mgr/dashboard: fix grafana dashboards

Fixes: https://tracker.ceph.com/issues/42542
Sort order was wrong for some dashboards,
fixed empty / buggy Top 3 clients IOPS by pool / Throughput - in Pools
Overall performance
fixed Avg utilization Multiple series found - in Host Overall
performance
Fixed invalid dimensions for plot - in OSD Overall performance

Signed-off-by: Radu Toader <radu.m.toader@gmail.com>
(cherry picked from commit 3beaf63761590fd35a72777fc56c0d5f4a6cad3b)

diff --git a/monitoring/grafana/dashboards/cephfs-overview.json b/monitoring/grafana/dashboards/cephfs-overview.json
index b131cc2079f1dc94abda27b650342a47a4a0055d..57922f55155a9f429ee5557d651d1fd1ee6cd5ad 100644 (file)
--- a/monitoring/grafana/dashboards/cephfs-overview.json
+++ b/monitoring/grafana/dashboards/cephfs-overview.json
@@ -109,7 +109,7 @@
       "title": "MDS Workload - $mds_servers",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -193,7 +193,7 @@
       "title": "Client Request Load - $mds_servers",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",

diff --git a/monitoring/grafana/dashboards/host-details.json b/monitoring/grafana/dashboards/host-details.json
index df68b65e34b5845905a579494aed118231e6272f..c4d139a68c007b5d227e231ead74ce7cb009b9eb 100644 (file)
--- a/monitoring/grafana/dashboards/host-details.json
+++ b/monitoring/grafana/dashboards/host-details.json
@@ -207,7 +207,7 @@
       "title": "CPU Utilisation",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -330,7 +330,7 @@
       "title": "RAM Usage",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -434,7 +434,7 @@
       "title": "Network Load",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -631,7 +631,7 @@
       "tableColumn": "",
       "targets": [
         {
-          "expr": "sum(ceph_osd_stat_bytes and on (ceph_daemon) ceph_disk_occupation{exported_instance=~\"($ceph_hosts).*\"})",
+          "expr": "sum(ceph_osd_stat_bytes and on (ceph_daemon) ceph_disk_occupation{instance=~\"($ceph_hosts).*\"})",
           "format": "time_series",
           "intervalFactor": 2,
           "refId": "A",
@@ -808,7 +808,7 @@
       "steppedLine": false,
       "targets": [
         {
-          "expr": "(irate(node_disk_writes_completed{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_writes_completed_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+          "expr": "(irate(node_disk_writes_completed{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_writes_completed_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
           "format": "time_series",
           "intervalFactor": 1,
           "legendFormat": "{{device}}({{ceph_daemon}}) writes",
@@ -817,7 +817,7 @@
           "textEditor": true
         },
         {
-          "expr": "(irate(node_disk_reads_completed{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_reads_completed_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+          "expr": "(irate(node_disk_reads_completed{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_reads_completed_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
           "format": "time_series",
           "hide": false,
           "intervalFactor": 1,
@@ -848,7 +848,7 @@
           "label": "Read (-) / Write (+)",
           "logBase": 1,
           "max": null,
-          "min": "0",
+          "min": null,
           "show": true
         },
         {
@@ -909,14 +909,14 @@
       "steppedLine": false,
       "targets": [
         {
-          "expr": "(irate(node_disk_bytes_written{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_written_bytes_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+          "expr": "(irate(node_disk_bytes_written{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_written_bytes_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
           "format": "time_series",
