通过vSphere API获取对象Statistics
预备知识点：
1、vim.PerformanceManager.MetricId() 通过counter_id获取到metric_id
2、vim.PerformanceManager.QuerySpec() 通过对象、metric_id、起始时间进行装配
3、content.perfManager.QueryPerf() 通过QuerySpec获取对象的Statistics
4、content.perfManager.perfCounter 获取所有Performance Counter相关信息
示例1： 获取ESXI主机的网络利用率
from pyVim.connect import SmartConnectNoSSL, Disconnect from pyVmomi import vim from datetime import timedelta import atexit def connect_vc(host, user, pwd, port): si = SmartConnectNoSSL(host=host, user=user, pwd=pwd, port=port) # disconnect this thing atexit.register(Disconnect, si) return si def build_query(content, vc_time, counter_id, obj, interval): metric_id = vim.PerformanceManager.MetricId(counterId=counter_id, instance="") start_time = vc_time - timedelta(minutes=(interval + 1)) end_time = vc_time - timedelta(minutes=1) query = vim.PerformanceManager.QuerySpec(intervalId=20, entity=obj, metricId=[metric_id], startTime=start_time, endTime=end_time) perf_results = content.perfManager.QueryPerf(querySpec=[query]) if perf_results: return perf_results else: pass def print_statistics(obj, content, vc_time, interval, perf_dict, ): stat_interval = interval * 3 # There are 3per 20s samples in each minute # Network usage (Tx/Rx) # statNetworkTx = BuildQuery(content, vchtime, (stat_check(perf_dict, 'net.usage.maximum')), obj, interval) # networkTx = (float(sum(statNetworkTx[0].value[0].value) * 8 / 1024) / statInt) # statNetworkRx = BuildQuery(content, vchtime, (stat_check(perf_dict, 'net.usage.minimum')), obj, interval) # networkRx = (float(sum(statNetworkRx[0].value[0].value) * 8 / 1024) / statInt) # Network utilization (combined transmit-rates and receive-rates) during the interval = 145 network_usage = build_query(content, vc_time, stat_check(perf_dict, 'net.usage.average'), obj, interval) try: print('statNetworkThroughput:%sMB' % (round((((sum(network_usage[0].value[0].value)) / 1024) / stat_interval), 2))) except TypeError: # 关机的ESXi主机无法获取到数据 pass def stat_check(perf_dict, counter_name): """通过performance counter名称获取counter id""" counter_id = perf_dict[counter_name] return counter_id def main(): username = 'administrator@vsphere.local' password = 'xxxxxx' vc_ip = '172.16.65.99' vc_port = '443' statistics_interval_time = 10 # 分钟为单位 si = connect_vc(host=vc_ip, user=username, pwd=password, port=vc_port) content = si.RetrieveContent() # Get vCenter date and time for use as baseline when querying for counters vc_time = si.CurrentTime() # 获取所有performance counter，并放入字典中 perf_dict = {} perf_list = content.perfManager.perfCounter for counter in perf_list: counter_full = "{}.{}.{}".format(counter.groupInfo.key, counter.nameInfo.key, counter.rollupType) perf_dict[counter_full] = counter.key # perf_dict包含了所有的perfCounter # 获取ESXi主机对象 container_view = content.viewManager.CreateContainerView(content.rootFolder, [vim.HostSystem], True) for obj in container_view.view: print_statistics(obj, content, vc_time, statistics_interval_time, perf_dict) # Start program if __name__ == "__main__": main()
示例2： 获取VM相关的performance
