Mailing List Archive

This is on a Newton Packstack.

I try to trigger alarms based on average cpu_util of a group of
instances. *Problem: *The alarm perpetually remains in state
"insufficient data".

Ceilometer is configured to use Gnocchi and the medium archive policy
(which stores data once a minute).  The intervals in pipeline.yaml are
set to 60.

I run two instances with high CPU usage. Both have a metadata item
"metering.server_group=hicpu". The alarm uses a query
"server_group==hicpu", has a granularity of 60 and evalution periods set
to 1. I expect it to be in state /alarm /or /ok /after less than 2 minutes.

From Gnocchi, I can retrieve measures, both of the two individual
instances and of aggregate measures.

*Why "insufficient data"? **How can I find out what's going on in Aodh's
mind? *More info below. Thanks.

Bernd Bausch

My alarm:

$ openstack alarm show cpuhigh-agg
+---------------------------+--------------------------------------------------+
| Field                     |
Value                                            |
+---------------------------+--------------------------------------------------+
| aggregation_method        |
sum                                              |
| alarm_actions             |
[u'http://localhost:1234']                       |
| alarm_id                  |
6adb333a-b306-470d-b673-2c8e72c7a468             |
| comparison_operator       |
gt                                               |
| description               | gnocchi_aggregation_by_resources_threshold
alarm |
|                           |
rule                                             |
| enabled                   |
True                                             |
| evaluation_periods        |
1                                                |
| granularity               |
60                                               |
| insufficient_data_actions |
[]                                               |
| metric                    |
cpu_util                                         |
| name                      |
cpuhigh-agg                                      |
| ok_actions                |
[u'http://localhost:1234']                       |
| project_id                |
55a05c4f3908490ca2419591837575ba                 |
| query                     | {"and": [{"=":
{"created_by_project_id":         |
|                           | "55a05c4f3908490ca2419591837575ba"}},
{"=":      |
|                           | {"server_group":
"hicpu"}}]}                     |
| repeat_actions            |
False                                            |
| resource_type             |
instance                                         |
| severity                  |
low                                              |
*| state                     | insufficient
data                                |*
| state_timestamp           |
2018-07-19T11:05:38.098000                       |
| threshold                 |
80.0                                             |
| time_constraints          |
[]                                               |
| timestamp                 |
2018-07-19T11:05:38.098000                       |
| type                      |
gnocchi_aggregation_by_resources_threshold       |
| user_id                   |
96ce6a7200a54c79add0cc27ded03422                 |
+---------------------------+--------------------------------------------------+

My instances look like this:

$ openstack server show cpu-user1
+--------------------------------------+---------------------------------------+
| Field                                |
Value                                 |
+--------------------------------------+---------------------------------------+
...
| project_id                           |
55a05c4f3908490ca2419591837575ba      |
| properties                           |
*metering.server_group='hicpu'*         |
| security_groups                      | [{u'name': u'default'},
{u'name':     |
|                                      |
u'ssh'}]                              |
| status                               |
ACTIVE                                |
...
+--------------------------------------+---------------------------------------+

Gnocchi contains enough data I would think:

gnocchi measures aggregation -m cpu_util --query server_group=hicpu
--aggregation sum --resource-type instance
+---------------------------+-------------+---------------+
| timestamp                 | granularity |         value |
+---------------------------+-------------+---------------+
| 2018-07-19T09:00:00+00:00 |      3600.0 | 676.454821872 |
| 2018-07-19T10:00:00+00:00 |      3600.0 | 927.148462196 |
| 2018-07-19T09:46:00+00:00 |        60.0 | 79.0149064873 |
| 2018-07-19T09:47:00+00:00 |        60.0 | 54.6575832468 |
| 2018-07-19T09:48:00+00:00 |        60.0 | 46.0457056053 |
| 2018-07-19T09:49:00+00:00 |        60.0 | 52.5139041993 |
| 2018-07-19T09:50:00+00:00 |        60.0 | 42.7994058262 |
| 2018-07-19T09:51:00+00:00 |        60.0 | 40.0215359957 |
...