Bosun Scollector: External Collectors

From WikiOD

Remarks[edit | edit source]

Scollector supports tcollector style external collectors that can be used to send metrics to Bosun via custom scripts or executables. External collectors are a great way to get started collecting data, but when possible it is recommended for applications to send data directly to Bosun or to update scollector so that it natively supports additional systems.

The ColDir configuration key specifies the external collector directory, which is usually set to something like /opt/scollector/collectors/ in Linux or C:\Program Files\scollector\collectors\ in Windows. It should contain numbered directories just like the ones used in OpenTSDB tcollector. Each directory represents how often scollector will try to invoke the collectors in that folder (example: 60 = every 60 seconds). Use a directory named 0 for any executables or scripts that will run continuously and create output on their own schedule. Any non-numeric named directories will be ignored, and a lib and etc directory are often used for library and config data shared by all collectors.

External collectors can use either the simple data output format from tcollector or they can send JSON data if they want to include metadata.

Powershell external collector script function[edit | edit source]

<#
    .DESCRIPTION
        Writes the metric out in bosun external collector format which is compatible with scollector external scripts
    .PARAMETER metric
        Name of the metric (eg : my.metric)
    .PARAMETER type
        Type of metric (counter, gauge, etc)
    .PARAMETER unit
        Type of unit (connections, operations, etc)
    .PARAMETER desc
        Description of the metric
    .PARAMETER value
        The current value for the metric
#>
function Write-Metric
{
param(
    [string]$metric,
    [string]$type,
    [string]$unit,
    [string]$desc,
    $value
)

$epoch = New-Object DateTime (1970,1,1)

$obj = @{
    metric = $metric
    name = "rate"
    value = $type
}

Write*Host (ConvertTo-Json $obj -Compress)

$obj.name="unit"
$obj.value=$unit

Write*Host (ConvertTo-Json $obj -Compress)

$obj.name="desc"
$obj.value=$desc

Write*Host (ConvertTo-Json $obj -Compress)

$output = @{
    metric = $metric
    timestamp= [int]([datetime]::UtcNow.Subtract($epoch).TotalSeconds)
    value=$value
    tags= @{
        host=$env:computername.ToLower()
    }
}

Write*Host (ConvertTo-Json $output -Compress)

}

Sample collector written in PowerShell[edit | edit source]

#Example of a PowerShell external collector. See http://bosun.org/scollector/external-collectors for details
#This file should be saved in C:\Program Files\scollector\collectors\0\mymetrics.ps1 since it is a continuous output script
#scollector.toml should have ColDir = 'C:\Program Files\scollector\collectors'

#Setup format strings and other variables
$epoch = New-Object DateTime (1970,1,1)
$MetricMetadata='{{"metric":"{0}","name":"{1}","value":"{2}"}}'
$MetricData='{{"metric":"{0}","timestamp":{1:F0},"value":{2:G}{3}}}'
$MetricTags=',"tags":{{{0}}}'
$Base="mymetric"

#Send metadata for each metric once on startup (Scollector will resend to Bosun periodically)
Write*Output ($MetricMetadata -f "$Base.test","rate","gauge") #See https://godoc.org/bosun.org/metadata#RateType
Write*Output ($MetricMetadata -f "$Base.test","unit","item")  #See https://godoc.org/bosun.org/metadata#Unit
Write*Output ($MetricMetadata -f "$Base.test","desc","A test metric")

#Create tags and send metrics
$tags=$MetricTags -f '"mykey":"myvalue"' #generate static tags here. Can append additional tags in the loop if needed. 
#Use $tags="" to exclude all tags but those added by Scollector.
Write*Output ($MetricData -f "$Base.test",[datetime]::UtcNow.Subtract($epoch).TotalSeconds,42.123,$tags)
do {
    $delay = Get-Random -Minimum 5 -Maximum 25
    sleep -Seconds $delay
    Write-Output ($MetricData -f "$Base.test",[datetime]::UtcNow.Subtract($epoch).TotalSeconds,$delay,$tags)
} while ($true)

#If a continuous output script ever exits scollector will restart it. If you just want periodic data every 60 seconds you 
#can use a /60/ folder instead of /0/ and allow the script to exit when finished sending a batch of metrics.

Twitter Collector written in Go[edit | edit source]

The following can be saved as main.go. After you update the EDITME settings and build the executable it can be used as a continuous external collector.

package main

import (
    "fmt"
    "log"
    "net/url"
    "strconv"
    "time"

    "github.com/ChimeraCoder/anaconda"
)

func main() {
    anaconda.SetConsumerKey("EDITME")
    anaconda.SetConsumerSecret("EDITME")
    api := anaconda.NewTwitterApi("EDITME", "EDITME")
    v := url.Values{}
    sr, err := api.GetSearch("stackoverflow", nil)
    if err != nil {
        log.Println(err)
    }
    var since_id int64 = 0
    for _, tweet := range sr {
        if tweet.Id > since_id {
            since_id = tweet.Id
        }
    }
    count := 0
    for {
        now := time.Now().Unix()
        v.Set("result_type", "recent")
        v.Set("since_id", strconv.FormatInt(since_id, 10))
        sr, err := api.GetSearch("stackoverflow", nil)
        if err != nil {
            log.Println(err)
        }
        for _, tweet := range sr {
            if tweet.Id > since_id {
                count += 1
                since_id = tweet.Id
            }
        }
        fmt.Println("twitter.tweet_count", now, count, "query=stackoverflow")
        time.Sleep(time.Second * 30)
    }
}

Hadoop HDFS disk usage written in Bash[edit | edit source]

This is a continuous collector that uses the hadoop fs -du -s /hbase/* command to get details about the HDFS disk usage. This metric is very useful for tracking space in an OpenTSDB system.