""" Python program that generates various statistics for one or more virtual machines A list of virtual machines can be provided as a comma separated list. """ from __future__ import print_function from pyVim.connect import SmartConnectNoSSL, Disconnect from pyVmomi import vmodl, vim from datetime import timedelta, datetime import atexit def connect_vc(host, user, pwd, port): si = SmartConnectNoSSL(host=host, user=user, pwd=pwd, port=port) # disconnect this thing atexit.register(Disconnect, si) return si def BuildQuery(content, vchtime, counterId, instance, vm, interval): perfManager = content.perfManager metricId = vim.PerformanceManager.MetricId(counterId=counterId, instance=instance) startTime = vchtime - timedelta(minutes=(interval + 1)) endTime = vchtime - timedelta(minutes=1) query = vim.PerformanceManager.QuerySpec(intervalId=20, entity=vm, metricId=[metricId], startTime=startTime, endTime=endTime) perfResults = perfManager.QueryPerf(querySpec=[query]) if perfResults: return perfResults else: print('ERROR: Performance results empty. TIP: Check time drift on source and vCenter server') print('Troubleshooting info:') print('vCenter/host date and time: {}'.format(vchtime)) print('Start perf counter time : {}'.format(startTime)) print('End perf counter time : {}'.format(endTime)) print(query) exit() def PrintVmInfo(vm, content, vchtime, interval, perf_dict, ): statInt = interval * 3 # There are 3 20s samples in each minute summary = vm.summary disk_list = [] network_list = [] # Convert limit and reservation values from -1 to None if vm.resourceConfig.cpuAllocation.limit == -1: vmcpulimit = "None" else: vmcpulimit = "{} Mhz".format(vm.resourceConfig.cpuAllocation.limit) if vm.resourceConfig.memoryAllocation.limit == -1: vmmemlimit = "None" else: vmmemlimit = "{} MB".format(vm.resourceConfig.cpuAllocation.limit) if vm.resourceConfig.cpuAllocation.reservation == 0: vmcpures = "None" else: vmcpures = "{} Mhz".format(vm.resourceConfig.cpuAllocation.reservation) if vm.resourceConfig.memoryAllocation.reservation == 0: vmmemres = "None" else: vmmemres = "{} MB".format(vm.resourceConfig.memoryAllocation.reservation) vm_hardware = vm.config.hardware for each_vm_hardware in vm_hardware.device: if (each_vm_hardware.key >= 2000) and (each_vm_hardware.key < 3000): disk_list.append('{} | {:.1f}GB | Thin: {} | {}'.format(each_vm_hardware.deviceInfo.label, each_vm_hardware.capacityInKB/1024/1024, each_vm_hardware.backing.thinProvisioned, each_vm_hardware.backing.fileName)) elif (each_vm_hardware.key >= 4000) and (each_vm_hardware.key < 5000): network_list.append('{} | {} | {}'.format(each_vm_hardware.deviceInfo.label, each_vm_hardware.deviceInfo.summary, each_vm_hardware.macAddress)) #CPU Ready Average statCpuReady = BuildQuery(content, vchtime, (StatCheck(perf_dict, 'cpu.ready.summation')), "", vm, interval) cpuReady = (float(sum(statCpuReady[0].value[0].value)) / statInt) #CPU Usage Average % - NOTE: values are type LONG so needs divided by 100 for percentage statCpuUsage = BuildQuery(content, vchtime, (StatCheck(perf_dict, 'cpu.usage.average')), "", vm, interval) cpuUsage = ((float(sum(statCpuUsage[0].value[0].value)) / statInt) / 100) #Memory Active Average MB statMemoryActive = BuildQuery(content, vchtime, (StatCheck(perf_dict, 'mem.active.average')), "", vm, interval) memoryActive = (float(sum(statMemoryActive[0].value[0].value) / 1024) / statInt) #Memory Shared statMemoryShared = BuildQuery(content, vchtime, (StatCheck(perf_dict, 'mem.shared.average')), "", vm, interval) memoryShared = (float(sum(statMemoryShared[0].value[0].value) / 1024) / statInt) #Memory Balloon statMemoryBalloon = BuildQuery(content, vchtime, (StatCheck(perf_dict, 'mem.vmmemctl.average')), "", vm, interval) memoryBalloon = (float(sum(statMemoryBalloon[0].value[0].value) / 1024) / statInt) #Memory Swapped statMemorySwapped = BuildQuery(content, vchtime, (StatCheck(perf_dict, 'mem.swapped.average')), "", vm, interval) memorySwapped = (float(sum(statMemorySwapped[0].value[0].value) / 1024) / statInt) #Datastore Average IO statDatastoreIoRead = BuildQuery(content, vchtime, (StatCheck(perf_dict, 'datastore.numberReadAveraged.average')), "*", vm, interval) DatastoreIoRead = (float(sum(statDatastoreIoRead[0].value[0].value)) / statInt) statDatastoreIoWrite = BuildQuery(content, vchtime, (StatCheck(perf_dict, 'datastore.numberWriteAveraged.average')), "*", vm, interval) DatastoreIoWrite = (float(sum(statDatastoreIoWrite[0].value[0].value)) / statInt) #Datastore Average Latency statDatastoreLatRead = BuildQuery(content, vchtime, (StatCheck(perf_dict, 'datastore.totalReadLatency.average')), "*", vm, interval) DatastoreLatRead = (float(sum(statDatastoreLatRead[0].value[0].value)) / statInt) statDatastoreLatWrite = BuildQuery(content, vchtime, (StatCheck(perf_dict, 'datastore.totalWriteLatency.average')), "*", vm, interval) DatastoreLatWrite = (float(sum(statDatastoreLatWrite[0].value[0].value)) / statInt) #Network usage (Tx/Rx) statNetworkTx = BuildQuery(content, vchtime, (StatCheck(perf_dict, 'net.transmitted.average')), "", vm, interval) networkTx = (float(sum(statNetworkTx[0].value[0].value) * 8 / 1024) / statInt) statNetworkRx = BuildQuery(content, vchtime, (StatCheck(perf_dict, 'net.received.average')), "", vm, interval) networkRx = (float(sum(statNetworkRx[0].value[0].value) * 8 / 1024) / statInt) print('networkRx:', networkRx) print('networkTx:', networkTx) print('\nNOTE: Any VM statistics are averages of the last {} minutes\n'.format(statInt / 3)) print('Server Name :', summary.config.name) print('Description :', summary.config.annotation) print('Guest :', summary.config.guestFullName) if vm.rootSnapshot: print('Snapshot Status : Snapshots present') else: print('Snapshot Status : No Snapshots') print('VM .vmx Path :', summary.config.vmPathName) try: print('Virtual Disks :', disk_list[0]) if len(disk_list) > 1: disk_list.pop(0) for each_disk in disk_list: print(' ', each_disk) except IndexError: pass print('Virtual NIC(s) :', network_list[0]) if len(network_list) > 1: network_list.pop(0) for each_vnic in network_list: print(' ', each_vnic) print('[VM] Limits : CPU: {}, Memory: {}'.format(vmcpulimit, vmmemlimit)) print('[VM] Reservations : CPU: {}, Memory: {}'.format(vmcpures, vmmemres)) print('[VM] Number of vCPUs :', summary.config.numCpu) print('[VM] CPU Ready : Average {:.1f} %, Maximum {:.1f} %'.format((cpuReady / 20000 * 100), ((float(max( statCpuReady[0].value[ 0].value)) / 20000 * 100)))) print('[VM] CPU (%) : {:.0f} %'.format(cpuUsage)) print('[VM] Memory : {} MB ({:.1f} GB)'.format(summary.config.memorySizeMB, (float(summary.config.memorySizeMB) / 1024))) print('[VM] Memory Shared : {:.0f} %, {:.0f} MB'.format( ((memoryShared / summary.config.memorySizeMB) * 100), memoryShared)) print('[VM] Memory Balloon : {:.0f} %, {:.0f} MB'.format( ((memoryBalloon / summary.config.memorySizeMB) * 100), memoryBalloon)) print('[VM] Memory Swapped : {:.0f} %, {:.0f} MB'.format( ((memorySwapped / summary.config.memorySizeMB) * 100), memorySwapped)) print('[VM] Memory Active : {:.0f} %, {:.0f} MB'.format( ((memoryActive / summary.config.memorySizeMB) * 100), memoryActive)) print('[VM] Datastore