Merge branch 'master' of https://github.com/ganglia/gmond_python_modules

ganglia · Jun 20, 2012 · a1f9d33 · a1f9d33
2 parents 2c8459d + 9303fdc
commit a1f9d33
Show file tree

Hide file tree

Showing 13 changed files with 2,644 additions and 8 deletions.
diff --git a/gpu/nvidia/README b/gpu/nvidia/README
@@ -26,11 +26,14 @@ The following metrics have been implemented:
 * gpu_graphics_speed
 * gpu_sm_speed
 * gpu_mem_speed
+* gpu_max_graphics_speed
+* gpu_max_sm_speed
+* gpu_max_mem_speed
 * gpu_temp
 * gpu_util
 * gpu_mem_util
 * gpu_mem_used
 * gpu_fan
 * gpu_power_usage
-* gpu_power_state
+* gpu_perf_state
 * gpu_ecc_mode
diff --git a/gpu/nvidia/conf.d/nvidia.pyconf b/gpu/nvidia/conf.d/nvidia.pyconf
@@ -73,9 +73,9 @@ collection_group {
   }
 
   metric {
-    name_match = "([\\S]+)_power_state"
-    name = "\\1_power_state"
-    title= "\\1 Power State"
+    name_match = "([\\S]+)_perf_state"
+    name = "\\1_perf_state"
+    title= "\\1 Performance State"
     value_threshold = 1.0
   }
 
@@ -124,4 +124,25 @@ collection_group {
     name = "\\1_mem_total"
     title = "\\1 Memory Total"
   }
+
+  metric {
+    name_match = "([\\S]+)_max_graphics_speed"
+    name = "\\1_max_graphics_speed"
+    title = "\\1 Max Graphics Speed"
+    value_threshold = 1.0
+  }
+
+  metric {
+    name_match = "([\\S]+)_max_sm_speed"
+    name = "\\1_max_sm_speed"
+    title = "\\1 Max SM Speed"
+    value_threshold = 1.0
+  }
+
+  metric {
+    name_match = "([\\S]+)_max_mem_speed"
+    name = "\\1_max_mem_speed"
+    title = "\\1 Max Memory Speed"
+    value_threshold = 1.0
+  }
 }
diff --git a/gpu/nvidia/python_modules/nvidia.py b/gpu/nvidia/python_modules/nvidia.py
@@ -1,6 +1,6 @@
 # NVIDIA GPU metric module using the Python bindings for NVML
 #
-# (C)opyright 2011 Bernard Li <bernard@vanhpc.org>
+# (C)opyright 2011, 2012 Bernard Li <bernard@vanhpc.org>
 # All Rights Reserved.
 #
 # Redistribution and use in source and binary forms, with or without
@@ -99,8 +99,8 @@ def gpu_device_handler(name):
         except NVMLError, nvmlError:
             if NVML_ERROR_NOT_SUPPORTED == nvmlError.value:
                 return 'N/A'
-    elif (metric == 'power_state'):
-        state = nvmlDeviceGetPowerState(gpu_device)
+    elif (metric == 'perf_state'):
+        state = nvmlDeviceGetPerformanceState(gpu_device)
         try:
             int(state)
             return "P%s" % state
@@ -114,6 +114,12 @@ def gpu_device_handler(name):
         return nvmlDeviceGetClockInfo(gpu_device, NVML_CLOCK_MEM)
     elif (metric == 'power_usage'):
         return nvmlDeviceGetPowerUsage(gpu_device)
+    elif (metric == 'max_graphics_speed'):
+        return nvmlDeviceGetMaxClockInfo(gpu_device, NVML_CLOCK_GRAPHICS)
+    elif (metric == 'max_sm_speed'):
+        return nvmlDeviceGetMaxClockInfo(gpu_device, NVML_CLOCK_SM)
+    elif (metric == 'max_mem_speed'):
+        return nvmlDeviceGetMaxClockInfo(gpu_device, NVML_CLOCK_MEM)
     else:
         print "Handler for %s not implemented, please fix in gpu_device_handler()" % metric
         os._exit(1)
@@ -138,13 +144,16 @@ def metric_init(params):
         build_descriptor('gpu%s_graphics_speed' % i, gpu_device_handler, default_time_max, 'uint', 'MHz', 'both', '%u', 'GPU%s Graphics Speed' % i, 'gpu')
         build_descriptor('gpu%s_sm_speed' % i, gpu_device_handler, default_time_max, 'uint', 'MHz', 'both', '%u', 'GPU%s SM Speed' % i, 'gpu')
         build_descriptor('gpu%s_mem_speed' % i, gpu_device_handler, default_time_max, 'uint', 'MHz', 'both', '%u', 'GPU%s Memory Speed' % i, 'gpu')
+        build_descriptor('gpu%s_max_graphics_speed' % i, gpu_device_handler, default_time_max, 'uint', 'MHz', 'zero', '%u', 'GPU%s Max Graphics Speed' % i, 'gpu')
+        build_descriptor('gpu%s_max_sm_speed' % i, gpu_device_handler, default_time_max, 'uint', 'MHz', 'zero', '%u', 'GPU%s Max SM Speed' % i, 'gpu')
+        build_descriptor('gpu%s_max_mem_speed' % i, gpu_device_handler, default_time_max, 'uint', 'MHz', 'zero', '%u', 'GPU%s Max Memory Speed' % i, 'gpu')
         build_descriptor('gpu%s_uuid' % i, gpu_device_handler, default_time_max, 'string', '', 'zero', '%s', 'GPU%s UUID' % i, 'gpu')
         build_descriptor('gpu%s_pci_id' % i, gpu_device_handler, default_time_max, 'string', '', 'zero', '%s', 'GPU%s PCI ID' % i, 'gpu')
         build_descriptor('gpu%s_temp' % i, gpu_device_handler, default_time_max, 'uint', 'C', 'both', '%u', 'Temperature of GPU %s' % i, 'gpu,temp')
         build_descriptor('gpu%s_mem_total' % i, gpu_device_handler, default_time_max, 'uint', 'KB', 'zero', '%u', 'GPU%s Total Memory' %i, 'gpu')
         build_descriptor('gpu%s_mem_used' % i, gpu_device_handler, default_time_max, 'uint', 'KB', 'both', '%u', 'GPU%s Used Memory' %i, 'gpu')
         build_descriptor('gpu%s_ecc_mode' % i, gpu_device_handler, default_time_max, 'string', '', 'zero', '%s', 'GPU%s ECC Mode' %i, 'gpu')
-        build_descriptor('gpu%s_power_state' % i, gpu_device_handler, default_time_max, 'string', '', 'zero', '%s', 'GPU%s Power State' %i, 'gpu')
+        build_descriptor('gpu%s_perf_state' % i, gpu_device_handler, default_time_max, 'string', '', 'zero', '%s', 'GPU%s Performance State' %i, 'gpu')
         build_descriptor('gpu%s_util' % i, gpu_device_handler, default_time_max, 'uint', '%', 'both', '%u', 'GPU%s Utilization' %i, 'gpu')
         build_descriptor('gpu%s_mem_util' % i, gpu_device_handler, default_time_max, 'uint', '%', 'both', '%u', 'GPU%s Memory Utilization' %i, 'gpu')
         build_descriptor('gpu%s_fan' % i, gpu_device_handler, default_time_max, 'uint', '%', 'both', '%u', 'GPU%s Fan Speed' %i, 'gpu')