Introduction
Good day. For half a year now, we have been using a script (more accurately, a set of scripts) that generates reports on the status of virtual machines (and more). I decided to share my experience in creating it along with the code. I welcome constructive criticism and hope this material might be useful to someone.
Formation of the need
We have many virtual machines (around 1500 VMs distributed across 3 vCenters). New ones are created and old ones are deleted quite often. To maintain order, several custom fields were added to vCenter to categorize VMs by subsystems, indicate whether they are for testing, and specify who created them and when. The human factor led to more than half of the machines having incomplete fields, complicating work. Every six months, someone would get frustrated and initiate a project to update this data, but the results would become outdated within a week and a half.
I want to clarify that everyone understands that there should be requests for creating machines, a process for their creation, and so on. However, unfortunately, that's not the case for us, but that’s not the subject of this article 🙂
In general, the decision was made to automate the verification of the correctness of the filled fields.
We decided that sending a daily email with a list of incorrectly filled machines to all responsible engineers and their managers would be a good start.
At that time, one of the colleagues had already implemented a PowerShell script that collected information on all machines across all vCenters every day on a schedule and generated three CSV documents (each for its respective vCenter), which were placed on a shared drive. It was decided to use this script as a basis and supplement it with checks using the R language, with which we had some experience.
During the development process, the solution expanded to include email notifications, a database with a main and historical table (more on this later), and log analysis of vSphere to find the actual creators of VMs and the time of their creation.
Development was carried out using RStudio Desktop and PowerShell ISE.
The script runs from a regular Windows virtual machine.
Description of the overall logic.
The overall logic of the scripts turned out to be as follows.
- We collect data on virtual machines using a PowerShell script, which we invoke via R, and merge the results into a single CSV. The interaction between the languages is done similarly. (It would have been possible to pass data directly from R to PowerShell as variables, but it's complicated, and having intermediate CSV files makes it easier to debug and share intermediate results with others).
- Using R, we generate acceptable parameters for the fields whose values we verify. — We create a Word document that will contain the values of these fields for insertion into an informational letter that will respond to colleagues' questions, "Well, how am I supposed to fill this out?".
- We load data for all VMs from the CSV using R, form a dataframe, remove unnecessary fields, and create an informational XLSX document that will contain a summary of all VMs, which we upload to a shared resource.
- We apply all field validation checks to the dataframe of all VMs and create a table containing only the VMs with incorrectly filled fields (and only those fields).
- The obtained list of VMs is sent to another PowerShell script, which will check the vCenter logs for events related to VM creation, allowing us to specify the estimated time of VM creation and the presumed creator. This is in case no one claims responsibility for the machine. This script does not run quickly, especially if there are many logs, so we only look at the last 2 weeks and use a workflow that allows us to search for information across multiple VMs simultaneously. The example script includes detailed comments on this mechanism. We save the result in a CSV, which we load back into R.
- We create a well-formatted XLSX document, which will highlight incorrectly filled fields in red, apply filters to some columns, and include additional columns containing the presumed creators and the VM creation time.
- We are drafting an email in which we attach a document describing acceptable field values, along with a table of incorrectly filled VMs. The text will specify the total number of incorrectly created VMs, a link to the main resource, and a motivational image. If there are no incorrectly filled VMs, we will send a different email with a more cheerful motivational image.
- We are recording data for all VMs in the SQL Server database considering the implemented mechanism of historical tables (a very interesting mechanism — more details to follow).
The scripts themselves.
The main file with the code in R.
# Путь к рабочей директории (нужно для корректной работы через виндовый планировщик заданий)
setwd("C:ScriptsgetVm")
#### Подгружаем необходимые пакеты ####
library(tidyverse)
library(xlsx)
library(mailR)
library(rmarkdown)
##### Определяем пути к исходным файлам и другие переменные #####
source(file = "const.R", local = T, encoding = "utf-8")
