2015-08-24 17:27:14 +00:00
package main
import (
"database/sql"
"flag"
"fmt"
2016-08-20 16:32:08 +00:00
"io/ioutil"
2016-08-06 01:54:41 +00:00
"math"
2015-08-24 17:27:14 +00:00
"net/http"
"os"
2016-09-07 20:29:30 +00:00
"strconv"
2016-01-06 19:19:41 +00:00
"time"
2016-11-16 15:16:45 +00:00
"regexp"
"errors"
2015-08-24 17:27:14 +00:00
2016-09-07 20:29:30 +00:00
"gopkg.in/yaml.v2"
2016-08-06 01:54:41 +00:00
2015-08-24 17:27:14 +00:00
_ "github.com/lib/pq"
"github.com/prometheus/client_golang/prometheus"
2016-08-06 01:54:41 +00:00
"github.com/prometheus/common/log"
2015-08-24 17:27:14 +00:00
)
2016-05-28 07:03:38 +00:00
var Version string = "0.0.1"
2016-05-28 04:19:55 +00:00
2015-08-24 17:27:14 +00:00
var (
listenAddress = flag . String (
2016-11-16 12:02:42 +00:00
"web.listen-address" , ":9187" ,
2015-08-24 17:27:14 +00:00
"Address to listen on for web interface and telemetry." ,
)
metricPath = flag . String (
"web.telemetry-path" , "/metrics" ,
"Path under which to expose metrics." ,
)
2016-06-28 23:56:47 +00:00
queriesPath = flag . String (
"extend.query-path" , "" ,
"Path to custom queries to run." ,
)
2016-08-20 16:32:08 +00:00
onlyDumpMaps = flag . Bool (
"dumpmaps" , false ,
"Do not run, simply dump the maps." ,
)
2015-08-24 17:27:14 +00:00
)
// Metric name parts.
const (
// Namespace for all metrics.
namespace = "pg"
// Subsystems.
2016-01-06 19:19:41 +00:00
exporter = "exporter"
2016-11-16 14:59:46 +00:00
// Metric label used for static string data thats handy to send to Prometheus
// e.g. version
staticLabelName = "static"
2015-08-24 17:27:14 +00:00
)
2016-11-16 14:59:46 +00:00
// Highest version of Postgres we have explicit behavior for. This is the
// assumed default if the version string does match any known versions.
const HighestSupportedVersion = "9.6"
2015-08-24 17:27:14 +00:00
// landingPage contains the HTML served at '/'.
// TODO: Make this nicer and more informative.
var landingPage = [ ] byte ( ` < html >
< head > < title > Postgres exporter < / title > < / head >
< body >
< h1 > Postgres exporter < / h1 >
< p > < a href = ' ` + *metricPath + ` ' > Metrics < / a > < / p >
< / body >
< / html >
` )
type ColumnUsage int
2016-01-06 19:19:41 +00:00
2015-08-24 17:27:14 +00:00
const (
2016-01-06 19:19:41 +00:00
DISCARD ColumnUsage = iota // Ignore this column
LABEL ColumnUsage = iota // Use this column as a label
COUNTER ColumnUsage = iota // Use this column as a counter
GAUGE ColumnUsage = iota // Use this column as a gauge
MAPPEDMETRIC ColumnUsage = iota // Use this column with the supplied mapping of text values
DURATION ColumnUsage = iota // This column should be interpreted as a text duration (and converted to milliseconds)
2015-08-24 17:27:14 +00:00
)
2016-11-16 14:59:46 +00:00
// Special case matric mappings
const (
// Which metric mapping should be acquired using "SHOW" queries
SHOW_METRIC = "pg_runtime_variables"
)
// Regex used to get the "short-version" from the postgres version field.
var versionRegex = regexp . MustCompile ( ` ^\w+ (\d+\.\d+) ` )
// Parses the version of postgres into the short version string we can use to
// match behaviors.
