-
Bug
-
Resolution: Done
-
Normal
-
None
-
None
-
None
-
5
-
False
-
None
-
False
-
COST-5108 - Attribute node attached storage to OpenShift Projects
-
-
Reproducer:
Fetch the testing MR
(Optional)
Modify the fixture
cost_get_aws_managed_disks
like bellow to get always the same disk size
If you don't do this, every run will pick a differnt size
aws_managed_disk_one_candidates = [
#(11, "st1"),
#(12, "sc1"),
(16, "sc1"),
#(32, "io1"),
#(64, "gp3"),
#(256, "io2"),
]
Run the following test with a breakpoint:
ENV_FOR_DYNACONF=local iqe tests plugin cost_management --trace -k 'test_api_ocp_on_aws_csi_storage_raw_calc' -raA
THis test will ingest an OCP on AWS cluster named test_cost_ocp_on_aws
Grab the unattributed cost
http://localhost:8000/api/cost-management/v1/reports/openshift/volumes/?group_by[project]=*&filter[resolution]=monthly&filter[time_scope_value]=-2&filter[time_scope_units]=month&filter[cluster]=test_cost_ocp_on_aws
value": 62.33726785258596
Trino disk capacity:
vol-aws_single_source_disk_one | 16 | 2024-08-14 00:00:00.000 | 499d5a1c-57fb-4f53-aa89-98b759d8e462 | 2024 | 08
The test will pass
Now run the test:
ENV_FOR_DYNACONF=local iqe tests plugin cost_management --trace -k 'test_api_ocp_on_aws_csi_storage_shared_disk_raw_calc' -raA
This test will ingest 3 OCP on AWS clusters, 2 of them with unatributed storage project:
test_cost_ocp_on_aws
and test_cost_ocp_charlie
Grab the same cost than before
http://localhost:8000/api/cost-management/v1/reports/openshift/volumes/?group_by[project]=*&filter[resolution]=monthly&filter[time_scope_value]=-2&filter[time_scope_units]=month&filter[cluster]=test_cost_ocp_on_aws
"value": 41.55817856839064
Trino disk Capacity:
vol-aws_single_source_disk_one | 16 | 2024-08-01 00:00:00.000 | bf37e7bd-1edf-4d30-ae49-51f9279c9c4d | 2024 | 08
The UUID is differnt because they are differnt sources, recreated
The cost for the other cluster is off, too
http://localhost:8000/api/cost-management/v1/reports/openshift/infrastructures/aws/storage/?group_by[project]=*&filter[resolution]=monthly&filter[time_scope_value]=-2&filter[time_scope_units]=month&filter[cluster]=test_cost_ocp_charlie
Storage unattributed = 1232.000000 vs the expected 1612.000000027125
My guess is that when calculating the unattributed cost, each cluster is not picking up the correct dik size in the trino table