# Проверяем существование файла со всеми ВМ и удаляем, если есть.
if (file.exists(filenameVmCreationRules)) {file.remove(filenameVmCreationRules)}
#### Создаём вордовский документ с допустимыми полями
render("VM_name_rules.Rmd",
output_format = word_document(),
output_file = filenameVmCreationRules)
# Проверяем существование файла со всеми ВМ и удаляем, если есть
if (file.exists(allVmXlsxPath)) {file.remove(allVmXlsxPath)}
#### Забираем данные по всем машинам через PowerShell скрипт. На выходе получим csv.
system(paste0("powershell -File ", getVmPsPath))
# Полный df
fullXslx_df <- allVmXlsxPath %>%
read.csv2(stringsAsFactors = FALSE)
# Checking the accuracy of filled fields
full_df <- fullXslx_df %>%
mutate(
# First, remove all extra spaces and tabs, then consider the comma as the separator, then check for inclusion in allowed values,
isSubsystemCorrect = Subsystem %>%
gsub("[[:space:]]", "", .) %>%
str_split(., ",") %>%
map(function(x) (all(x %in% AllowedValues$Subsystem))) %>%
as.logical(),
isOwnerCorrect = Owner %in% AllowedValues$Owner,
isCategoryCorrect = Category %in% AllowedValues$Category,
isCreatorCorrect = (!is.na(Creator) & Creator != ''),
isCreation.DateCorrect = map(Creation.Date, IsDate)
)
# Checking the existence of a file with all VMs and removing it if it exists.
if (file.exists(filenameAll)) {file.remove(filenameAll)}
#### Creating an xslx file with the report ####
# General data on a separate sheet
full_df %>% write.xlsx(file=filenameAll,
sheetName=names[1],
col.names=TRUE,
row.names=FALSE,
append=FALSE)
#### Creating an xslx file with incorrectly filled fields ####
# Creating df
incorrect_df <- full_df %>%
select(VM.Name,
IP.s,
Owner,
Subsystem,
Creator,
Category,
Creation.Date,
isOwnerCorrect,
isSubsystemCorrect,
isCategoryCorrect,
isCreatorCorrect,
vCenter.Name) %>%
filter(isSubsystemCorrect == F |
isOwnerCorrect == F |
isCategoryCorrect == F |
isCreatorCorrect == F)
# Checking the existence of a file with all VMs and removing it if it exists.
if (file.exists(filenameIncVM)) {file.remove(filenameIncVM)}
# Saving the list of VMs with unfilled fields in csv
incorrect_df %>%
select(VM.Name) %>%
write_csv2(path = filenameIncVM, append = FALSE)
# Filtering for insertion into the email
incorrect_df_filtered <- incorrect_df %>%
select(VM.Name,
IP.s,
Owner,
Subsystem,
Category,
Creator,
vCenter.Name,
Creation.Date
)
# Counting the number of rows
numberOfRows <- nrow(incorrect_df)
#### Начало условия ####
# Дальше либо у нас есть неправильно заполненные поля, либо нет.
# Если есть - запускаем ещё один скрипт
if (numberOfRows > 0) {
# Checking the existence of a file with creators and removing it if it exists.
if (file.exists(creatorsFilePath)) {file.remove(creatorsFilePath)}
# Running a PowerShell script that will find the creators of the found VMs. The output will be a csv.
system(paste0("powershell -File ", getCreatorsPath))
# Reading the file with creators
creators_df <- creatorsFilePath %>%
read.csv2(stringsAsFactors = FALSE)
# Filtering for insertion into the email, adding data from the table with creators
incorrect_df_filtered <- incorrect_df_filtered %>%
select(VM.Name,
IP.s,
Owner,
Subsystem,
Category,
Creator,
vCenter.Name,
Creation.Date
) %>%
left_join(creators_df, by = "VM.Name") %>%
rename(`Proposed Creator` = CreatedBy,
`Proposed Creation Date` = CreatedOn)
# Creating the email body
emailBody <- paste0(
'<html>
<h3>Good day, dear colleagues.</h3>
<p>You can view the complete and up-to-date information about virtual machines on disk H: here:<p>
<p>\server.ruVM', sourceFileFormat, '</p>
<p>Attached is a list of VMs with <strong>incorrectly filled</strong> fields. Their total is <strong>', numberOfRows, '</strong>.</p>
<p>The table now has 2 additional columns. <strong>Expected creator</strong> and <strong>Expected creation date</strong>, which are extracted from vCenter logs for the past 2 weeks</p>
<p>Creators of the machines are requested to verify the data and fill in the fields correctly. The rules for filling in the fields are also attached.</p>
<p><img src="data/meme.jpg"></p>
</html>'
)