#!/bin/bash
while true; do
    while read -r bytes raw_bytes path; do
        echo "hdfs.du $(date +"%s") $bytes path=$path"
        #https://community.cloudera.com/t5/Storage-Random-Access-HDFS/hdfs-du-format-change/td-p/27192 KMB 2015-08-24T12:01:20Z
        echo "hdfs.du.raw $(date +"%s") $raw_bytes path=$path"
    done < <(hadoop fs -du -s  /hbase/*)
    sleep 30
done

StackExchange.Exceptional collector written in Go with Metadata[edit | edit source]

The following Go file can be compiled into a continuous external collector that will query a MSSQL server database that uses the StackExchange.Exceptional schema. It will query multiple servers/databases for all exceptions since UTC 00:00 to convert the raw entries into a counter. It also uses the bosun.org/metadata package to include metadata for the exceptional.exceptions.count metric.

/*
Exceptional is an scollector external collector for StackExchange.Exceptional.
*/
package main

import (
    "database/sql"
    "encoding/json"
    "fmt"
    "log"
    "strings"
    "time"

    "bosun.org/metadata"
    "bosun.org/opentsdb"

    _ "github.com/denisenkom/go-mssqldb"
)

func mssqlConnect(server, database, user, pass, port string) (*sql.DB, error) {
    dsn := fmt.Sprintf("server=%s;port=%s;database=%s;user id=%s;password=%s", server, port, database, user, pass)
    return sql.Open("mssql", dsn)
}

type Exceptions struct {
    GUID            string
    ApplicationName string
    MachineName     string
    CreationDate    time.Time
    Type            string
    IsProtected     int
    Host            string
    Url             string
    HTTPMethod      string
    IPAddress       string
    Source          string
    Message         string
    Detail          string
    StatusCode      int
    SQL             string
    DeletionDate    time.Time
    FullJson        string
    ErrorHash       int
    DuplicateCount  int
}

type ExceptionsCount struct {
    ApplicationName string
    MachineName     string
    Count           int64
    Source          string
}

type ExceptionsDB struct {
    Server     string
    DBName     string
    DBPassword string
    DBPort     string
    Source     string
}

const (
    defaultPassword = "EnterPasswordHere"
    defaultPort     = "1433"

    metric     = "exceptional.exceptions.count"
    descMetric = "The number of exceptions thrown per second by applications and machines. Data is queried from multiple sources. See status instances for details on exceptions."
)

func main() {
    mds := []metadata.Metasend{
        {
            Metric: metric,
            Name:   "rate",
            Value:  "counter",
        },
        {
            Metric: metric,
            Name:   "unit",
            Value:  metadata.Error,
        },
        {
            Metric: metric,
            Name:   "desc",
            Value:  descMetric,
        },
    }
    for _, m := range mds {
        b, err := json.Marshal(m)
        if err != nil {
            log.Fatal(err)
        }
        fmt.Println(string(b))
    }
    instances := [...]ExceptionsDB{
        {"NY_AG", "NY.Exceptions", defaultPassword, defaultPort, "NY_Status"},
        {"CO-SQL", "CO.Exceptions", defaultPassword, defaultPort, "CO_Status"},
        {"NY-INTSQL", "Int.Exceptions", defaultPassword, defaultPort, "INT_Status"},
    }
    for _, exdb := range instances {
        go run(exdb)
    }
    select {}
}

func run(exdb ExceptionsDB) {
    const interval = time.Second * 30

    query := func() {
        // Database name is the same as the username
        db, err := mssqlConnect(exdb.Server, exdb.DBName, exdb.DBName, exdb.DBPassword, exdb.DBPort)
        if err != nil {
            log.Println(err)
        }
        defer db.Close()
        var results []ExceptionsCount
        sqlQuery := `
        SELECT ApplicationName, MachineName, MAX(Count) as Count FROM
        (
            --New since UTC rollover
            SELECT ApplicationName, MachineName, Sum(DuplicateCount) as Count from Exceptions
            WHERE CreationDate > CONVERT (date, GETUTCDATE())
            GROUP BY MachineName, ApplicationName
            UNION --Zero out any app/machine combos that had exceptions in last 24 hours
            SELECT DISTINCT ex.ApplicationName, ex.MachineName, 0 as Count from Exceptions ex WHERE ex.CreationDate Between Convert(Date, GETUTCDATE()-1) And Convert(Date, GETUTCDATE())
        ) as T
        GROUP By T.MachineName, T.ApplicationName`
        rows, err := db.Query(sqlQuery)
        if err != nil {
            log.Println(err)
            return
        }
        defer rows.Close()
        for rows.Next() {
            var r ExceptionsCount
            if err := rows.Scan(&r.ApplicationName, &r.MachineName, &r.Count); err != nil {
                log.Println(err)
                continue
            }
            r.Source = exdb.Source
            results = append(results, r)
        }
        if err := rows.Err(); err != nil {
            log.Println(err)
        }
        if len(results) > 0 {
            now := time.Now().Unix()
            for _, r := range results {
                application, err := opentsdb.Clean(r.ApplicationName)
                if err != nil {
                    log.Println(err)
                    continue
                }
                db := opentsdb.DataPoint{
                    Metric:    metric,
                    Timestamp: now,
                    Value:     r.Count,
                    Tags: opentsdb.TagSet{
                        "application": application,
                        "machine":     strings.ToLower(r.MachineName),
                        "source":      r.Source,
                    },
                }
                b, err := db.MarshalJSON()
                if err != nil {
                    log.Println(err)
                    continue
                }
                fmt.Println(string(b))
            }
        }
    }

    for {
        wait := time.After(interval)
        query()
        <-wait
    }
}

Credit:Stack_Overflow_Documentation