FAU config

fau-systems/cc-metric-collector/alex/collectors.json

@@ -0,0 +1,145 @@
+{
+  "nfs4stat" : {},
+  "memstat" : {
+    "numa_stats": true,
+    "node_stats": true
+  },
+  "cpustat" : {},
+  "loadavg" : {},
+  "schedstat": {},
+  "nvidia" : {
+    "use_pci_info_as_type_id": true
+  },
+  "lustrestat" : {
+    "send_all_metrics" : true,
+    "use_sudo": false,
+    "send_diff_values": true,
+    "send_derived_values": true,
+    "send_abs_values": false
+  },
+  "netstat" : {
+    "include_devices" : [
+      "enp1s0",
+      "enp70s0f0",
+      "enp195s0f0"
+    ],
+    "send_abs_values": true,
+    "send_derived_values": true
+  },
+  "diskstat" : {},
+  "iostat" : {},
+  "ibstat" : {
+    "send_abs_values": true,
+    "send_derived_values": true
+  },
+  "ipmistat" : {
+    "send_abs_values": true,
+    "send_derived_values": true
+  },
+  "tempstat" : {
+    "tag_override" : {
+      "hwmon0" : {
+        "type" : "socket",
+        "type-id" : "0"
+      },
+      "hwmon1" : {
+        "type" : "socket",
+        "type-id" : "1"
+      }
+    }
+  },
+  "likwid": {
+    "force_overwrite" : true,
+    "invalid_to_zero" : true,
+    "access_mode" : "accessdaemon",
+    "accessdaemon_path" : "/apps/likwid/system/sbin",
+    "liblikwid_path": "/apps/likwid/system/lib/liblikwid.so",
+    "eventsets": [
+      {
+        "events": {
+          "FIXC1": "ACTUAL_CPU_CLOCK",
+          "FIXC2": "MAX_CPU_CLOCK",
+          "PMC0": "RETIRED_INSTRUCTIONS",
+          "PMC1": "CPU_CLOCKS_UNHALTED",
+          "PMC2": "RETIRED_SSE_AVX_FLOPS_ALL",
+          "PMC3": "MERGE",
+          "DFC0": "DRAM_CHANNEL_0",
+          "DFC1": "DRAM_CHANNEL_1",
+          "DFC2": "DRAM_CHANNEL_2",
+          "DFC3": "DRAM_CHANNEL_3"
+        },
+        "metrics": [
+          {
+            "name": "ipc",
+            "calc": "PMC0/PMC1",
+            "type": "hwthread",
+            "publish": true
+          },
+          {
+            "name": "flops_any",
+            "calc": "1E-9*PMC2/time",
+            "unit": "GFlops/s",
+            "type": "hwthread",
+            "publish": true
+          },
+          {
+            "name": "clock",
+            "calc": "1E-6*(FIXC1/FIXC2)/inverseClock",
+            "type": "hwthread",
+            "unit": "MHz",
+            "publish": true
+          },
+          {
+            "name": "mem1",
+            "calc": "1E-9*(DFC0+DFC1+DFC2+DFC3)*64.0/time",
+            "unit": "Gbyte/s",
+            "type": "socket",
+            "publish": false
+          }
+        ]
+      },
+      {
+        "events": {
+          "DFC0": "DRAM_CHANNEL_4",
+          "DFC1": "DRAM_CHANNEL_5",
+          "DFC2": "DRAM_CHANNEL_6",
+          "DFC3": "DRAM_CHANNEL_7",
+          "PWR0": "RAPL_CORE_ENERGY",
+          "PWR1": "RAPL_PKG_ENERGY"
+        },
+        "metrics": [
+          {
+            "name": "core_power",
+            "calc": "PWR0/time",
+            "unit": "Watt",
+            "type": "hwthread",
+            "publish": true
+          },
+          {
+            "name": "cpu_power",
+            "calc": "PWR1/time",
+            "type": "socket",
+            "unit": "Watt",
+            "publish": true
+          },
+          {
+            "name": "mem2",
+            "calc": "1E-9*(DFC0+DFC1+DFC2+DFC3)*64.0/time",
+            "unit": "Gbyte/s",
+            "type": "socket",
+            "publish": false
+          }
+        ]
+      }
+    ],
+    "globalmetrics": [
+      {
+        "name": "mem_bw",
+        "calc": "mem1+mem2",
+        "type": "socket",
+        "unit": "Gbyte/s",
+        "publish": true
+      }
+    ]
+  }
+}

fau-systems/cc-metric-collector/alex/config.json

@@ -0,0 +1,8 @@
+{
+  "sinks": "/etc/cc-metric-collector/sinks.json",
+  "collectors" : "/etc/cc-metric-collector/collectors.json",
+  "receivers" : "/etc/cc-metric-collector/receivers.json",
+  "router" : "/etc/cc-metric-collector/router.json",
+  "interval": "60s",
+  "duration": "10s"
+}