# Checking the existence of the file
if (file.exists(filenameIncorrect)) {file.remove(filenameIncorrect)}
# Creating a nice table with formats, etc.
source(file = "email.R", local = T, encoding = "utf-8")
#### Forming an email with poorly signed machines ####
send.mail(from = emailParams$from,
to = emailParams$to,
subject = "VMs with incorrectly filled fields",
body = emailBody,
encoding = "utf-8",
html = TRUE,
inline = TRUE,
smtp = emailParams$smtpParams,
authenticate = TRUE,
send = TRUE,
attach.files = c(filenameIncorrect, filenameVmCreationRules),
debug = FALSE)
#### Next will be a block if there are no VM issues ####
} else {
# Forming the body of the email
emailBody <- paste0(
'<html>
<h3>Good afternoon, dear colleagues</h3>
<p>You can view the complete and up-to-date information about virtual machines on disk H: here:<p>
<p>\server.ruVM', sourceFileFormat, '</p>
<p>Additionally, at this moment, all VM fields are correctly filled</p>
<p><img src="data/meme_correct.jpg"></p>
</html>'
)
#### Forming an email without poorly filled VMs ####
send.mail(from = emailParams$from,
to = emailParams$to,
subject = "Summary information",
body = emailBody,
encoding = "utf-8",
html = TRUE,
inline = TRUE,
smtp = emailParams$smtpParams,
authenticate = TRUE,
send = TRUE,
debug = FALSE)
}
####### Writing data to the database #####
source(file = "DB.R", local = T, encoding = "utf-8")
A script to retrieve the list of VMs using PowerShell.
# Данные для подключения и другие переменные
$vCenterNames = @(
"vcenter01",
"vcenter02",
"vcenter03"
)
$vCenterUsername = "myusername"
$vCenterPassword = "mypassword"
$filename = "C:ScriptsgetVmdataallvmall-vm-$(get-date -f yyyy-MM-dd).csv"
$destinationSMB = "server.rumyfolder$vm"
$IP0=""
$IP1=""
$IP2=""
$IP3=""
$IP4=""
$IP5=""
# Подключение ко всем vCenter, что содержатся в переменной. Будет работать, если логин и пароль одинаковые (например, доменные)
Connect-VIServer -Server $vCenterNames -User $vCenterUsername -Password $vCenterPassword
write-host ""
# Создаём функцию с циклом по всем vCenter-ам
function Get-VMinventory {
# В этой переменной будет списко всех ВМ, как объектов
$AllVM = Get-VM | Sort Name
$cnt = $AllVM.Count
$count = 1
# Начинаем цикл по всем ВМ и собираем необходимые параметры каждого объекта
foreach ($vm in $AllVM) {
$StartTime = $(get-date)
$IP0 = $vm.Guest.IPAddress[0]
$IP1 = $vm.Guest.IPAddress[1]
$IP2 = $vm.Guest.IPAddress[2]
$IP3 = $vm.Guest.IPAddress[3]
$IP4 = $vm.Guest.IPAddress[4]
$IP5 = $vm.Guest.IPAddress[5]
If ($IP0 -ne $null) {If ($IP0.Contains(":") -ne 0) {$IP0=""}}
If ($IP1 -ne $null) {If ($IP1.Contains(":") -ne 0) {$IP1=""}}
If ($IP2 -ne $null) {If ($IP2.Contains(":") -ne 0) {$IP2=""}}
If ($IP3 -ne $null) {If ($IP3.Contains(":") -ne 0) {$IP3=""}}
If ($IP4 -ne $null) {If ($IP4.Contains(":") -ne 0) {$IP4=""}}
If ($IP5 -ne $null) {If ($IP5.Contains(":") -ne 0) {$IP5=""}}
$cluster = $vm | Get-Cluster | Select-Object -ExpandProperty name
$Bootime = $vm.ExtensionData.Runtime.BootTime
$TotalHDDs = $vm.ProvisionedSpaceGB -as [int]
$CreationDate = $vm.CustomFields.Item("CreationDate") -as [string]
$Creator = $vm.CustomFields.Item("Creator") -as [string]