           "intervalFactor": 1,
           "legendFormat": "{{device}}({{ceph_daemon}}) write",
           "refId": "B"
         },
         {
-          "expr": "(irate(node_disk_bytes_read{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_read_bytes_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+          "expr": "(irate(node_disk_bytes_read{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_read_bytes_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
           "format": "time_series",
           "intervalFactor": 1,
           "legendFormat": "{{device}}({{ceph_daemon}}) read",
@@ -946,7 +946,7 @@
           "label": "Read (-) / Write (+)",
           "logBase": 1,
           "max": null,
-          "min": "0",
+          "min": null,
           "show": true
         },
         {
@@ -1002,11 +1002,11 @@
       "steppedLine": false,
       "targets": [
         {
-          "expr": "max by(instance,device) (((\n  (irate(node_disk_write_time_ms{ instance=~\"($ceph_hosts).*\"}[5m]) )\n  / clamp_min(irate(node_disk_writes_completed{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001) or \n  (irate(node_disk_read_time_ms{ instance=~\"($ceph_hosts).*\"}[5m]) )\n  / clamp_min(irate(node_disk_reads_completed{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001)\n  ) / 1000) or\n  (irate(node_disk_write_time_seconds_total{ instance=~\"($ceph_hosts).*\"}[5m]) )\n  / clamp_min(irate(node_disk_writes_completed_total{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001) or \n  (irate(node_disk_read_time_seconds_total{ instance=~\"($ceph_hosts).*\"}[5m]) )\n  / clamp_min(irate(node_disk_reads_completed_total{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001)\n  ) *\n  on(instance,device) group_left(osd_id) label_replace(ceph_disk_occupation,\"osd_id\",\"$1\",\"ceph_daemon\",\"osd.(.*)\")",
+          "expr": "max by(instance,device) ((irate(node_disk_write_time_seconds_total{ instance=~\"($ceph_hosts).*\"}[5m]) )  / clamp_min(irate(node_disk_writes_completed_total{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001) or   (irate(node_disk_read_time_seconds_total{ instance=~\"($ceph_hosts).*\"}[5m]) )  / clamp_min(irate(node_disk_reads_completed_total{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001)  ) *  on(instance,device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation{instance=~\"($ceph_hosts).*\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
           "format": "time_series",
           "hide": false,
           "intervalFactor": 1,
-          "legendFormat": "{{device}}({{osd_id}})",
+          "legendFormat": "{{device}}({{ceph_daemon}})",
           "refId": "D"
         }
       ],
@@ -1058,6 +1058,7 @@
       "datasource": "$datasource",
       "description": "Show disk utilization % (util) of any OSD devices on the host by the physical device name and associated OSD id.",
       "fill": 1,
+      "fillGradient": 0,
       "gridPos": {
         "h": 9,
         "w": 11,
@@ -1077,8 +1078,11 @@
       "lines": true,
       "linewidth": 1,
       "links": [],
-      "minSpan": 12,
+      "maxPerRow": 2,
       "nullPointMode": "connected",
+      "options": {
+        "dataLinks": []
+      },
       "percentage": false,
       "pointradius": 5,
       "points": false,
@@ -1089,24 +1093,17 @@
       "steppedLine": false,
       "targets": [
         {
-          "expr": "(\n  (irate(node_disk_io_time_ms[5m]) / 10 ) or\n  irate(node_disk_io_time_seconds_total[5m]) * 100)\n* on(instance, device) group_left(osd_id) label_replace(ceph_disk_occupation{instance=~\"($ceph_hosts).*\"}, \"osd_id\",\"$1\",\"ceph_daemon\",\"osd.(.*)\")\n",
+          "expr": "((irate(node_disk_io_time_ms{instance=~\"($ceph_hosts).*\"}[5m]) / 10 ) or  irate(node_disk_io_time_seconds_total{instance=~\"($ceph_hosts).*\"}[5m]) * 100)* on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation{instance=~\"($ceph_hosts).*\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