fau-systems/cc-metric-collector/alex/receivers.json

@@ -0,0 +1 @@
+{}

fau-systems/cc-metric-collector/alex/router.json

@@ -0,0 +1,58 @@
+{
+    "add_tags" : [
+        {
+            "key" : "cluster",
+            "value" : "alex",
+            "if" : "*"
+        }
+    ],
+    "rename_metrics" : {
+        "load_one" : "cpu_load",
+        "cpu_load_core" : "cpu_load",
+        "net_bytes_in_bw" : "net_bytes_in",
+        "net_bytes_out_bw" : "net_bytes_out",
+        "net_pkts_in_bw" : "net_pkts_in",
+        "net_pkts_out_bw" : "net_pkts_out",
+        "ib_recv_bw" : "ib_recv",
+        "ib_xmit_bw" : "ib_xmit",
+        "ib_recv_pkts_bw": "ib_recv_pkts",
+        "ib_xmit_pkts_bw": "ib_xmit_pkts",
+        "lustre_read_bytes_diff" : "lustre_read_bytes",
+        "lustre_read_requests_diff" : "lustre_read_requests",
+        "lustre_write_bytes_diff" : "lustre_write_bytes",
+        "lustre_write_requests_diff" : "lustre_write_requests",
+        "lustre_open_diff" : "lustre_open",
+        "lustre_close_diff" : "lustre_close",
+        "lustre_setattr_diff" : "lustre_setattr",
+        "lustre_getattr_diff" : "lustre_getattr",
+        "lustre_statfs_diff": "lustre_statfs",
+        "lustre_inode_permission_diff" : "lustre_inode_permission",
+        "nv_util" : "acc_utilization",
+        "nv_fb_mem_used" : "acc_mem_used",
+        "nv_power_usage" : "acc_power"
+    },
+    "drop_metrics" : [
+        "net_bytes_in",
+        "net_bytes_out",
+        "ib_recv",
+        "ib_xmit",
+        "ib_recv_pkts",
+        "ib_xmit_pkts",
+        "net_pkts_in",
+        "net_pkts_out",
+        "lustre_read_bytes",
+        "lustre_read_requests",
+        "lustre_write_bytes",
+        "lustre_write_requests"
+    ],
+    "interval_timestamp" : false,
+    "num_cache_intervals" : 0,
+    "change_unit_prefix": {
+        "mem_used": "G",
+        "swap_used": "G",
+        "mem_total": "G",
+        "swap_total": "G",
+        "cpufreq": "M"
+    },
+    "normalize_metrics" : true
+}

fau-systems/cc-metric-collector/alex/sinks.json

@@ -0,0 +1,26 @@
+{
+    "nhrinflux": {
+        "type": "influxasync",
+        "host": "monitoring-test.nhr.uni-erlangen.de",
+        "port": "8086",
+        "organization": "ClusterCockpit",
+        "database": "alex",
+        "password": "XYZ",
+        "ssl": true,
+        "meta_as_tags": [
+            "unit"
+        ]
+    },
+    "alexstore": {
+        "type": "http",
+        "url": "http://monitoring.nhr.fau.de:8082/api/write?cluster=alex",
+        "jwt": "XYZ",
+        "meta_as_tags": [
+            "unit"
+        ],
+        "idle_connection_timeout": "60s",
+        "flush_delay": "2s",
+        "max_retries": 1,
+        "timeout": "10s"
+    }
+}

fau-systems/cc-metric-collector/alex/sinks_debug.json

@@ -0,0 +1,33 @@
+{
+    "fritzganglia": {
+        "type": "libganglia",
+        "gmond_config": "/etc/ganglia/gmond.conf",
+        "libganglia_path": "libganglia.so.0",
+        "add_ganglia_group": true
+    },
+    "nhrinflux": {
+        "type": "influxasync",
+        "host": "monitoring-test.nhr.uni-erlangen.de",
+        "port": "8086",
+        "organization": "ClusterCockpit",
+        "database": "fritz_neu",
+        "password": "XZY",
+        "ssl": true,
+        "meta_as_tags": [
+            "unit"
+        ]
+    },
+    "fritzstore": {
+        "type": "http",
+        "url": "http://monitoring.nhr.fau.de:8082/api/write?cluster=fritz",
+        "jwt": "XZY",
+        "meta_as_tags": [
+            "unit"
+        ],
+        "idle_connection_timeout": "60s"
+    },
+    "debugstdout": {
+        "type": "stdout",
+        "output_file": "/tmp/debug.log"
+    }
+}