$Category = $vm.CustomFields.Item("Category") -as [string]
$Owner = $vm.CustomFields.Item("Owner") -as [string]
$Subsystem = $vm.CustomFields.Item("Subsystem") -as [string]
$IPS = $vm.CustomFields.Item("IP") -as [string]
$vCPU = $vm.NumCpu
$CorePerSocket = $vm.ExtensionData.config.hardware.NumCoresPerSocket
$Sockets = $vCPU/$CorePerSocket
$Id = $vm.Id.Split('-')[2] -as [int]
# Собираем все параметры в один объект
$Vmresult = New-Object PSObject
$Vmresult | add-member -MemberType NoteProperty -Name "Id" -Value $Id
$Vmresult | add-member -MemberType NoteProperty -Name "VM Name" -Value $vm.Name
$Vmresult | add-member -MemberType NoteProperty -Name "Cluster" -Value $cluster
$Vmresult | add-member -MemberType NoteProperty -Name "Esxi Host" -Value $VM.VMHost
$Vmresult | add-member -MemberType NoteProperty -Name "IP Address 1" -Value $IP0
$Vmresult | add-member -MemberType NoteProperty -Name "IP Address 2" -Value $IP1
$Vmresult | add-member -MemberType NoteProperty -Name "IP Address 3" -Value $IP2
$Vmresult | add-member -MemberType NoteProperty -Name "IP Address 4" -Value $IP3
$Vmresult | add-member -MemberType NoteProperty -Name "IP Address 5" -Value $IP4
$Vmresult | add-member -MemberType NoteProperty -Name "IP Address 6" -Value $IP5
$Vmresult | add-member -MemberType NoteProperty -Name "vCPU" -Value $vCPU
$Vmresult | Add-Member -MemberType NoteProperty -Name "CPU Sockets" -Value $Sockets
$Vmresult | Add-Member -MemberType NoteProperty -Name "Core per Socket" -Value $CorePerSocket
$Vmresult | add-member -MemberType NoteProperty -Name "RAM (GB)" -Value $vm.MemoryGB
$Vmresult | add-member -MemberType NoteProperty -Name "Total-HDD (GB)" -Value $TotalHDDs
$Vmresult | add-member -MemberType NoteProperty -Name "Power State" -Value $vm.PowerState
$Vmresult | add-member -MemberType NoteProperty -Name "OS" -Value $VM.ExtensionData.summary.config.guestfullname
$Vmresult | Add-Member -MemberType NoteProperty -Name "Boot Time" -Value $Bootime
$Vmresult | add-member -MemberType NoteProperty -Name "VMTools Status" -Value $vm.ExtensionData.Guest.ToolsStatus
$Vmresult | add-member -MemberType NoteProperty -Name "VMTools Version" -Value $vm.ExtensionData.Guest.ToolsVersion
$Vmresult | add-member -MemberType NoteProperty -Name "VMTools Version Status" -Value $vm.ExtensionData.Guest.ToolsVersionStatus
$Vmresult | add-member -MemberType NoteProperty -Name "VMTools Running Status" -Value $vm.ExtensionData.Guest.ToolsRunningStatus
$Vmresult | add-member -MemberType NoteProperty -Name "Creation Date" -Value $CreationDate
$Vmresult | add-member -MemberType NoteProperty -Name "Creator" -Value $Creator
$Vmresult | add-member -MemberType NoteProperty -Name "Category" -Value $Category
$Vmresult | add-member -MemberType NoteProperty -Name "Owner" -Value $Owner
$Vmresult | add-member -MemberType NoteProperty -Name "Subsystem" -Value $Subsystem
$Vmresult | add-member -MemberType NoteProperty -Name "IP's" -Value $IPS
$Vmresult | add-member -MemberType NoteProperty -Name "vCenter Name" -Value $vm.Uid.Split('@')[1].Split(':')[0]
# Считаем общее и оставшееся время выполнения и выводим на экран результаты. Использовалось для тестирования, но по факту оказалось очень удобно.