           "format": "time_series",
           "hide": false,
           "intervalFactor": 1,
-          "legendFormat": "{{device}}({{osd_id}})",
-          "refId": "C",
-          "step": 10,
-          "textEditor": true
-        },
-        {
-          "expr": "(irate(node_disk_io_time_ms{instance=~\"($ceph_hosts).*\"}[5m] * 1000) or irate(node_disk_writes_completed_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
-          "format": "time_series",
-          "intervalFactor": 1,
+          "legendFormat": "{{device}}({{ceph_daemon}})",
           "refId": "A"
         }
       ],
       "thresholds": [],
       "timeFrom": null,
+      "timeRegions": [],
       "timeShift": null,
       "title": "$ceph_hosts Disk utilisation",
       "tooltip": {
@@ -1124,7 +1121,7 @@
       },
       "yaxes": [
         {
-          "format": "short",
+          "format": "percent",
           "label": "%Util",
           "logBase": 1,
           "max": "100",

diff --git a/monitoring/grafana/dashboards/hosts-overview.json b/monitoring/grafana/dashboards/hosts-overview.json
index dbb7bdda74a79e02d6f0e1b6ab48f6bd727441c0..96e739d11fd14913e3294a35b435f6da99c87e5c 100644 (file)
--- a/monitoring/grafana/dashboards/hosts-overview.json
+++ b/monitoring/grafana/dashboards/hosts-overview.json
@@ -431,7 +431,7 @@
       "tableColumn": "",
       "targets": [
         {
-          "expr": "avg (\n  ((irate(node_disk_io_time_ms[5m]) / 10 ) or\n   (irate(node_disk_io_time_seconds_total[5m]) * 100)\n  ) *\n  on(instance, device) label_replace(label_replace(ceph_disk_occupation{instance=~\"($osd_hosts).*\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")\n)",
+          "expr" : "avg (\n  ((irate(node_disk_io_time_ms[5m]) / 10 ) or\n   (irate(node_disk_io_time_seconds_total[5m]) * 100)\n  ) *\n  on(instance, device) label_replace(label_replace(ceph_disk_occupation{instance=~\"($osd_hosts).*\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")\n)",
           "format": "time_series",
           "instant": true,
           "intervalFactor": 1,
@@ -581,11 +581,12 @@
       ],
       "thresholds": [],
       "timeFrom": null,
+      "timeRegions": [],
       "timeShift": null,
       "title": "CPU Busy - Top 10 Hosts",
       "tooltip": {
         "shared": true,
-        "sort": 1,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -667,11 +668,12 @@
       ],
       "thresholds": [],
       "timeFrom": null,
+      "timeRegions": [],
       "timeShift": null,
       "title": "Network Load - Top 10",
       "tooltip": {
         "shared": true,
-        "sort": 1,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",

diff --git a/monitoring/grafana/dashboards/osd-device-details.json b/monitoring/grafana/dashboards/osd-device-details.json
index e78e6a77d6e4590b7a142703e87419f41f24f402..6b70778df4231617ee12037236590ee767897a2e 100644 (file)
--- a/monitoring/grafana/dashboards/osd-device-details.json
+++ b/monitoring/grafana/dashboards/osd-device-details.json
@@ -109,7 +109,7 @@
       "title": "$osd Latency",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -205,7 +205,7 @@
       "title": "$osd R/W IOPS",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -301,7 +301,7 @@
       "title": "$osd R/W Bytes",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -390,14 +390,14 @@
       "steppedLine": false,
       "targets": [
         {
-          "expr": "(irate(node_disk_read_time_seconds_total[1m]) / irate(node_disk_reads_completed_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\"))",
+          "expr": "(irate(node_disk_read_time_seconds_total[1m]) / irate(node_disk_reads_completed_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\"))",
           "format": "time_series",
