X-Git-Url: https://git.proxmox.com/?a=blobdiff_plain;f=ceph%2Fmonitoring%2Fgrafana%2Fdashboards%2Fhosts-overview.json;h=804aa51cc9df3defb9a3fe967f38e832ab9f32de;hb=f6b5b4d738b87d88d2de35127b6b0e41eae2a272;hp=dbb7bdda74a79e02d6f0e1b6ab48f6bd727441c0;hpb=81eedcae9c6d684efdae63dc481d8be7c714897d;p=ceph.git

diff --git a/ceph/monitoring/grafana/dashboards/hosts-overview.json b/ceph/monitoring/grafana/dashboards/hosts-overview.json
index dbb7bdda7..804aa51cc 100644
--- a/ceph/monitoring/grafana/dashboards/hosts-overview.json
+++ b/ceph/monitoring/grafana/dashboards/hosts-overview.json
@@ -133,6 +133,7 @@
       "datasource": "$datasource",
       "decimals": 0,
       "description": "Average CPU busy across all hosts (OSD, RGW, MON etc) within the cluster",
+      "decimals": 2,
       "format": "percentunit",
       "gauge": {
         "maxValue": 100,
@@ -216,6 +217,7 @@
       "datasource": "$datasource",
       "decimals": 0,
       "description": "Average Memory Usage across all hosts in the cluster (excludes buffer/cache usage)",
+      "decimals": 2,
       "format": "percentunit",
       "gauge": {
         "maxValue": 100,
@@ -267,7 +269,7 @@
       "tableColumn": "",
       "targets": [
         {
-          "expr": "avg (((node_memory_MemTotal{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_MemTotal_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"})- (\n  (node_memory_MemFree{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_MemFree_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"})  + \n  (node_memory_Cached{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_Cached_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}) + \n  (node_memory_Buffers{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_Buffers_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}) +\n  (node_memory_Slab{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_Slab_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"})\n  )) /\n (node_memory_MemTotal{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_MemTotal_bytes{instance=~\"[[osd_hosts]]|[[rgw_hosts]]|[[mon_hosts]]|[[mds_hosts]].*\"} ))",
+          "expr": "avg (((node_memory_MemTotal{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_MemTotal_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"})- (\n  (node_memory_MemFree{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_MemFree_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"})  + \n  (node_memory_Cached{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_Cached_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}) + \n  (node_memory_Buffers{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_Buffers_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}) +\n  (node_memory_Slab{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_Slab_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"})\n  )) /\n (node_memory_MemTotal{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or node_memory_MemTotal_bytes{instance=~\"($osd_hosts|$rgw_hosts|$mon_hosts|$mds_hosts).*\"} ))",
           "format": "time_series",
           "instant": true,
           "intervalFactor": 1,
@@ -431,7 +433,7 @@
       "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  label_replace((irate(node_disk_io_time_ms[5m]) / 10 ) or\n   (irate(node_disk_io_time_seconds_total[5m]) * 100), \"instance\", \"$1\", \"instance\", \"([^.:]*).*\"\n  ) *\n  on(instance, device) label_replace(label_replace(ceph_disk_occupation{instance=~\"($osd_hosts).*\"}, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^.:]*).*\")\n)",
           "format": "time_series",
           "instant": true,
           "intervalFactor": 1,
@@ -572,7 +574,7 @@
       "steppedLine": false,
       "targets": [
         {
-          "expr": "topk(10,( 1 - (\n    avg by(instance) \n      (irate(node_cpu_seconds_total{mode='idle',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]) or\n       irate(node_cpu{mode='idle',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]))\n    )\n  )\n)",
+          "expr": "topk(10,100 * ( 1 - (\n    avg by(instance) \n      (irate(node_cpu_seconds_total{mode='idle',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]) or\n       irate(node_cpu{mode='idle',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[1m]))\n    )\n  )\n)",
           "format": "time_series",
           "intervalFactor": 1,
           "legendFormat": "{{instance}}",
@@ -581,11 +583,12 @@
       ],
       "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 +670,12 @@
       ],
       "thresholds": [],
       "timeFrom": null,
+      "timeRegions": [],
       "timeShift": null,
-      "title": "Network Load - Top 10",
+      "title": "Network Load - Top 10 Hosts",
       "tooltip": {
         "shared": true,
-        "sort": 1,
+        "sort": 2,
         "value_type": "individual"
       },
       "type": "graph",
@@ -738,7 +742,7 @@
         "multi": false,
         "name": "osd_hosts",
         "options": [],
-        "query": "label_values(ceph_disk_occupation, instance)",
+        "query": "label_values(ceph_disk_occupation, exported_instance)",
         "refresh": 1,
         "regex": "([^.]*).*",
         "skipUrlSync": false,