func parseVersion ( versionString string ) string {
submatches := versionRegex . FindStringSubmatch ( versionString )
if len ( submatches ) > 1 {
return submatches [ 1 ]
}
log . Debugln ( "Could not parse postgres version regex:" , versionString )
return ""
}
2015-08-24 17:27:14 +00:00
// User-friendly representation of a prometheus descriptor map
type ColumnMapping struct {
2016-01-06 19:19:41 +00:00
usage ColumnUsage
2015-08-24 17:27:14 +00:00
description string
2016-01-06 19:19:41 +00:00
mapping map [ string ] float64 // Optional column mapping for MAPPEDMETRIC
2015-08-24 17:27:14 +00:00
}
// Groups metric maps under a shared set of labels
type MetricMapNamespace struct {
2016-01-06 19:19:41 +00:00
labels [ ] string // Label names for this namespace
columnMappings map [ string ] MetricMap // Column mappings in this namespace
2015-08-24 17:27:14 +00:00
}
// Stores the prometheus metric description which a given column will be mapped
// to by the collector
type MetricMap struct {
2016-01-06 19:19:41 +00:00
discard bool // Should metric be discarded during mapping?
vtype prometheus . ValueType // Prometheus valuetype
desc * prometheus . Desc // Prometheus descriptor
conversion func ( interface { } ) ( float64 , bool ) // Conversion function to turn PG result into float64
2015-08-24 17:27:14 +00:00
}
// Metric descriptors for dynamically created metrics.
2015-12-09 05:41:28 +00:00
var variableMaps = map [ string ] map [ string ] ColumnMapping {
2016-01-06 19:19:41 +00:00
"pg_runtime_variable" : map [ string ] ColumnMapping {
"max_connections" : { GAUGE , "Sets the maximum number of concurrent connections." , nil } ,
"max_files_per_process" : { GAUGE , "Sets the maximum number of simultaneously open files for each server process." , nil } ,
"max_function_args" : { GAUGE , "Shows the maximum number of function arguments." , nil } ,
"max_identifier_length" : { GAUGE , "Shows the maximum identifier length." , nil } ,
"max_index_keys" : { GAUGE , "Shows the maximum number of index keys." , nil } ,
"max_locks_per_transaction" : { GAUGE , "Sets the maximum number of locks per transaction." , nil } ,
"max_pred_locks_per_transaction" : { GAUGE , "Sets the maximum number of predicate locks per transaction." , nil } ,
"max_prepared_transactions" : { GAUGE , "Sets the maximum number of simultaneously prepared transactions." , nil } ,
2015-12-09 05:41:28 +00:00
//"max_stack_depth" : { GAUGE, "Sets the maximum number of concurrent connections.", nil }, // No dehumanize support yet
2016-01-06 19:19:41 +00:00
"max_standby_archive_delay" : { DURATION , "Sets the maximum delay before canceling queries when a hot standby server is processing archived WAL data." , nil } ,
"max_standby_streaming_delay" : { DURATION , "Sets the maximum delay before canceling queries when a hot standby server is processing streamed WAL data." , nil } ,
"max_wal_senders" : { GAUGE , "Sets the maximum number of simultaneously running WAL sender processes." , nil } ,
2015-12-09 05:41:28 +00:00
} ,
}
2016-08-20 16:32:08 +00:00
func dumpMaps ( ) {
for name , cmap := range metricMaps {
query , ok := queryOverrides [ name ]
if ok {
fmt . Printf ( "%s: %s\n" , name , query )
} else {
fmt . Println ( name )
}
for column , details := range cmap {
fmt . Printf ( " %-40s %v\n" , column , details )
}
fmt . Println ( )
}
}
2016-01-06 19:19:41 +00:00
var metricMaps = map [ string ] map [ string ] ColumnMapping {
"pg_stat_bgwriter" : map [ string ] ColumnMapping {
"checkpoints_timed" : { COUNTER , "Number of scheduled checkpoints that have been performed" , nil } ,
"checkpoints_req" : { COUNTER , "Number of requested checkpoints that have been performed" , nil } ,
"checkpoint_write_time" : { COUNTER , "Total amount of time that has been spent in the portion of checkpoint processing where files are written to disk, in milliseconds" , nil } ,
"checkpoint_sync_time" : { COUNTER , "Total amount of time that has been spent in the portion of checkpoint processing where files are synchronized to disk, in milliseconds" , nil } ,
"buffers_checkpoint" : { COUNTER , "Number of buffers written during checkpoints" , nil } ,