           "intervalFactor": 1,
           "legendFormat": "{{instance}}/{{device}} Reads",
           "refId": "A"
         },
         {
-          "expr": "(irate(node_disk_write_time_seconds_total[1m]) / irate(node_disk_writes_completed_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\"))",
+          "expr": "(irate(node_disk_write_time_seconds_total[1m]) / irate(node_disk_writes_completed_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\"))",
           "format": "time_series",
           "intervalFactor": 1,
           "legendFormat": "{{instance}}/{{device}} Writes",
@@ -410,7 +410,7 @@
       "title": "Physical Device Latency for $osd",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -486,14 +486,14 @@
       "steppedLine": false,
       "targets": [
         {
-          "expr": "irate(node_disk_writes_completed_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+          "expr": "irate(node_disk_writes_completed_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
           "format": "time_series",
           "intervalFactor": 1,
           "legendFormat": "{{device}} on {{instance}} Writes",
           "refId": "A"
         },
         {
-          "expr": "irate(node_disk_reads_completed_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+          "expr": "irate(node_disk_reads_completed_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
           "format": "time_series",
           "intervalFactor": 1,
           "legendFormat": "{{device}} on {{instance}} Reads",
@@ -506,7 +506,7 @@
       "title": "Physical Device R/W IOPS for $osd",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -582,14 +582,14 @@
       "steppedLine": false,
       "targets": [
         {
-          "expr": "irate(node_disk_read_bytes_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+          "expr": "irate(node_disk_read_bytes_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
           "format": "time_series",
           "intervalFactor": 1,
           "legendFormat": "{{instance}} {{device}} Reads",
           "refId": "A"
         },
         {
-          "expr": "irate(node_disk_written_bytes_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+          "expr": "irate(node_disk_written_bytes_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
           "format": "time_series",
           "intervalFactor": 1,
           "legendFormat": "{{instance}} {{device}} Writes",
@@ -602,7 +602,7 @@
       "title": "Physical Device R/W Bytes for $osd",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -673,7 +673,7 @@
       "steppedLine": false,
       "targets": [
         {
-          "expr": "irate(node_disk_io_time_seconds_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+          "expr": "irate(node_disk_io_time_seconds_total[1m]) and on (instance, device) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"$osd\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
           "format": "time_series",
           "intervalFactor": 1,
           "legendFormat": "{{device}} on {{instance}}",
@@ -686,7 +686,7 @@
       "title": "Physical Device Util% for $osd",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",

diff --git a/monitoring/grafana/dashboards/osds-overview.json b/monitoring/grafana/dashboards/osds-overview.json
index 3fc6d3152c25dca12c8bbd9c636c589135b64b3b..d81c6e3646e1a7ba2374cd56d72f5d9c0a10b652 100644 (file)
--- a/monitoring/grafana/dashboards/osds-overview.json
+++ b/monitoring/grafana/dashboards/osds-overview.json
@@ -112,7 +112,7 @@
       "title": "OSD Read Latencies",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -289,7 +289,7 @@
       "title": "OSD Write Latencies",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -786,7 +786,7 @@
       "title": "Read/Write Profile",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",