$elapsedTime = $(get-date) - $StartTime
$totalTime = "{0:HH:mm:ss}" -f ([datetime]($elapsedTime.Ticks*($cnt - $count)))
clear-host
Write-Host "Processing" $count "from" $cnt
Write-host "Progress:" ([math]::Round($count/$cnt*100, 2)) "%"
Write-host "You have about " $totalTime "for cofee"
Write-host ""
$count++
# Выводим результат, чтобы цикл "знал" что является результатом выполнения одного прохода
$Vmresult
}
}
# Вызываем получившуюся функцию и сразу выгружаем результат в csv
$allVm = Get-VMinventory | Export-CSV -Path $filename -NoTypeInformation -UseCulture -Force
# Пытаемся выложить полученный файл в нужное нам место и, в случае ошибки, пишем лог.
try
{
Copy-Item $filename -Destination $destinationSMB -Force -ErrorAction SilentlyContinue
}
catch
{
$error | Export-CSV -Path $filename".error" -NoTypeInformation -UseCulture -Force
}A PowerShell script that extracts the creators of the virtual machines and their creation dates from the logs.
# Путь к файлу, из которого будем доставать список VM
$VMfilePath = "C:ScriptsgetVmcreators_VMcreators_VM_$(get-date -f yyyy-MM-dd).csv"
# Путь к файлу, в который будем записывать результат
$filePath = "C:ScriptsgetVmdatacreatorscreators-$(get-date -f yyyy-MM-dd).csv"
# Создаём вокрфлоу
Workflow GetCreators-Wf
{
# Параметры, которые можно будет передать при вызове скрипта
param([string[]]$VMfilePath)
# Параметры, которые доступны только внутри workflow
$vCenterUsername = "myusername"
$vCenterPassword = "mypassword"
$daysToLook = 14
$start = (get-date).AddDays(-$daysToLook)
$finish = get-date
# Значения, которые будут вписаны в csv для машин, по которым не будет ничего найдено
$UnknownUser = "UNKNOWN"
$UnknownCreatedTime = "0000-00-00"
# Определяем параметры подключения и выводной файл, которые будут доступны во всём скрипте.
$vCenterNames = @(
"vcenter01",
"vcenter02",
"vcenter03"
)
# Получаем список VM из csv и загружаем соответствующие объекты
$list = Import-Csv $VMfilePath -UseCulture | select -ExpandProperty VM.Name
# Цикл, который будет выполняться параллельно (по 5 машин за раз)
foreach -parallel ($row in $list)
{
# Это скрипт, который видит только свои переменные и те, которые ему переданы через $Using
InlineScript {
# Время начала выполнения отдельного блока
$StartTime = $(get-date)
Write-Host ""
Write-Host "Processing $Using:row started at $StartTime"
Write-Host ""
# Подключение оборачиваем в переменную, чтобы информация о нём не мешалась в консоли
$con = Connect-VIServer -Server $Using:vCenterNames -User $Using:vCenterUsername -Password $Using:vCenterPassword
# Получаем объект vm
$vm = Get-VM -Name $Using:row
# Ниже 2 одинаковые команды. Одна с фильтром по времени, вторая - без. Можно пользоваться тем,
$Event = $vm | Get-VIEvent -Start $Using:start -Finish $Using:finish -Types Info | Where { $_.Gettype().Name -eq "VmBeingDeployedEvent" -or $_.Gettype().Name -eq "VmCreatedEvent" -or $_.Gettype().Name -eq "VmRegisteredEvent" -or $_.Gettype().Name -eq "VmClonedEvent"}
# $Event = $vm | Get-VIEvent -Types Info | Where { $_.Gettype().Name -eq "VmBeingDeployedEvent" -or $_.Gettype().Name -eq "VmCreatedEvent" -or $_.Gettype().Name -eq "VmRegisteredEvent" -or $_.Gettype().Name -eq "VmClonedEvent"}
# Заполняем параметры в зависимости от того, удалось ли в логах найти что-то
If (($Event | Measure-Object).Count -eq 0){
$User = $Using:UnknownUser
$Created = $Using:UnknownCreatedTime
$CreatedFormat = $Using:UnknownCreatedTime
} Else {
If ($Event.Username -eq "" -or $Event.Username -eq $null) {
$User = $Using:UnknownUser
} Else {
$User = $Event.Username
} # Else
$CreatedFormat = $Event.CreatedTime
# Один из коллег отдельно просил, чтобы время было в таком формате, поэтому дублируем его. А в БД пойдёт нормальный формат.