"buffers_clean" : { COUNTER , "Number of buffers written by the background writer" , nil } ,
"maxwritten_clean" : { COUNTER , "Number of times the background writer stopped a cleaning scan because it had written too many buffers" , nil } ,
"buffers_backend" : { COUNTER , "Number of buffers written directly by a backend" , nil } ,
"buffers_backend_fsync" : { COUNTER , "Number of times a backend had to execute its own fsync call (normally the background writer handles those even when the backend does its own write)" , nil } ,
"buffers_alloc" : { COUNTER , "Number of buffers allocated" , nil } ,
"stats_reset" : { COUNTER , "Time at which these statistics were last reset" , nil } ,
} ,
"pg_stat_database" : map [ string ] ColumnMapping {
"datid" : { LABEL , "OID of a database" , nil } ,
"datname" : { LABEL , "Name of this database" , nil } ,
"numbackends" : { GAUGE , "Number of backends currently connected to this database. This is the only column in this view that returns a value reflecting current state; all other columns return the accumulated values since the last reset." , nil } ,
"xact_commit" : { COUNTER , "Number of transactions in this database that have been committed" , nil } ,
"xact_rollback" : { COUNTER , "Number of transactions in this database that have been rolled back" , nil } ,
"blks_read" : { COUNTER , "Number of disk blocks read in this database" , nil } ,
"blks_hit" : { COUNTER , "Number of times disk blocks were found already in the buffer cache, so that a read was not necessary (this only includes hits in the PostgreSQL buffer cache, not the operating system's file system cache)" , nil } ,
"tup_returned" : { COUNTER , "Number of rows returned by queries in this database" , nil } ,
"tup_fetched" : { COUNTER , "Number of rows fetched by queries in this database" , nil } ,
"tup_inserted" : { COUNTER , "Number of rows inserted by queries in this database" , nil } ,
"tup_updated" : { COUNTER , "Number of rows updated by queries in this database" , nil } ,
"tup_deleted" : { COUNTER , "Number of rows deleted by queries in this database" , nil } ,
"conflicts" : { COUNTER , "Number of queries canceled due to conflicts with recovery in this database. (Conflicts occur only on standby servers; see pg_stat_database_conflicts for details.)" , nil } ,
"temp_files" : { COUNTER , "Number of temporary files created by queries in this database. All temporary files are counted, regardless of why the temporary file was created (e.g., sorting or hashing), and regardless of the log_temp_files setting." , nil } ,
"temp_bytes" : { COUNTER , "Total amount of data written to temporary files by queries in this database. All temporary files are counted, regardless of why the temporary file was created, and regardless of the log_temp_files setting." , nil } ,
"deadlocks" : { COUNTER , "Number of deadlocks detected in this database" , nil } ,
"blk_read_time" : { COUNTER , "Time spent reading data file blocks by backends in this database, in milliseconds" , nil } ,
"blk_write_time" : { COUNTER , "Time spent writing data file blocks by backends in this database, in milliseconds" , nil } ,
"stats_reset" : { COUNTER , "Time at which these statistics were last reset" , nil } ,
} ,
"pg_stat_database_conflicts" : map [ string ] ColumnMapping {
"datid" : { LABEL , "OID of a database" , nil } ,
"datname" : { LABEL , "Name of this database" , nil } ,
"confl_tablespace" : { COUNTER , "Number of queries in this database that have been canceled due to dropped tablespaces" , nil } ,
"confl_lock" : { COUNTER , "Number of queries in this database that have been canceled due to lock timeouts" , nil } ,
"confl_snapshot" : { COUNTER , "Number of queries in this database that have been canceled due to old snapshots" , nil } ,
"confl_bufferpin" : { COUNTER , "Number of queries in this database that have been canceled due to pinned buffers" , nil } ,
"confl_deadlock" : { COUNTER , "Number of queries in this database that have been canceled due to deadlocks" , nil } ,
} ,
"pg_locks" : map [ string ] ColumnMapping {
"datname" : { LABEL , "Name of this database" , nil } ,
