summaryrefslogtreecommitdiffstats
path: root/monitoring/ceph-mixin
diff options
context:
space:
mode:
authorAashish Sharma <aasharma@li-e74156cc-2f67-11b2-a85c-e98659a63c5c.ibm.com>2024-01-25 14:52:35 +0100
committerAashish Sharma <aasharma@li-e74156cc-2f67-11b2-a85c-e98659a63c5c.ibm.com>2024-02-07 09:57:31 +0100
commit65e671472088fcd63bb8870c8d8bd2e947d1887c (patch)
tree027572259a14423bafefb3ed5d0e4af91d46207a /monitoring/ceph-mixin
parentmgr/dashboard: Fixing RGW graph panels (diff)
downloadceph-65e671472088fcd63bb8870c8d8bd2e947d1887c.tar.xz
ceph-65e671472088fcd63bb8870c8d8bd2e947d1887c.zip
mgr/dashboards: add generated json files
Signed-off-by: Aashish Sharma <aasharma@redhat.com>
Diffstat (limited to 'monitoring/ceph-mixin')
-rw-r--r--monitoring/ceph-mixin/dashboards_out/radosgw-detail.json8
-rw-r--r--monitoring/ceph-mixin/dashboards_out/radosgw-overview.json14
-rw-r--r--monitoring/ceph-mixin/tests_dashboards/features/radosgw-detail.feature12
-rw-r--r--monitoring/ceph-mixin/tests_dashboards/features/radosgw_overview.feature24
4 files changed, 29 insertions, 29 deletions
diff --git a/monitoring/ceph-mixin/dashboards_out/radosgw-detail.json b/monitoring/ceph-mixin/dashboards_out/radosgw-detail.json
index 18ba369460a..e7c6f91e226 100644
--- a/monitoring/ceph-mixin/dashboards_out/radosgw-detail.json
+++ b/monitoring/ceph-mixin/dashboards_out/radosgw-detail.json
@@ -114,14 +114,14 @@
"steppedLine": false,
"targets": [
{
- "expr": "sum by (instance_id) (\n rate(ceph_rgw_get_initial_lat_sum{job=~\"$job\"}[$__rate_interval]) /\n rate(ceph_rgw_get_initial_lat_count{job=~\"$job\"}[$__rate_interval])\n) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\", ceph_daemon=~\"$rgw_servers\"}\n",
+ "expr": "sum by (instance_id) (\n rate(ceph_rgw_op_get_obj_lat_sum{job=~\"$job\"}[$__rate_interval]) /\n rate(ceph_rgw_op_get_obj_lat_count{job=~\"$job\"}[$__rate_interval])\n) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\", ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "GET {{ceph_daemon}}",
"refId": "A"
},
{
- "expr": "sum by (instance_id) (\n rate(ceph_rgw_put_initial_lat_sum{job=~\"$job\"}[$__rate_interval]) /\n rate(ceph_rgw_put_initial_lat_count{job=~\"$job\"}[$__rate_interval])\n) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\", ceph_daemon=~\"$rgw_servers\"}\n",
+ "expr": "sum by (instance_id) (\n rate(ceph_rgw_op_put_obj_lat_sum{job=~\"$job\"}[$__rate_interval]) /\n rate(ceph_rgw_op_put_obj_lat_count{job=~\"$job\"}[$__rate_interval])\n) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\", ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "PUT {{ceph_daemon}}",
@@ -216,14 +216,14 @@
"steppedLine": false,
"targets": [
{
- "expr": "rate(ceph_rgw_get_b{job=~\"$job\"}[$__rate_interval]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\", ceph_daemon=~\"$rgw_servers\"}\n",
+ "expr": "rate(ceph_rgw_op_get_obj_bytes{job=~\"$job\"}[$__rate_interval]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\", ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "GETs {{ceph_daemon}}",
"refId": "A"
},
{
- "expr": "rate(ceph_rgw_put_b{job=~\"$job\"}[$__rate_interval]) *\n on (instance_id) group_left (ceph_daemon)\n ceph_rgw_metadata{job=~\"$job\", ceph_daemon=~\"$rgw_servers\"}\n",
+ "expr": "rate(ceph_rgw_op_put_obj_bytes{job=~\"$job\"}[$__rate_interval]) *\n on (instance_id) group_left (ceph_daemon)\n ceph_rgw_metadata{job=~\"$job\", ceph_daemon=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "PUTs {{ceph_daemon}}",
diff --git a/monitoring/ceph-mixin/dashboards_out/radosgw-overview.json b/monitoring/ceph-mixin/dashboards_out/radosgw-overview.json
index 92e416764e7..3cb8a7c76ac 100644
--- a/monitoring/ceph-mixin/dashboards_out/radosgw-overview.json
+++ b/monitoring/ceph-mixin/dashboards_out/radosgw-overview.json
@@ -108,14 +108,14 @@
"steppedLine": false,