$Created = $Event.CreatedTime.ToString('yyyy-MM-dd')
} # Else
Write-Host "Creator for $vm is $User. Creating object."
# Создаём объект. Добавляем параметры.
$Vmresult = New-Object PSObject
$Vmresult | add-member -MemberType NoteProperty -Name "VM Name" -Value $vm.Name
$Vmresult | add-member -MemberType NoteProperty -Name "CreatedBy" -Value $User
$Vmresult | add-member -MemberType NoteProperty -Name "CreatedOn" -Value $CreatedFormat
$Vmresult | add-member -MemberType NoteProperty -Name "CreatedOnFormat" -Value $Created
# Выводим результаты
$Vmresult
} # Inline
} # ForEach
}
$Creators = GetCreators-Wf $VMfilePath
# Записываем результат в файл
$Creators | select 'VM Name', CreatedBy, CreatedOn | Export-Csv -Path $filePath -NoTypeInformation -UseCulture -Force
Write-Host "CSV generetion finisghed at $(get-date). PROFIT"
The library deserves special attention. , which allowed the attachment to the email to be formatted visually (as management prefers), rather than just as a csv table.
Generating an attractive xlsx document with the list of incorrectly filled machines.
# Создаём новую книгу
# Возможные значения : "xls" и "xlsx"
wb<-createWorkbook(type="xlsx")
# Стили для имён рядов и колонок в таблицах
TABLE_ROWNAMES_STYLE <- CellStyle(wb) + Font(wb, isBold=TRUE)
TABLE_COLNAMES_STYLE <- CellStyle(wb) + Font(wb, isBold=TRUE) +
Alignment(wrapText=TRUE, horizontal="ALIGN_CENTER") +
Border(color="black", position=c("TOP", "BOTTOM"),
pen=c("BORDER_THIN", "BORDER_THICK"))
# Создаём новый лист
sheet <- createSheet(wb, sheetName = names[2])
# Добавляем таблицу
addDataFrame(incorrect_df_filtered,
sheet, startRow=1, startColumn=1, row.names=FALSE, byrow=FALSE,
colnamesStyle = TABLE_COLNAMES_STYLE,
rownamesStyle = TABLE_ROWNAMES_STYLE)
# Меняем ширину, чтобы форматирование было автоматическим
autoSizeColumn(sheet = sheet, colIndex=c(1:ncol(incorrect_df)))
# Добавляем фильтры
addAutoFilter(sheet, cellRange = "C1:G1")
# Определяем стиль
fo2 <- Fill(foregroundColor="red")
cs2 <- CellStyle(wb,
fill = fo2,
dataFormat = DataFormat("@"))
# Находим ряды с неверно заполненным полем Владельца и применяем к ним определённый стиль
rowsOwner <- getRows(sheet, rowIndex = (which(!incorrect_df$isOwnerCorrect) + 1))
cellsOwner <- getCells(rowsOwner, colIndex = which( colnames(incorrect_df_filtered) == "Owner" ))
lapply(names(cellsOwner), function(x) setCellStyle(cellsOwner[[x]], cs2))
# Находим ряды с неверно заполненным полем Подсистемы и применяем к ним определённый стиль
rowsSubsystem <- getRows(sheet, rowIndex = (which(!incorrect_df$isSubsystemCorrect) + 1))
cellsSubsystem <- getCells(rowsSubsystem, colIndex = which( colnames(incorrect_df_filtered) == "Subsystem" ))
lapply(names(cellsSubsystem), function(x) setCellStyle(cellsSubsystem[[x]], cs2))
# Аналогично по Категории
rowsCategory <- getRows(sheet, rowIndex = (which(!incorrect_df$isCategoryCorrect) + 1))
cellsCategory <- getCells(rowsCategory, colIndex = which( colnames(incorrect_df_filtered) == "Category" ))
lapply(names(cellsCategory), function(x) setCellStyle(cellsCategory[[x]], cs2))
# Создатель
rowsCreator <- getRows(sheet, rowIndex = (which(!incorrect_df$isCreatorCorrect) + 1))
cellsCreator <- getCells(rowsCreator, colIndex = which( colnames(incorrect_df_filtered) == "Creator" ))
lapply(names(cellsCreator), function(x) setCellStyle(cellsCreator[[x]], cs2))
# Сохраняем файл
saveWorkbook(wb, filenameIncorrect)The output looks something like this:
There was also an interesting nuance in configuring the Windows scheduler. It was difficult to find the right parameters for permissions and settings to get everything running as needed. Ultimately, an R library was found that automatically creates a task to execute the R script and doesn't forget to include a file for logs. Then you can manually adjust the task later.