"mode" : { LABEL , "Type of Lock" , nil } ,
"count" : { GAUGE , "Number of locks" , nil } ,
2015-08-24 17:27:14 +00:00
} ,
2016-01-06 19:19:41 +00:00
"pg_stat_replication" : map [ string ] ColumnMapping {
"pid" : { DISCARD , "Process ID of a WAL sender process" , nil } ,
"usesysid" : { DISCARD , "OID of the user logged into this WAL sender process" , nil } ,
"usename" : { DISCARD , "Name of the user logged into this WAL sender process" , nil } ,
"application_name" : { DISCARD , "Name of the application that is connected to this WAL sender" , nil } ,
"client_addr" : { LABEL , "IP address of the client connected to this WAL sender. If this field is null, it indicates that the client is connected via a Unix socket on the server machine." , nil } ,
"client_hostname" : { DISCARD , "Host name of the connected client, as reported by a reverse DNS lookup of client_addr. This field will only be non-null for IP connections, and only when log_hostname is enabled." , nil } ,
"client_port" : { DISCARD , "TCP port number that the client is using for communication with this WAL sender, or -1 if a Unix socket is used" , nil } ,
"backend_start" : { DISCARD , "with time zone Time when this process was started, i.e., when the client connected to this WAL sender" , nil } ,
"backend_xmin" : { DISCARD , "The current backend's xmin horizon." , nil } ,
"state" : { LABEL , "Current WAL sender state" , nil } ,
"sent_location" : { DISCARD , "Last transaction log position sent on this connection" , nil } ,
"write_location" : { DISCARD , "Last transaction log position written to disk by this standby server" , nil } ,
"flush_location" : { DISCARD , "Last transaction log position flushed to disk by this standby server" , nil } ,
"replay_location" : { DISCARD , "Last transaction log position replayed into the database on this standby server" , nil } ,
"sync_priority" : { DISCARD , "Priority of this standby server for being chosen as the synchronous standby" , nil } ,
"sync_state" : { DISCARD , "Synchronous state of this standby server" , nil } ,
2016-09-08 13:24:42 +00:00
"slot_name" : { LABEL , "A unique, cluster-wide identifier for the replication slot" , nil } ,
"plugin" : { DISCARD , "The base name of the shared object containing the output plugin this logical slot is using, or null for physical slots" , nil } ,
"slot_type" : { DISCARD , "The slot type - physical or logical" , nil } ,
"datoid" : { DISCARD , "The OID of the database this slot is associated with, or null. Only logical slots have an associated database" , nil } ,
"database" : { DISCARD , "The name of the database this slot is associated with, or null. Only logical slots have an associated database" , nil } ,
"active" : { DISCARD , "True if this slot is currently actively being used" , nil } ,
"active_pid" : { DISCARD , "Process ID of a WAL sender process" , nil } ,
"xmin" : { DISCARD , "The oldest transaction that this slot needs the database to retain. VACUUM cannot remove tuples deleted by any later transaction" , nil } ,
"catalog_xmin" : { DISCARD , "The oldest transaction affecting the system catalogs that this slot needs the database to retain. VACUUM cannot remove catalog tuples deleted by any later transaction" , nil } ,
"restart_lsn" : { DISCARD , "The address (LSN) of oldest WAL which still might be required by the consumer of this slot and thus won't be automatically removed during checkpoints" , nil } ,
2016-01-06 19:19:41 +00:00
"pg_current_xlog_location" : { DISCARD , "pg_current_xlog_location" , nil } ,
"pg_xlog_location_diff" : { GAUGE , "Lag in bytes between master and slave" , nil } ,
2015-08-24 17:27:14 +00:00
} ,
2016-01-06 19:19:41 +00:00
"pg_stat_activity" : map [ string ] ColumnMapping {
"datname" : { LABEL , "Name of this database" , nil } ,
"state" : { LABEL , "connection state" , nil } ,
"count" : { GAUGE , "number of connections in this state" , nil } ,
"max_tx_duration" : { GAUGE , "max duration in seconds any active transaction has been running" , nil } ,
2015-08-24 17:27:14 +00:00
} ,
}
2016-01-06 19:19:41 +00:00
// Overriding queries for namespaces above.
