openstack/watcher
Code Issues Proposed changes

Merge "Handle deprecated "cpu_util" metric"

Browse Source
This commit is contained in:
Zuul 2023-10-27 09:47:38 +00:00 committed by Gerrit Code Review
commit 40e93407c7
9 changed files with 74 additions and 31 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
doc/source/configuration/configuring.rst
View File

@ -372,7 +372,7 @@ You can configure and install Ceilometer by following the documentation below :
#. https://docs.openstack.org/ceilometer/latest
The built-in strategy 'basic_consolidation' provided by watcher requires
"**compute.node.cpu.percent**" and "**cpu_util**" measurements to be collected
"**compute.node.cpu.percent**" and "**cpu**" measurements to be collected
by Ceilometer.
The measurements available depend on the hypervisors that OpenStack manages on
the specific implementation.

2
doc/source/contributor/plugin/strategy-plugin.rst
View File

@ -300,6 +300,6 @@ Using that you can now query the values for that specific metric:
.. code-block:: py
avg_meter = self.datasource_backend.statistic_aggregation(
instance.uuid, 'cpu_util', self.periods['instance'],
instance.uuid, 'instance_cpu_usage', self.periods['instance'],
self.granularity,
aggregation=self.aggregation_method['instance'])

3
doc/source/strategies/basic-server-consolidation.rst
View File

@ -26,8 +26,7 @@ metric service name plugins comment
``compute_monitors`` option
to ``cpu.virt_driver`` in
the nova.conf.
``cpu_util`` ceilometer_ none cpu_util has been removed
since Stein.
``cpu`` ceilometer_ none
============================ ============ ======= ===========================
.. _ceilometer: https://docs.openstack.org/ceilometer/latest/admin/telemetry-measurements.html#openstack-compute

3
doc/source/strategies/vm_workload_consolidation.rst
View File

@ -22,8 +22,7 @@ The *vm_workload_consolidation* strategy requires the following metrics:
============================ ============ ======= =========================
metric service name plugins comment
============================ ============ ======= =========================
``cpu_util`` ceilometer_ none cpu_util has been removed
since Stein.
``cpu`` ceilometer_ none
``memory.resident`` ceilometer_ none
``memory`` ceilometer_ none
``disk.root.size`` ceilometer_ none

17
doc/source/strategies/workload-stabilization.rst
View File

@ -27,9 +27,8 @@ metric service name plugins comment
to ``cpu.virt_driver`` in the
nova.conf.
``hardware.memory.used`` ceilometer_ SNMP_
``cpu_util`` ceilometer_ none cpu_util has been removed
since Stein.
``memory.resident`` ceilometer_ none
``cpu`` ceilometer_ none
``instance_ram_usage`` ceilometer_ none
============================ ============ ======= =============================
.. _ceilometer: https://docs.openstack.org/ceilometer/latest/admin/telemetry-measurements.html#openstack-compute
@ -107,10 +106,10 @@ parameter type default Value description
period of all received ones.
==================== ====== ===================== =============================
.. |metrics| replace:: ["cpu_util", "memory.resident"]
.. |thresholds| replace:: {"cpu_util": 0.2, "memory.resident": 0.2}
.. |weights| replace:: {"cpu_util_weight": 1.0, "memory.resident_weight": 1.0}
.. |instance_metrics| replace:: {"cpu_util": "compute.node.cpu.percent", "memory.resident": "hardware.memory.used"}
.. |metrics| replace:: ["instance_cpu_usage", "instance_ram_usage"]
.. |thresholds| replace:: {"instance_cpu_usage": 0.2, "instance_ram_usage": 0.2}
.. |weights| replace:: {"instance_cpu_usage_weight": 1.0, "instance_ram_usage_weight": 1.0}
.. |instance_metrics| replace:: {"instance_cpu_usage": "compute.node.cpu.percent", "instance_ram_usage": "hardware.memory.used"}
.. |periods| replace:: {"instance": 720, "node": 600}
Efficacy Indicator
@ -136,8 +135,8 @@ How to use it ?
at1 workload_balancing --strategy workload_stabilization
$ openstack optimize audit create -a at1 \
-p thresholds='{"memory.resident": 0.05}' \
-p metrics='["memory.resident"]'
-p thresholds='{"instance_ram_usage": 0.05}' \
-p metrics='["instance_ram_usage"]'
External Links
--------------

