdown-tests/files/start_live_migration.sh
Executable
+49
View File
@@ -0,0 +1,49 @@
#!/bin/bash
source /opt/stack/devstack/openrc admin
set -x
set -e
timeout=196
server_lm=$1
image_id=$(openstack image list -f value -c ID | awk 'NR==1{print $1}')
flavor_id=$(openstack flavor list -f value -c ID | awk 'NR==1{print $1}')
network_id=$(openstack network list --no-share -f value -c ID | awk 'NR==1{print $1}')
echo "Creating test server on subnode for graceful shutdown live migration test"
openstack --os-compute-api-version 2.74 server create --image ${image_id} --flavor ${flavor_id} \
--nic net-id=${network_id} --host ${SUBNODE_HOSTNAME} --wait ${server_lm}
echo "Starting live migration of ${server_lm} to ${CONTROLLER_HOSTNAME}"
openstack server migrate --live-migration \
--host ${CONTROLLER_HOSTNAME} ${server_lm}
# Wait for the migration to be in progress before returning so that the
# SIGTERM can be sent while the migrations are in progress.
count=0
while true; do
migration_status=$(openstack server migration list ${server_lm} \
-f value -c Status 2>/dev/null | head -1)
server_status=$(openstack server show ${server_lm} \
-f value -c status 2>/dev/null)
task_state=$(openstack server show ${server_lm} \
-f value -c OS-EXT-STS:task_state 2>/dev/null)
if [ "${migration_status}" == "preparing" ] || \
[ "${migration_status}" == "running" ] || \
[ "${task_state}" == "migrating" ]; then
echo "Live migration is in progress (status: ${migration_status}, task_state: ${task_state})"
break
elif [ "${migration_status}" == "completed" ] || \
{ [ "${server_status}" == "ACTIVE" ] && \
{ [ "${task_state}" == "None" ] || [ -z "${task_state}" ]; }; }; then
echo "Live migration has already completed"
exit 2
fi
count=$((count+1))
if [ ${count} -eq ${timeout} ]; then
echo "Timed out waiting for migrations to start"
exit 2
fi
done
roles/run-graceful-shutdown-tests/files/verify_live_migration.sh
Executable
+45
View File
@@ -0,0 +1,45 @@
#!/bin/bash
source /opt/stack/devstack/openrc admin
set -x
set -e
server=$1
# Wait for the server to finish live migration and become ACTIVE with
# no task_state, which indicates the migration has completed.
timeout=360
count=0
migration_start=$(date +%s)
while true; do
status=$(openstack server show ${server} -f value -c status)
task_state=$(openstack server show ${server} -f value -c OS-EXT-STS:task_state)
if [ "${status}" == "ACTIVE" ] && { [ "${task_state}" == "None" ] || [ -z "${task_state}" ]; }; then
migration_end=$(date +%s)
migration_duration=$((migration_end - migration_start))
echo "Migration is completed in ${migration_duration} seconds."
break
fi
if [ "${status}" == "ERROR" ]; then
echo "Server went to ERROR status during live migration"
exit 3
fi
sleep 5
count=$((count+1))
if [ ${count} -eq ${timeout} ]; then
echo "Timed out waiting for live migration to complete"
exit 5
fi
done
# Make sure the server moved to the controller.
host=$(openstack server show ${server} -f value -c OS-EXT-SRV-ATTR:host)
if [[ ${host} != ${CONTROLLER_HOSTNAME} ]]; then
echo "Unexpected host ${host} for server after live migration during graceful shutdown."
exit 4
fi
echo "Live migration during graceful shutdown completed successfully"
echo "Server ${server} is ACTIVE on ${host}"
roles/run-graceful-shutdown-tests/tasks/main.yaml
+56
View File
@@ -0,0 +1,56 @@
- name: Graceful shutdown source compute live migration
block:
- name: Start live migrations of test servers
become: true
become_user: stack
script: "start_live_migration.sh server-lm1"
environment:
SUBNODE_HOSTNAME: "{{ hostvars['compute1']['ansible_hostname'] }}"
CONTROLLER_HOSTNAME: "{{ hostvars['controller']['ansible_hostname'] }}"
register: start_live_migrations_result
failed_when: start_live_migrations_result.rc not in [0, 2]
- name: Set fact if migrations completed or timed out before SIGTERM to source compute
set_fact:
live_migrations_completed_or_timeout: "{{ start_live_migrations_result.rc == 2 }}"
- name: Run graceful shutdown tests
when: not live_migrations_completed_or_timeout
block:
- name: Send SIGTERM to source compute to start the source compute graceful shutdown
delegate_to: compute1
become: true
shell: "kill -15 $(systemctl show devstack@n-cpu -p MainPID --value)"
- name: Verify live migration is completed during graceful shutdown
become: true
become_user: stack
script: "verify_live_migration.sh server-lm1"
environment:
CONTROLLER_HOSTNAME: "{{ hostvars['controller']['ansible_hostname'] }}"
# Sleep for 180 sec: default graceful_shutdown_timeout
- name: Sleep for 180 seconds to allow source compute graceful shutdown to complete
pause:
seconds: 180
- name: Verify compute service is stopped after graceful shutdown
become: true
become_user: stack
script: "start_and_verify_compute_service.sh {{ hostvars['compute1']['ansible_hostname'] }} inactive"
- name: Start and verify subnode compute service is running
become: true
become_user: stack
script: "start_and_verify_compute_service.sh {{ hostvars['compute1']['ansible_hostname'] }}"
- name: Cleanup test servers
become: true
become_user: stack
script: "cleanup_test_servers.sh server-lm1"
ignore_errors: true
- name: Fail if any test is skipped
fail:
msg: "One or more test is skipped due to operation is either completed or timed out before SIGTERM signal."
when: live_migrations_completed_or_timeout