monitoring/grafana: Add tests for radosgw panels

Some of the expressions modified in c40290390d7 were not covered by any tests,
especially those in the `radosgw-detail.json` dashboard.

This commit fills in those gaps.

Signed-off-by: Benoît Knecht <bknecht@protonmail.ch>

2 files changed, 174 insertions, 0 deletions

diff --git a/monitoring/grafana/dashboards/tests/features/radosgw-detail.feature b/monitoring/grafana/dashboards/tests/features/radosgw-detail.feature
new file mode 100644
index 00000000000..bcc793a21a5
--- /dev/null
+++ b/monitoring/grafana/dashboards/tests/features/radosgw-detail.feature
@@ -0,0 +1,139 @@
+Feature: RGW Host Detail Dashboard
+
+Scenario: "Test $rgw_servers GET/PUT Latencies - GET"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_get_initial_lat_sum{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 10 50 100 |
+    | ceph_rgw_get_initial_lat_count{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 20 60 80 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.foo", hostname="localhost", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 1 1 |
+  When interval is `30s`
+  And variable `rgw_servers` is `rgw.foo`
+  Then Grafana panel `$rgw_servers GET/PUT Latencies` with legend `GET {{ceph_daemon}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.foo", instance_id="58892247"} | 2.5000000000000004 |
+
+Scenario: "Test $rgw_servers GET/PUT Latencies - PUT"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_put_initial_lat_sum{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 15 35 55 |
+    | ceph_rgw_put_initial_lat_count{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 10 30 50 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.foo", hostname="localhost", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 1 1 |
+  When interval is `30s`
+  And variable `rgw_servers` is `rgw.foo`
+  Then Grafana panel `$rgw_servers GET/PUT Latencies` with legend `PUT {{ceph_daemon}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.foo", instance_id="58892247"} | 1 |
+
+Scenario: "Test Bandwidth by HTTP Operation - GET"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_get_b{instance="127.0.0.1", instance_id="92806566", job="ceph"} | 10 50 100 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.1", hostname="localhost", instance="127.0.0.1", instance_id="92806566", job="ceph"} | 1 1 1 |
+  When interval is `30s`
+  And variable `rgw_servers` is `rgw.1`
+  Then Grafana panel `Bandwidth by HTTP Operation` with legend `GETs {{ceph_daemon}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.1", instance="127.0.0.1", instance_id="92806566", job="ceph"} | 1.6666666666666667 |
+
+Scenario: "Test Bandwidth by HTTP Operation - PUT"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_put_b{instance="127.0.0.1", instance_id="92806566", job="ceph"} | 5 20 50 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.1", hostname="localhost", instance="127.0.0.1", instance_id="92806566", job="ceph"} | 1 1 1 |
+  When interval is `30s`
+  And variable `rgw_servers` is `rgw.1`
+  Then Grafana panel `Bandwidth by HTTP Operation` with legend `PUTs {{ceph_daemon}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.1", instance="127.0.0.1", instance_id="92806566", job="ceph"} | 1 |
+
+Scenario: "Test HTTP Request Breakdown - Requests Failed"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_failed_req{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 5 7 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.foo", hostname="localhost", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 1 1 |
+  When interval is `30s`
+  And variable `rgw_servers` is `rgw.foo`
+  Then Grafana panel `HTTP Request Breakdown` with legend `Requests Failed {{ceph_daemon}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.foo", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 6.666666666666667e-02 |
+
+Scenario: "Test HTTP Request Breakdown - GET"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_get{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 100 150 170 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.foo", hostname="localhost", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 1 1 |
+  When interval is `30s`
+  And variable `rgw_servers` is `rgw.foo`
+  Then Grafana panel `HTTP Request Breakdown` with legend `GETs {{ceph_daemon}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.foo", instance="127.0.0.1", instance_id="58892247", job="ceph"} | .6666666666666666 |
+
+Scenario: "Test HTTP Request Breakdown - PUT"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_put{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 70 90 160 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.foo", hostname="localhost", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 1 1 |
+  When interval is `30s`
+  And variable `rgw_servers` is `rgw.foo`
+  Then Grafana panel `HTTP Request Breakdown` with legend `PUTs {{ceph_daemon}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.foo", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 2.3333333333333335 |
+
+Scenario: "Test HTTP Request Breakdown - Other"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_req{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 175 250 345 |
+    | ceph_rgw_get{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 100 150 170 |
+    | ceph_rgw_put{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 70 90 160 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.foo", hostname="localhost", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 1 1 |
+  When interval is `30s`
+  And variable `rgw_servers` is `rgw.foo`
+  Then Grafana panel `HTTP Request Breakdown` with legend `Other {{ceph_daemon}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.foo", instance="127.0.0.1", instance_id="58892247", job="ceph"} | .16666666666666652 |
+
+Scenario: "Test Workload Breakdown - Failures"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_failed_req{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 5 7 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.foo", hostname="localhost", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 1 1 |
+  When interval is `30s`
+  And variable `rgw_servers` is `rgw.foo`
+  Then Grafana panel `Workload Breakdown` with legend `Failures {{ceph_daemon}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.foo", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 6.666666666666667e-02 |
+
+Scenario: "Test Workload Breakdown - GETs"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_get{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 100 150 170 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.foo", hostname="localhost", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 1 1 |
+  When interval is `30s`
+  And variable `rgw_servers` is `rgw.foo`
+  Then Grafana panel `Workload Breakdown` with legend `GETs {{ceph_daemon}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.foo", instance="127.0.0.1", instance_id="58892247", job="ceph"} | .6666666666666666 |
+
+Scenario: "Test Workload Breakdown - PUTs"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_put{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 70 90 160 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.foo", hostname="localhost", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 1 1 |
+  When interval is `30s`
+  And variable `rgw_servers` is `rgw.foo`
+  Then Grafana panel `Workload Breakdown` with legend `PUTs {{ceph_daemon}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.foo", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 2.3333333333333335 |
+
+Scenario: "Test Workload Breakdown - Other"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_req{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 175 250 345 |
+    | ceph_rgw_get{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 100 150 170 |
+    | ceph_rgw_put{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 70 90 160 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.foo", hostname="localhost", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 1 1 |
+  When interval is `30s`
+  And variable `rgw_servers` is `rgw.foo`
+  Then Grafana panel `Workload Breakdown` with legend `Other (DELETE,LIST) {{ceph_daemon}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.foo", instance="127.0.0.1", instance_id="58892247", job="ceph"} | .16666666666666652 |
diff --git a/monitoring/grafana/dashboards/tests/features/radosgw_overview.feature b/monitoring/grafana/dashboards/tests/features/radosgw_overview.feature
index 3e9724ee2dc..69e46b1d511 100644
--- a/monitoring/grafana/dashboards/tests/features/radosgw_overview.feature
+++ b/monitoring/grafana/dashboards/tests/features/radosgw_overview.feature
@@ -32,6 +32,17 @@ Scenario: "Test Total Requests/sec by RGW Instance"
     | metrics | values |
     | {rgw_host="1"} | 1.6666666666666667 |
 