var queryOverrides = map [ string ] string {
"pg_locks" : `
SELECT pg_database . datname , tmp . mode , COALESCE ( count , 0 ) as count FROM
( VALUES ( ' accesssharelock ' ) , ( ' rowsharelock ' ) , ( ' rowexclusivelock ' ) , ( ' shareupdateexclusivelock ' ) , ( ' sharelock ' ) , ( ' sharerowexclusivelock ' ) , ( ' exclusivelock ' ) , ( ' accessexclusivelock ' ) ) AS tmp ( mode ) CROSS JOIN pg_database
LEFT JOIN
( SELECT database , lower ( mode ) AS mode , count ( * ) AS count
FROM pg_locks WHERE database IS NOT NULL
GROUP BY database , lower ( mode )
) AS tmp2
ON tmp . mode = tmp2 . mode and pg_database . oid = tmp2 . database ORDER BY 1 ` ,
"pg_stat_replication" : `
2016-11-16 13:16:21 +00:00
SELECT * , pg_current_xlog_location ( ) , pg_xlog_location_diff ( pg_current_xlog_location ( ) , replay_location ) : : float FROM pg_stat_replication ` ,
2016-01-06 19:19:41 +00:00
"pg_stat_activity" : `
SELECT
pg_database . datname ,
tmp . state ,
COALESCE ( count , 0 ) as count ,
COALESCE ( max_tx_duration , 0 ) as max_tx_duration
FROM
( VALUES ( ' active ' ) , ( ' idle ' ) , ( ' idle in transaction ' ) , ( ' idle in transaction ( aborted ) ' ) , ( ' fastpath function call ' ) , ( ' disabled ' ) ) as tmp ( state ) CROSS JOIN pg_database
LEFT JOIN
( SELECT
datname ,
state ,
count ( * ) AS count ,
MAX ( EXTRACT ( EPOCH FROM now ( ) - xact_start ) ) : : float AS max_tx_duration
FROM pg_stat_activity GROUP BY datname , state ) as tmp2
ON tmp . state = tmp2 . state AND pg_database . datname = tmp2 . datname ` ,
}
2016-06-29 00:32:47 +00:00
// Add queries to the metricMaps and queryOverrides maps
2016-06-29 22:09:42 +00:00
func addQueries ( queriesPath string ) ( err error ) {
2016-06-29 00:32:47 +00:00
var extra map [ string ] interface { }
content , err := ioutil . ReadFile ( queriesPath )
if err != nil {
return err
}
err = yaml . Unmarshal ( content , & extra )
if err != nil {
return err
}
for metric , specs := range extra {
for key , value := range specs . ( map [ interface { } ] interface { } ) {
switch key . ( string ) {
case "query" :
query := value . ( string )
queryOverrides [ metric ] = query
case "metrics" :
for _ , c := range value . ( [ ] interface { } ) {
column := c . ( map [ interface { } ] interface { } )
for n , a := range column {
var cmap ColumnMapping
2016-08-20 16:32:08 +00:00
metric_map , ok := metricMaps [ metric ]
if ! ok {
metric_map = make ( map [ string ] ColumnMapping )
}
2016-06-29 00:32:47 +00:00
name := n . ( string )
for attr_key , attr_val := range a . ( map [ interface { } ] interface { } ) {
2016-08-20 16:32:08 +00:00
switch attr_key . ( string ) {
2016-06-29 00:32:47 +00:00
case "usage" :
2016-06-29 22:09:42 +00:00
usage , err := stringToColumnUsage ( attr_val . ( string ) )
2016-06-29 00:32:47 +00:00
if err != nil {
return err
}
cmap . usage = usage
case "description" :