Average IO : Read: {:.0f} IOPS, Write: {:.0f} IOPS'.format(DatastoreIoRead, DatastoreIoWrite)) print('[VM] Datastore Average Latency : Read: {:.0f} ms, Write: {:.0f} ms'.format(DatastoreLatRead, DatastoreLatWrite)) print('[VM] Overall Network Usage : Transmitted {:.3f} Mbps, Received {:.3f} Mbps'.format(networkTx, networkRx)) print('[Host] Name : {}'.format(summary.runtime.host.name)) print('[Host] CPU Detail : Processor Sockets: {}, Cores per Socket {}'.format( summary.runtime.host.summary.hardware.numCpuPkgs, (summary.runtime.host.summary.hardware.numCpuCores / summary.runtime.host.summary.hardware.numCpuPkgs))) print('[Host] CPU Type : {}'.format(summary.runtime.host.summary.hardware.cpuModel)) print('[Host] CPU Usage : Used: {} Mhz, Total: {} Mhz'.format( summary.runtime.host.summary.quickStats.overallCpuUsage, (summary.runtime.host.summary.hardware.cpuMhz * summary.runtime.host.summary.hardware.numCpuCores))) print('[Host] Memory Usage : Used: {:.0f} GB, Total: {:.0f} GB\n'.format( (float(summary.runtime.host.summary.quickStats.overallMemoryUsage) / 1024), (float(summary.runtime.host.summary.hardware.memorySize) / 1024 / 1024 / 1024))) def StatCheck(perf_dict, counter_name): counter_key = perf_dict[counter_name] return counter_key def GetProperties(content, viewType, props, specType): # Build a view and get basic properties for all Virtual Machines objView = content.viewManager.CreateContainerView(content.rootFolder, viewType, True) tSpec = vim.PropertyCollector.TraversalSpec(name='tSpecName', path='view', skip=False, type=vim.view.ContainerView) pSpec = vim.PropertyCollector.PropertySpec(all=False, pathSet=props, type=specType) oSpec = vim.PropertyCollector.ObjectSpec(obj=objView, selectSet=[tSpec], skip=False) pfSpec = vim.PropertyCollector.FilterSpec(objectSet=[oSpec], propSet=[pSpec], reportMissingObjectsInResults=False) retOptions = vim.PropertyCollector.RetrieveOptions() totalProps = [] retProps = content.propertyCollector.RetrievePropertiesEx(specSet=[pfSpec], options=retOptions) totalProps += retProps.objects while retProps.token: retProps = content.propertyCollector.ContinueRetrievePropertiesEx(token=retProps.token) totalProps += retProps.objects objView.Destroy() # Turn the output in retProps into a usable dictionary of values gpOutput = [] for eachProp in totalProps: propDic = {} for prop in eachProp.propSet: propDic[prop.name] = prop.val propDic['moref'] = eachProp.obj gpOutput.append(propDic) return gpOutput def main(): username = 'administrator@vsphere.local' password = 'xxxxxx' vc_ip = '172.16.65.99' vc_port = '443' customization_spec_name = 'Ubuntu_Customization' si = connect_vc(host=vc_ip, user=username, pwd=password, port=vc_port) content = si.RetrieveContent() # Get vCenter date and time for use as baseline when querying for counters vchtime = si.CurrentTime() # Get all the performance counters perf_dict = {} perfList = content.perfManager.perfCounter for counter in perfList: counter_full = "{}.{}.{}".format(counter.groupInfo.key, counter.nameInfo.key, counter.rollupType) perf_dict[counter_full] = counter.key retProps = GetProperties(content, [vim.VirtualMachine], ['name', 'runtime.powerState'], vim.VirtualMachine) #Find VM supplied as arg and use Managed Object Reference (moref) for the PrintVmInfo for vm in retProps: PrintVmInfo(vm['moref'], content, vchtime, 20, perf_dict) break # Start program if __name__ == "__main__": main()