+Scenario: "Test GET Latencies by RGW Instance"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_get_initial_lat_sum{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 10 50 100 |
+    | ceph_rgw_get_initial_lat_count{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 20 60 80 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.foo", hostname="localhost", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 1 1 |
+  When interval is `30s`
+  Then Grafana panel `GET Latencies by RGW Instance` with legend `{{rgw_host}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.foo", instance="127.0.0.1", instance_id="58892247", job="ceph", rgw_host="foo"} | 2.5000000000000004 |
+
 Scenario: "Test Bandwidth Consumed by Type- GET"
   Given the following series:
     | metrics | values |
@@ -52,6 +63,30 @@ Scenario: "Test Bandwidth Consumed by Type- PUT"
     | metrics | values |
     | {} | 1 |
 
+Scenario: "Test Bandwidth by RGW Instance"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_get_b{instance="127.0.0.1", instance_id="92806566", job="ceph"} | 10 50 100 |
+    | ceph_rgw_put_b{instance="127.0.0.1", instance_id="92806566", job="ceph"} | 5 20 50 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.1", hostname="localhost", instance="127.0.0.1", instance_id="92806566", job="ceph"} | 1 1 1 |
+  When evaluation time is `1m`
+  And interval is `30s`
+  Then Grafana panel `Bandwidth by RGW Instance` with legend `{{rgw_host}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.1", instance_id="92806566", rgw_host="1"} | 2.666666666666667 |
+
+Scenario: "Test PUT Latencies by RGW Instance"
+  Given the following series:
+    | metrics | values |
+    | ceph_rgw_put_initial_lat_sum{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 15 35 55 |
+    | ceph_rgw_put_initial_lat_count{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 10 30 50 |
+    | ceph_rgw_metadata{ceph_daemon="rgw.foo", hostname="localhost", instance="127.0.0.1", instance_id="58892247", job="ceph"} | 1 1 1 |
+  When evaluation time is `1m`
+  And interval is `30s`
+  Then Grafana panel `PUT Latencies by RGW Instance` with legend `{{rgw_host}}` shows:
+    | metrics | values |
+    | {ceph_daemon="rgw.foo", instance="127.0.0.1", instance_id="58892247", job="ceph", rgw_host="foo"} | 1 |
+
 Scenario: "Test Total backend responses by HTTP code"
   Given the following series:
     | metrics | values |