cmap . description = attr_val . ( string )
}
}
cmap . mapping = nil
metric_map [ name ] = cmap
metricMaps [ metric ] = metric_map
}
}
}
}
}
return
}
2015-08-24 17:27:14 +00:00
// Turn the MetricMap column mapping into a prometheus descriptor mapping.
func makeDescMap ( metricMaps map [ string ] map [ string ] ColumnMapping ) map [ string ] MetricMapNamespace {
var metricMap = make ( map [ string ] MetricMapNamespace )
2016-01-06 19:19:41 +00:00
2015-08-24 17:27:14 +00:00
for namespace , mappings := range metricMaps {
thisMap := make ( map [ string ] MetricMap )
2016-01-06 19:19:41 +00:00
2015-08-24 17:27:14 +00:00
// Get the constant labels
var constLabels [ ] string
for columnName , columnMapping := range mappings {
if columnMapping . usage == LABEL {
2016-01-06 19:19:41 +00:00
constLabels = append ( constLabels , columnName )
2015-08-24 17:27:14 +00:00
}
}
2016-01-06 19:19:41 +00:00
2015-08-24 17:27:14 +00:00
for columnName , columnMapping := range mappings {
switch columnMapping . usage {
2016-01-06 19:19:41 +00:00
case DISCARD , LABEL :
thisMap [ columnName ] = MetricMap {
discard : true ,
conversion : func ( in interface { } ) ( float64 , bool ) {
return math . NaN ( ) , true
} ,
}
case COUNTER :
thisMap [ columnName ] = MetricMap {
vtype : prometheus . CounterValue ,
desc : prometheus . NewDesc ( fmt . Sprintf ( "%s_%s" , namespace , columnName ) , columnMapping . description , constLabels , nil ) ,
conversion : func ( in interface { } ) ( float64 , bool ) {
return dbToFloat64 ( in )
} ,
}
case GAUGE :
thisMap [ columnName ] = MetricMap {
vtype : prometheus . GaugeValue ,
desc : prometheus . NewDesc ( fmt . Sprintf ( "%s_%s" , namespace , columnName ) , columnMapping . description , constLabels , nil ) ,
conversion : func ( in interface { } ) ( float64 , bool ) {
return dbToFloat64 ( in )
} ,
}
case MAPPEDMETRIC :
thisMap [ columnName ] = MetricMap {
vtype : prometheus . GaugeValue ,
desc : prometheus . NewDesc ( fmt . Sprintf ( "%s_%s" , namespace , columnName ) , columnMapping . description , constLabels , nil ) ,
conversion : func ( in interface { } ) ( float64 , bool ) {
text , ok := in . ( string )
if ! ok {
return math . NaN ( ) , false
}
val , ok := columnMapping . mapping [ text ]
if ! ok {
return math . NaN ( ) , false
}
return val , true
} ,
}
case DURATION :
thisMap [ columnName ] = MetricMap {
vtype : prometheus . GaugeValue ,
desc : prometheus . NewDesc ( fmt . Sprintf ( "%s_%s_milliseconds" , namespace , columnName ) , columnMapping . description , constLabels , nil ) ,
conversion : func ( in interface { } ) ( float64 , bool ) {
var durationString string
switch t := in . ( type ) {
case [ ] byte :
durationString = string ( t )
case string :
durationString = t
default :
log . Errorln ( "DURATION conversion metric was not a string" )
return math . NaN ( ) , false
}