A snippet of R code with two examples that creates a task in the Windows scheduler.
library(taskscheduleR)
myscript <- file.path(getwd(), "all_vm.R")
## Running the script after 62 seconds
taskscheduler_create(taskname = "getAllVm", rscript = myscript,
schedule = "ONCE", starttime = format(Sys.time() + 62, "%H:%M"))
## Running the script daily at 09:10
taskscheduler_create(taskname = "getAllVmDaily", rscript = myscript,
schedule = "WEEKLY",
days = c("MON", "TUE", "WED", "THU", "FRI"),
starttime = "02:00")
## Deleting tasks
taskscheduler_delete(taskname = "getAllVm")
taskscheduler_delete(taskname = "getAllVmDaily")
# Checking logs (last 4 lines)
tail(readLines("all_vm.log"), sep = "n", n = 4)Separately about the database.
After setting up the script, other questions began to arise. For example, I wanted to find the date when the VM was deleted, but the logs in vCenter have already been cleared. Since the script saves files to a folder every day and doesn't clean them (we do that manually when we remember), you can look through the old files and find the first one in which this VM is absent. But that's not ideal.
I wanted to create a historical database.
The functionality of MS SQL SERVER came to the rescue — system-versioned temporal table. It's usually translated as temporal tables (not temporary tables).
You can read in detail in .
In brief, we create a table, specify that it will be versioned, and SQL Server creates two additional datetime columns in this table (the record creation date and the record expiration date) and an additional table where changes will be logged. As a result, we get up-to-date information and, through simple queries, examples of which are provided in the documentation, we can see either the life cycle of a specific virtual machine or the state of all VMs at a given point in time.
From a performance standpoint, the transaction writing to the main table won’t be completed until the transaction writing to the temporal table is finished. That is to say, for tables with a high volume of write operations, this functionality should be implemented with caution, but in our case, it's really a cool feature.
To ensure that the mechanism works correctly, I had to write a small piece of code in R that would compare the new table with the data for all VMs stored in the database and only write the changed rows. The code isn't particularly complex, it uses the compareDF library, but I will also provide it below.