24
doc/source/strategies/workload_balance.rst
View File

@ -24,8 +24,7 @@ The *workload_balance* strategy requires the following metrics:
======================= ============ ======= =========================
metric service name plugins comment
======================= ============ ======= =========================
``cpu_util`` ceilometer_ none cpu_util has been removed
since Stein.
``cpu`` ceilometer_ none
``memory.resident`` ceilometer_ none
======================= ============ ======= =========================
@ -65,15 +64,16 @@ Configuration
Strategy parameters are:
============== ====== ============= ====================================
parameter type default Value description
============== ====== ============= ====================================
``metrics`` String 'cpu_util' Workload balance base on cpu or ram
utilization. choice: ['cpu_util',
'memory.resident']
``threshold`` Number 25.0 Workload threshold for migration
``period`` Number 300 Aggregate time period of ceilometer
============== ====== ============= ====================================
============== ====== ==================== ====================================
parameter type default Value description
============== ====== ==================== ====================================
``metrics`` String 'instance_cpu_usage' Workload balance base on cpu or ram
utilization. Choices:
['instance_cpu_usage',
'instance_ram_usage']
``threshold`` Number 25.0 Workload threshold for migration
``period`` Number 300 Aggregate time period of ceilometer
============== ====== ==================== ====================================
Efficacy Indicator
------------------
@ -95,7 +95,7 @@ How to use it ?
at1 workload_balancing --strategy workload_balance
$ openstack optimize audit create -a at1 -p threshold=26.0 \
-p period=310 -p metrics=cpu_util
-p period=310 -p metrics=instance_cpu_usage
External Links
--------------

32
watcher/decision_engine/datasources/gnocchi.py
View File

@ -39,7 +39,7 @@ class GnocchiHelper(base.DataSourceBase):
host_inlet_temp='hardware.ipmi.node.temperature',
host_airflow='hardware.ipmi.node.airflow',
host_power='hardware.ipmi.node.power',
instance_cpu_usage='cpu_util',
instance_cpu_usage='cpu',
instance_ram_usage='memory.resident',
instance_ram_allocated='memory',
instance_l3_cache_usage='cpu_l3_cache',
@ -96,6 +96,25 @@ class GnocchiHelper(base.DataSourceBase):
resource_id = resources[0]['id']
if meter_name == "instance_cpu_usage":
if resource_type != "instance":
LOG.warning("Unsupported resource type for metric "
"'instance_cpu_usage': ", resource_type)
return
# The "cpu_util" gauge (percentage) metric has been removed.
# We're going to obtain the same result by using the rate of change
# aggregate operation.
if aggregate not in ("mean", "rate:mean"):
LOG.warning("Unsupported aggregate for instance_cpu_usage "
"metric: %s. "
"Supported aggregates: mean, rate:mean ",
aggregate)
return
# TODO(lpetrut): consider supporting other aggregates.
aggregate = "rate:mean"
raw_kwargs = dict(
metric=meter,
start=start_time,
@ -122,6 +141,17 @@ class GnocchiHelper(base.DataSourceBase):
# Airflow from hardware.ipmi.node.airflow is reported as
# 1/10 th of actual CFM
return_value *= 10
if meter_name == "instance_cpu_usage":
# "rate:mean" can return negative values for migrated vms.
return_value = max(0, return_value)
# We're converting the cumulative cpu time (ns) to cpu usage
# percentage.
vcpus = resource.vcpus
if not vcpus:
LOG.warning("instance vcpu count not set, assuming 1")
vcpus = 1
return_value *= 100 / (granularity * 10e+8) / vcpus
return return_value

2
watcher/decision_engine/strategy/strategies/workload_balance.py
View File

@ -295,7 +295,7 @@ class WorkloadBalance(base.WorkloadStabilizationBaseStrategy):
self.threshold)
return self.solution
# choose the server with largest cpu_util
# choose the server with largest cpu usage
source_nodes = sorted(source_nodes,
reverse=True,
key=lambda x: (x[self._meter]))

20
watcher/tests/decision_engine/datasources/test_gnocchi_helper.py
View File

@ -40,17 +40,25 @@ class TestGnocchiHelper(base.BaseTestCase):
self.addCleanup(stat_agg_patcher.stop)
def test_gnocchi_statistic_aggregation(self, mock_gnocchi):
vcpus = 2
mock_instance = mock.Mock(
id='16a86790-327a-45f9-bc82-45839f062fdc',
vcpus=vcpus)
gnocchi = mock.MagicMock()
# cpu time rate of change (ns)
mock_rate_measure = 360 * 10e+8 * vcpus * 5.5 / 100
expected_result = 5.5
expected_measures = [["2017-02-02T09:00:00.000000", 360, 5.5]]
expected_measures = [
["2017-02-02T09:00:00.000000", 360, mock_rate_measure]]
gnocchi.metric.get_measures.return_value = expected_measures
mock_gnocchi.return_value = gnocchi
helper = gnocchi_helper.GnocchiHelper()
result = helper.statistic_aggregation(
resource=mock.Mock(id='16a86790-327a-45f9-bc82-45839f062fdc'),
resource=mock_instance,
resource_type='instance',
meter_name='instance_cpu_usage',
period=300,
@ -59,6 +67,14 @@ class TestGnocchiHelper(base.BaseTestCase):
)
self.assertEqual(expected_result, result)
gnocchi.metric.get_measures.assert_called_once_with(
metric="cpu",
start=mock.ANY,
stop=mock.ANY,
resource_id=mock_instance.uuid,
granularity=360,
aggregation="rate:mean")
def test_gnocchi_statistic_series(self, mock_gnocchi):
gnocchi = mock.MagicMock()
expected_result = {