diff --git a/monitoring/grafana/dashboards/pool-overview.json b/monitoring/grafana/dashboards/pool-overview.json
index 404d719ead0afa68868978194a17dacd48795fb3..d8654599a9fd593abaa65312005567d87629538b 100644 (file)
--- a/monitoring/grafana/dashboards/pool-overview.json
+++ b/monitoring/grafana/dashboards/pool-overview.json
@@ -82,7 +82,7 @@
       "title": "Top $topk Client IOPS by Pool",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -180,7 +180,7 @@
       "title": "Top $topk Client Throughput by Pool",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -236,7 +236,7 @@
       "scroll": true,
       "showHeader": true,
       "sort": {
-        "col": 6,
+        "col": 5,
         "desc": true
       },
       "styles": [
@@ -382,7 +382,7 @@
       "scroll": true,
       "showHeader": true,
       "sort": {
-        "col": 6,
+        "col": 5,
         "desc": true
       },
       "styles": [
@@ -651,19 +651,19 @@
       },
       {
         "current": {
-          "text": "3",
-          "value": "3"
+          "text": "15",
+          "value": "15"
         },
         "hide": 0,
         "label": "Top K",
         "name": "topk",
         "options": [
           {
-            "text": "3",
-            "value": "3"
+            "text": "15",
+            "value": "15"
           }
         ],
-        "query": "3",
+        "query": "15",
         "skipUrlSync": false,
         "type": "textbox"
       }

diff --git a/monitoring/grafana/dashboards/radosgw-detail.json b/monitoring/grafana/dashboards/radosgw-detail.json
index 8aa02414a8fab56be0243e70b04a12fcd93329ff..648abab89ce962844ac794d7b37d3bfc17b4bff4 100644 (file)
--- a/monitoring/grafana/dashboards/radosgw-detail.json
+++ b/monitoring/grafana/dashboards/radosgw-detail.json
@@ -90,22 +90,23 @@
       "steppedLine": false,
       "targets": [
         {
-          "expr": "rate(ceph_rgw_get_initial_lat_sum{ceph_daemon=~\"($rgw_servers)\"}[30s]) / rate(ceph_rgw_get_initial_lat_count{ceph_daemon=~\"($rgw_servers)\"}[30s])",
+          "expr": "sum by (ceph_daemon) (rate(ceph_rgw_get_initial_lat_sum{ceph_daemon=~\"($rgw_servers)\"}[30s]) / rate(ceph_rgw_get_initial_lat_count{ceph_daemon=~\"($rgw_servers)\"}[30s]))",
           "format": "time_series",
           "intervalFactor": 1,
-          "legendFormat": "GET",
+          "legendFormat": "GET {{ceph_daemon}}",
           "refId": "A"
         },
         {
-          "expr": "rate(ceph_rgw_put_initial_lat_sum{ceph_daemon=~\"($rgw_servers)\"}[30s]) / rate(ceph_rgw_put_initial_lat_count{ceph_daemon=~\"($rgw_servers)\"}[30s])",
+          "expr": "sum by (ceph_daemon)(rate(ceph_rgw_put_initial_lat_sum{ceph_daemon=~\"($rgw_servers)\"}[30s]) / rate(ceph_rgw_put_initial_lat_count{ceph_daemon=~\"($rgw_servers)\"}[30s]))",
           "format": "time_series",
           "intervalFactor": 1,
-          "legendFormat": "PUT",
+          "legendFormat": "PUT {{ceph_daemon}}",
           "refId": "B"
         }
       ],
       "thresholds": [],
       "timeFrom": null,
+      "timeRegions": [],
       "timeShift": null,
       "title": "$rgw_servers GET/PUT Latencies",
       "tooltip": {
@@ -147,6 +148,7 @@
       "dashes": false,
       "datasource": "$datasource",
       "fill": 1,
+      "fillGradient": 0,
       "gridPos": {
         "h": 8,
         "w": 7,
@@ -180,19 +182,20 @@
           "expr": "rate(ceph_rgw_get_b{ceph_daemon=~\"[[rgw_servers]]\"}[30s])",
           "format": "time_series",
           "intervalFactor": 1,
-          "legendFormat": "GETs",
+          "legendFormat": "GETs {{ceph_daemon}}",
           "refId": "B"
         },
         {
           "expr": "rate(ceph_rgw_put_b{ceph_daemon=~\"[[rgw_servers]]\"}[30s])",
           "format": "time_series",
           "intervalFactor": 1,
-          "legendFormat": "PUTs",
+          "legendFormat": "PUTs {{ceph_daemon}}",