R code for writing data to the database
# Подцепляем пакеты
library(odbc)
library(compareDF)
# Формируем коннект
con <- dbConnect(odbc(),
Driver = "ODBC Driver 13 for SQL Server",
Server = DBParams$server,
Database = DBParams$database,
UID = DBParams$UID,
PWD = DBParams$PWD,
Port = 1433)
#### Проверяем есть ли таблица. Если нет - создаём. ####
if (!dbExistsTable(con, DBParams$TblName)) {
#### Создаём таблицу ####
create <- dbSendStatement(
con,
paste0(
'CREATE TABLE ',
DBParams$TblName,
'(
[Id] [int] NOT NULL PRIMARY KEY CLUSTERED,
[VM.Name] [varchar](255) NULL,
[Cluster] [varchar](255) NULL,
[Esxi.Host] [varchar](255) NULL,
[IP.Address.1] [varchar](255) NULL,
[IP.Address.2] [varchar](255) NULL,
[IP.Address.3] [varchar](255) NULL,
[IP.Address.4] [varchar](255) NULL,
[IP.Address.5] [varchar](255) NULL,
[IP.Address.6] [varchar](255) NULL,
[vCPU] [int] NULL,
[CPU.Sockets] [int] NULL,
[Core.per.Socket] [int] NULL,
[RAM..GB.] [int] NULL,
[Total.HDD..GB.] [int] NULL,
[Power.State] [varchar](255) NULL,
[OS] [varchar](255) NULL,
[Boot.Time] [varchar](255) NULL,
[VMTools.Status] [varchar](255) NULL,
[VMTools.Version] [int] NULL,
[VMTools.Version.Status] [varchar](255) NULL,
[VMTools.Running.Status] [varchar](255) NULL,
[Creation.Date] [varchar](255) NULL,
[Creator] [varchar](255) NULL,
[Category] [varchar](255) NULL,
[Owner] [varchar](255) NULL,
[Subsystem] [varchar](255) NULL,
[IP.s] [varchar](255) NULL,
[vCenter.Name] [varchar](255) NULL,
DateFrom datetime2 GENERATED ALWAYS AS ROW START NOT NULL,
DateTo datetime2 GENERATED ALWAYS AS ROW END NOT NULL,
PERIOD FOR SYSTEM_TIME (DateFrom, DateTo)
) ON [PRIMARY]
WITH (SYSTEM_VERSIONING = ON (HISTORY_TABLE = ', DBParams$TblHistName,'));'
)
)
# Отправляем подготовленный запрос
dbClearResult(create)
} # if
#### Начало работы с таблицей ####
# Обозначаем таблицу, с которой будем работать
allVM_db_con <- tbl(con, DBParams$TblName)
#### Сравниваем таблицы ####
# Собираем данные с таблицы (убираем служебные временные поля)
allVM_db <- allVM_db_con %>%
select(c(-"DateTo", -"DateFrom")) %>%
collect()
# Создаём таблицу со сравнением объектов. Сравниваем по Id
# Удалённые объекты там будут помечены через -, созданные через +, изменённые через - и +
ctable_VM <- fullXslx_df %>%
compare_df(allVM_db,
c("Id"))
#### Удаление строк ####
# Выдираем Id виртуалок, записи о которых надо удалить
remove_Id <- ctable_VM$comparison_df %>%
filter(chng_type == "-") %>%
select(Id)
# Проверяем, что есть записи (если записей нет - и удалять ничего не нужно)
if (remove_Id %>% nrow() > 0) {
# Конструируем шаблон для запроса на удаление данных
delete <- dbSendStatement(con,
paste0('
DELETE
FROM ',
DBParams$TblName,
' WHERE "Id"=?
') # paste
) # send
# Создаём запрос на удаление данных
dbBind(delete, remove_Id)
# Отправляем подготовленный запрос
dbClearResult(delete)
} # if
#### Добавление строк ####
# Выделяем таблицу, содержащую строки, которые нужно добавить.
allVM_add <- ctable_VM$comparison_df %>%
filter(chng_type == "+") %>%
select(-chng_type)
# Проверяем, есть ли строки, которые нужно добавить и добавляем (если нет - не добавляем)
if (allVM_add %>% nrow() > 0) {
# Пишем таблицу со всеми необходимыми данными
dbWriteTable(con,
DBParams$TblName,
allVM_add,
overwrite = FALSE,
append = TRUE)
} # if
#### Не забываем сделать дисконнект ####
dbDisconnect(con)Total
As a result of implementing the script, order has been maintained over several months. Occasionally, incorrectly filled VMs appear, but the script serves as a good reminder, and rarely does a VM make it onto the list two days in a row.
There was also a foundation laid for historical data analysis.
It's clear that much of this can be implemented not 'on the knee,' but through specialized software, but the task was interesting and can be considered optional.
R has once again proven to be an excellent universal language, perfectly suited not only for solving statistical problems but also serving as a great 'bridge' between other data sources.
Source: habr.com
