Skip to content

Commit

Permalink
Merge pull request ceph#55314 from cloudbehl/rgw-dashboard-json
Browse files Browse the repository at this point in the history 
mgr/dashboard: Fixing RGW graph panels


Reviewed-by: Aashish Sharma <[email protected]>
  • Loading branch information
@aaSharma14
aaSharma14 authored Feb 13, 2024
2 parents 137ea0f + 65e6714 commit a85baa8
Show file tree
Hide file tree
Showing 6 changed files with 51 additions and 51 deletions.
36 changes: 18 additions & 18 deletions monitoring/ceph-mixin/dashboards/rgw.libsonnet
View file
Original file line number Diff line number Diff line change
Expand Up @@ -263,8 +263,8 @@ local u = import 'utils.libsonnet';
'short',
|||
label_replace(
rate(ceph_rgw_get_initial_lat_sum{%(matchers)s}[$__rate_interval]) /
rate(ceph_rgw_get_initial_lat_count{%(matchers)s}[$__rate_interval]) *
rate(ceph_rgw_op_get_obj_lat_sum{%(matchers)s}[$__rate_interval]) /
rate(ceph_rgw_op_get_obj_lat_count{%(matchers)s}[$__rate_interval]) *
on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{%(matchers)s},
"rgw_host", "$1", "ceph_daemon", "rgw.(.*)"
)
Expand All @@ -279,8 +279,8 @@ local u = import 'utils.libsonnet';
$.addTargetSchema(
|||
label_replace(
rate(ceph_rgw_put_initial_lat_sum{%(matchers)s}[$__rate_interval]) /
rate(ceph_rgw_put_initial_lat_count{%(matchers)s}[$__rate_interval]) *
rate(ceph_rgw_op_put_obj_lat_sum{%(matchers)s}[$__rate_interval]) /
rate(ceph_rgw_op_put_obj_lat_count{%(matchers)s}[$__rate_interval]) *
on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{%(matchers)s},
"rgw_host", "$1", "ceph_daemon", "rgw.(.*)"
)
Expand Down Expand Up @@ -316,8 +316,8 @@ local u = import 'utils.libsonnet';
'short',
|||
label_replace(
rate(ceph_rgw_get_initial_lat_sum{%(matchers)s}[$__rate_interval]) /
rate(ceph_rgw_get_initial_lat_count{%(matchers)s}[$__rate_interval]) *
rate(ceph_rgw_op_get_obj_lat_sum{%(matchers)s}[$__rate_interval]) /
rate(ceph_rgw_op_get_obj_lat_count{%(matchers)s}[$__rate_interval]) *
on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{%(matchers)s},
"rgw_host", "$1", "ceph_daemon", "rgw.(.*)"
)
Expand All @@ -333,14 +333,14 @@ local u = import 'utils.libsonnet';
'Total bytes transferred in/out of all radosgw instances within the cluster',
'bytes',
'short',
'sum(rate(ceph_rgw_get_b{%(matchers)s}[$__rate_interval]))' % $.matchers(),
'sum(rate(ceph_rgw_op_get_obj_bytes{%(matchers)s}[$__rate_interval]))' % $.matchers(),
'GETs',
0,
8,
8,
6
).addTargets(
[$.addTargetSchema('sum(rate(ceph_rgw_put_b{%(matchers)s}[$__rate_interval]))' % $.matchers(),
[$.addTargetSchema('sum(rate(ceph_rgw_op_put_obj_bytes{%(matchers)s}[$__rate_interval]))' % $.matchers(),
'PUTs')]
),
RgwOverviewPanel(
Expand All @@ -350,8 +350,8 @@ local u = import 'utils.libsonnet';
'short',
|||
label_replace(sum by (instance_id) (
rate(ceph_rgw_get_b{%(matchers)s}[$__rate_interval]) +
rate(ceph_rgw_put_b{%(matchers)s}[$__rate_interval])) *
rate(ceph_rgw_op_get_obj_bytes{%(matchers)s}[$__rate_interval]) +
rate(ceph_rgw_op_put_obj_bytes{%(matchers)s}[$__rate_interval])) *
on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{%(matchers)s},
"rgw_host", "$1", "ceph_daemon", "rgw.(.*)"
)
Expand All @@ -369,8 +369,8 @@ local u = import 'utils.libsonnet';
'short',
|||
label_replace(
rate(ceph_rgw_put_initial_lat_sum{%(matchers)s}[$__rate_interval]) /
rate(ceph_rgw_put_initial_lat_count{%(matchers)s}[$__rate_interval]) *
rate(ceph_rgw_op_put_obj_lat_sum{%(matchers)s}[$__rate_interval]) /
rate(ceph_rgw_op_put_obj_lat_count{%(matchers)s}[$__rate_interval]) *
on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{%(matchers)s},
"rgw_host", "$1", "ceph_daemon", "rgw.(.*)"
)
Expand Down Expand Up @@ -736,14 +736,14 @@ local u = import 'utils.libsonnet';
'short',
|||
sum by (instance_id) (
rate(ceph_rgw_get_initial_lat_sum{%(matchers)s}[$__rate_interval]) /
rate(ceph_rgw_get_initial_lat_count{%(matchers)s}[$__rate_interval])
rate(ceph_rgw_op_get_obj_lat_sum{%(matchers)s}[$__rate_interval]) /