"targets": [
{
- "expr": "label_replace(\n rate(ceph_rgw_get_initial_lat_sum{job=~\"$job\"}[$__rate_interval]) /\n rate(ceph_rgw_get_initial_lat_count{job=~\"$job\"}[$__rate_interval]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\"},\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
+ "expr": "label_replace(\n rate(ceph_rgw_op_get_obj_lat_sum{job=~\"$job\"}[$__rate_interval]) /\n rate(ceph_rgw_op_get_obj_lat_count{job=~\"$job\"}[$__rate_interval]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\"},\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "GET {{rgw_host}}",
"refId": "A"
},
{
- "expr": "label_replace(\n rate(ceph_rgw_put_initial_lat_sum{job=~\"$job\"}[$__rate_interval]) /\n rate(ceph_rgw_put_initial_lat_count{job=~\"$job\"}[$__rate_interval]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\"},\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
+ "expr": "label_replace(\n rate(ceph_rgw_op_put_obj_lat_sum{job=~\"$job\"}[$__rate_interval]) /\n rate(ceph_rgw_op_put_obj_lat_count{job=~\"$job\"}[$__rate_interval]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\"},\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "PUT {{rgw_host}}",
@@ -305,7 +305,7 @@
"steppedLine": false,
"targets": [
{
- "expr": "label_replace(\n rate(ceph_rgw_get_initial_lat_sum{job=~\"$job\"}[$__rate_interval]) /\n rate(ceph_rgw_get_initial_lat_count{job=~\"$job\"}[$__rate_interval]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\"},\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
+ "expr": "label_replace(\n rate(ceph_rgw_op_get_obj_lat_sum{job=~\"$job\"}[$__rate_interval]) /\n rate(ceph_rgw_op_get_obj_lat_count{job=~\"$job\"}[$__rate_interval]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\"},\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{rgw_host}}",
@@ -400,14 +400,14 @@
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(ceph_rgw_get_b{job=~\"$job\"}[$__rate_interval]))",
+ "expr": "sum(rate(ceph_rgw_op_get_obj_bytes{job=~\"$job\"}[$__rate_interval]))",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "GETs",
"refId": "A"
},
{
- "expr": "sum(rate(ceph_rgw_put_b{job=~\"$job\"}[$__rate_interval]))",
+ "expr": "sum(rate(ceph_rgw_op_put_obj_bytes{job=~\"$job\"}[$__rate_interval]))",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "PUTs",
@@ -502,7 +502,7 @@
"steppedLine": false,
"targets": [
{
- "expr": "label_replace(sum by (instance_id) (\n rate(ceph_rgw_get_b{job=~\"$job\"}[$__rate_interval]) +\n rate(ceph_rgw_put_b{job=~\"$job\"}[$__rate_interval])) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\"},\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
+ "expr": "label_replace(sum by (instance_id) (\n rate(ceph_rgw_op_get_obj_bytes{job=~\"$job\"}[$__rate_interval]) +\n rate(ceph_rgw_op_put_obj_bytes{job=~\"$job\"}[$__rate_interval])) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\"},\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{rgw_host}}",
@@ -597,7 +597,7 @@
"steppedLine": false,
"targets": [
{
- "expr": "label_replace(\n rate(ceph_rgw_put_initial_lat_sum{job=~\"$job\"}[$__rate_interval]) /\n rate(ceph_rgw_put_initial_lat_count{job=~\"$job\"}[$__rate_interval]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\"},\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
+ "expr": "label_replace(\n rate(ceph_rgw_op_put_obj_lat_sum{job=~\"$job\"}[$__rate_interval]) /\n rate(ceph_rgw_op_put_obj_lat_count{job=~\"$job\"}[$__rate_interval]) *\n on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{job=~\"$job\"},\n \"rgw_host\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{rgw_host}}",
diff --git a/monitoring/ceph-mixin/tests_dashboards/features/radosgw-detail.feature b/monitoring/ceph-mixin/tests_dashboards/features/radosgw-detail.feature
index e0016c5077d..f819fc6dc41 100644