           "refId": "A"
         }
       ],
       "thresholds": [],
       "timeFrom": null,
+      "timeRegions": [],
       "timeShift": null,
       "title": "Bandwidth by HTTP Operation",
       "tooltip": {
@@ -241,6 +244,7 @@
       "dashes": false,
       "datasource": "$datasource",
       "fill": 1,
+      "fillGradient": 0,
       "gridPos": {
         "h": 8,
         "w": 7,
@@ -274,33 +278,34 @@
           "expr": "rate(ceph_rgw_failed_req{ceph_daemon=~\"[[rgw_servers]]\"}[30s])",
           "format": "time_series",
           "intervalFactor": 1,
-          "legendFormat": "Requests Failed",
+          "legendFormat": "Requests Failed {{ceph_daemon}}",
           "refId": "B"
         },
         {
           "expr": "rate(ceph_rgw_get{ceph_daemon=~\"[[rgw_servers]]\"}[30s])",
           "format": "time_series",
           "intervalFactor": 1,
-          "legendFormat": "GETs",
+          "legendFormat": "GETs {{ceph_daemon}}",
           "refId": "C"
         },
         {
           "expr": "rate(ceph_rgw_put{ceph_daemon=~\"[[rgw_servers]]\"}[30s])",
           "format": "time_series",
           "intervalFactor": 1,
-          "legendFormat": "PUTs",
+          "legendFormat": "PUTs {{ceph_daemon}}",
           "refId": "D"
         },
         {
           "expr": "rate(ceph_rgw_req{ceph_daemon=~\"[[rgw_servers]]\"}[30s]) -\n  (rate(ceph_rgw_get{ceph_daemon=~\"[[rgw_servers]]\"}[30s]) +\n   rate(ceph_rgw_put{ceph_daemon=~\"[[rgw_servers]]\"}[30s]))",
           "format": "time_series",
           "intervalFactor": 1,
-          "legendFormat": "Other",
+          "legendFormat": "Other {{ceph_daemon}}",
           "refId": "A"
         }
       ],
       "thresholds": [],
       "timeFrom": null,
+      "timeRegions": [],
       "timeShift": null,
       "title": "HTTP Request Breakdown",
       "tooltip": {
@@ -374,28 +379,28 @@
           "expr": "rate(ceph_rgw_failed_req{ceph_daemon=~\"[[rgw_servers]]\"}[30s])",
           "format": "time_series",
           "intervalFactor": 1,
-          "legendFormat": "Failures",
+          "legendFormat": "Failures {{ceph_daemon}}",
           "refId": "A"
         },
         {
           "expr": "rate(ceph_rgw_get{ceph_daemon=~\"[[rgw_servers]]\"}[30s])",
           "format": "time_series",
           "intervalFactor": 1,
-          "legendFormat": "GETs",
+          "legendFormat": "GETs {{ceph_daemon}}",
           "refId": "B"
         },
         {
           "expr": "rate(ceph_rgw_put{ceph_daemon=~\"[[rgw_servers]]\"}[30s])",
           "format": "time_series",
           "intervalFactor": 1,
-          "legendFormat": "PUTs",
+          "legendFormat": "PUTs {{ceph_daemon}}",
           "refId": "C"
         },
         {
           "expr": "rate(ceph_rgw_req{ceph_daemon=~\"[[rgw_servers]]\"}[30s]) -\n  (rate(ceph_rgw_get{ceph_daemon=~\"[[rgw_servers]]\"}[30s]) +\n   rate(ceph_rgw_put{ceph_daemon=~\"[[rgw_servers]]\"}[30s]))",
           "format": "time_series",
           "intervalFactor": 1,
-          "legendFormat": "Other (DELETE,LIST)",
+          "legendFormat": "Other (DELETE,LIST) {{ceph_daemon}}",
           "refId": "D"
         }
       ],

diff --git a/monitoring/grafana/dashboards/rbd-overview.json b/monitoring/grafana/dashboards/rbd-overview.json
index 37eb650a6489fc0584c76a8ff74ab0c224f44f12..f3df003ec49ec62f696df7f0585eec2432a7716c 100644 (file)
--- a/monitoring/grafana/dashboards/rbd-overview.json
+++ b/monitoring/grafana/dashboards/rbd-overview.json
@@ -105,7 +105,7 @@
       "title": "IOPS",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -200,7 +200,7 @@
       "title": "Throughput",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -293,7 +293,7 @@
       "title": "Average Latency",
       "tooltip": {
         "shared": true,
-        "sort": 0,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",