rate(ceph_rgw_op_get_obj_lat_count{%(matchers)s}[$__rate_interval])
) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{%(matchers)s, ceph_daemon=~"$rgw_servers"}
||| % $.matchers(),
|||
sum by (instance_id) (
rate(ceph_rgw_put_initial_lat_sum{%(matchers)s}[$__rate_interval]) /
rate(ceph_rgw_put_initial_lat_count{%(matchers)s}[$__rate_interval])
rate(ceph_rgw_op_put_obj_lat_sum{%(matchers)s}[$__rate_interval]) /
rate(ceph_rgw_op_put_obj_lat_count{%(matchers)s}[$__rate_interval])
) * on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{%(matchers)s, ceph_daemon=~"$rgw_servers"}
||| % $.matchers(),
'GET {{ceph_daemon}}',
Expand All @@ -760,11 +760,11 @@ local u = import 'utils.libsonnet';
'bytes',
'short',
|||
rate(ceph_rgw_get_b{%(matchers)s}[$__rate_interval]) *
rate(ceph_rgw_op_get_obj_bytes{%(matchers)s}[$__rate_interval]) *
on (instance_id) group_left (ceph_daemon) ceph_rgw_metadata{%(matchers)s, ceph_daemon=~"$rgw_servers"}
||| % $.matchers(),
|||
rate(ceph_rgw_put_b{%(matchers)s}[$__rate_interval]) *
rate(ceph_rgw_op_put_obj_bytes{%(matchers)s}[$__rate_interval]) *
on (instance_id) group_left (ceph_daemon)
ceph_rgw_metadata{%(matchers)s, ceph_daemon=~"$rgw_servers"}
||| % $.matchers(),
Expand Down
8 changes: 4 additions & 4 deletions monitoring/ceph-mixin/dashboards_out/radosgw-detail.json
View file
Original file line number Diff line number Diff line change
Expand Up @@ -114,14 +114,14 @@
"steppedLine": false,
"targets": [
{
"expr": "sum by (instance_id) (\n rate(ceph_rgw_get_initial_lat_sum{job=~\"$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=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "PUT {{ceph_daemon}}",
Expand Down Expand Up @@ -216,14 +216,14 @@
"steppedLine": false,
"targets": [
{
"expr": "rate(ceph_rgw_get_b{job=~\"$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=~\"$rgw_servers\"}\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "PUTs {{ceph_daemon}}",
Expand Down
14 changes: 7 additions & 7 deletions monitoring/ceph-mixin/dashboards_out/radosgw-overview.json
View file
Original file line number Diff line number Diff line change
Expand Up @@ -108,14 +108,14 @@
"steppedLine": false,
"targets": [
{
"expr": "label_replace(\n rate(ceph_rgw_get_initial_lat_sum{job=~\"$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\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "PUT {{rgw_host}}",
Expand Down Expand Up @@ -305,7 +305,7 @@
"steppedLine": false,
"targets": [
{
"expr": "label_replace(\n rate(ceph_rgw_get_initial_lat_sum{job=~\"$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\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{rgw_host}}",
Expand Down Expand Up @@ -400,14 +400,14 @@
"steppedLine": false,
"targets": [
{
"expr": "sum(rate(ceph_rgw_get_b{job=~\"$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\"}[$__rate_interval]))",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "PUTs",
Expand Down Expand Up @@ -502,7 +502,7 @@
"steppedLine": false,
"targets": [
{
"expr": "label_replace(sum by (instance_id) (\n rate(ceph_rgw_get_b{job=~\"$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\", \"$1\", \"ceph_daemon\", \"rgw.(.*)\"\n)\n",
"format": "time_series",
"intervalFactor": 1,
"legendFormat": "{{rgw_host}}",
Expand Down Expand Up @@ -597,7 +597,7 @@
"steppedLine": false,
"targets": [
{
"expr": "label_replace(\n rate(ceph_rgw_put_initial_lat_sum{job=~\"$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}}",
Expand Down
12 changes: 6 additions & 6 deletions monitoring/ceph-mixin/tests_dashboards/features/radosgw-detail.feature
View file
Original file line number Diff line number Diff line change
Expand Up @@ -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`
Expand All @@ -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`
Expand All @@ -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`
Expand All @@ -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`
Expand Down
Loading

0 comments on commit a85baa8

Please sign in to comment.