--- a/monitoring/ceph-mixin/tests_dashboards/features/radosgw-detail.feature
+++ b/monitoring/ceph-mixin/tests_dashboards/features/radosgw-detail.feature
@@ -3,8 +3,8 @@ 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_op_get_obj_lat_sum{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 10 50 100 |
+ | ceph_rgw_op_get_obj_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`
@@ -15,8 +15,8 @@ Scenario: "Test $rgw_servers GET/PUT Latencies - GET"
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_op_put_obj_lat_sum{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 15 35 55 |
+ | ceph_rgw_op_put_obj_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`
@@ -27,7 +27,7 @@ Scenario: "Test $rgw_servers GET/PUT Latencies - PUT"
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_op_get_obj_bytes{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`
@@ -38,7 +38,7 @@ Scenario: "Test Bandwidth by HTTP Operation - GET"
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_op_put_obj_bytes{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`
diff --git a/monitoring/ceph-mixin/tests_dashboards/features/radosgw_overview.feature b/monitoring/ceph-mixin/tests_dashboards/features/radosgw_overview.feature
index 642e4397878..14a6e5543a4 100644
--- a/monitoring/ceph-mixin/tests_dashboards/features/radosgw_overview.feature
+++ b/monitoring/ceph-mixin/tests_dashboards/features/radosgw_overview.feature
@@ -3,8 +3,8 @@ Feature: RGW Overview Dashboard
Scenario: "Test Average GET Latencies"
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_op_get_obj_lat_sum{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 10 50 100 |
+ | ceph_rgw_op_get_obj_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 `Average GET/PUT Latencies by RGW Instance` with legend `GET {{rgw_host}}` shows:
@@ -14,8 +14,8 @@ Scenario: "Test Average GET Latencies"
Scenario: "Test Average PUT Latencies"
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_op_put_obj_lat_sum{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 15 35 55 |
+ | ceph_rgw_op_put_obj_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`
Then Grafana panel `Average GET/PUT Latencies by RGW Instance` with legend `PUT {{rgw_host}}` shows:
@@ -35,8 +35,8 @@ Scenario: "Test Total Requests/sec by RGW Instance"
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_op_get_obj_lat_sum{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 10 50 100 |
+ | ceph_rgw_op_get_obj_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:
@@ -46,7 +46,7 @@ Scenario: "Test GET Latencies by RGW Instance"
Scenario: "Test Bandwidth Consumed by Type- 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_op_get_obj_bytes{instance="127.0.0.1", instance_id="92806566", job="ceph"} | 10 50 100 |
When evaluation time is `1m`
And interval is `30s`
Then Grafana panel `Bandwidth Consumed by Type` with legend `GETs` shows:
@@ -56,7 +56,7 @@ Scenario: "Test Bandwidth Consumed by Type- GET"
Scenario: "Test Bandwidth Consumed by Type- 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_op_put_obj_bytes{instance="127.0.0.1", instance_id="92806566", job="ceph"} | 5 20 50 |
When evaluation time is `1m`
And interval is `30s`
Then Grafana panel `Bandwidth Consumed by Type` with legend `PUTs` shows:
@@ -66,8 +66,8 @@ Scenario: "Test Bandwidth Consumed by Type- PUT"
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_op_get_obj_bytes{instance="127.0.0.1", instance_id="92806566", job="ceph"} | 10 50 100 |
+ | ceph_rgw_op_put_obj_bytes{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`
@@ -78,8 +78,8 @@ Scenario: "Test Bandwidth by RGW Instance"
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_op_put_obj_lat_sum{instance="127.0.0.1", instance_id="58892247", job="ceph"} | 15 35 55 |
+ | ceph_rgw_op_put_obj_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`