2016-05-06 01:07:27 +00:00
if durationString == "-1" {
return math . NaN ( ) , false
}
2016-01-06 19:19:41 +00:00
d , err := time . ParseDuration ( durationString )
if err != nil {
log . Errorln ( "Failed converting result to metric:" , columnName , in , err )
return math . NaN ( ) , false
}
return float64 ( d / time . Millisecond ) , true
} ,
}
2015-08-24 17:27:14 +00:00
}
}
2016-01-06 19:19:41 +00:00
metricMap [ namespace ] = MetricMapNamespace { constLabels , thisMap }
2015-08-24 17:27:14 +00:00
}
2016-01-06 19:19:41 +00:00
2015-08-24 17:27:14 +00:00
return metricMap
}
2016-06-29 22:09:42 +00:00
// convert a string to the corresponding ColumnUsage
func stringToColumnUsage ( s string ) ( u ColumnUsage , err error ) {
2016-08-20 16:32:08 +00:00
switch s {
2016-06-29 22:09:42 +00:00
case "DISCARD" :
u = DISCARD
case "LABEL" :
u = LABEL
case "COUNTER" :
u = COUNTER
case "GAUGE" :
u = GAUGE
case "MAPPEDMETRIC" :
u = MAPPEDMETRIC
case "DURATION" :
u = DURATION
default :
err = fmt . Errorf ( "wrong ColumnUsage given : %s" , s )
}
return
}
2015-08-27 18:26:52 +00:00
// Convert database.sql types to float64s for Prometheus consumption. Null types are mapped to NaN. string and []byte
// types are mapped as NaN and !ok
func dbToFloat64 ( t interface { } ) ( float64 , bool ) {
switch v := t . ( type ) {
case int64 :
return float64 ( v ) , true
case float64 :
return v , true
case time . Time :
return float64 ( v . Unix ( ) ) , true
2015-12-09 05:41:28 +00:00
case [ ] byte :
// Try and convert to string and then parse to a float64
strV := string ( v )
result , err := strconv . ParseFloat ( strV , 64 )
if err != nil {
return math . NaN ( ) , false
}
return result , true
2016-07-28 10:59:28 +00:00
case string :
result , err := strconv . ParseFloat ( v , 64 )
if err != nil {
2016-08-06 01:54:41 +00:00
log . Infoln ( "Could not parse string:" , err )
2016-07-28 10:59:28 +00:00
return math . NaN ( ) , false
}
return result , true
2015-08-27 18:26:52 +00:00
case nil :
return math . NaN ( ) , true
default :
return math . NaN ( ) , false
}
}
// Convert database.sql to string for Prometheus labels. Null types are mapped to empty strings.
func dbToString ( t interface { } ) ( string , bool ) {
switch v := t . ( type ) {
case int64 :
return fmt . Sprintf ( "%v" , v ) , true
case float64 :
return fmt . Sprintf ( "%v" , v ) , true
case time . Time :
return fmt . Sprintf ( "%v" , v . Unix ( ) ) , true
case nil :
return "" , true
case [ ] byte :
// Try and convert to string
return string ( v ) , true
case string :
return v , true
default :
return "" , false
}
}
2015-12-09 05:41:28 +00:00
// Exporter collects Postgres metrics. It implements prometheus.Collector.
2015-08-24 17:27:14 +00:00
type Exporter struct {
dsn string
duration , error prometheus . Gauge
totalScrapes prometheus . Counter
2016-01-06 19:19:41 +00:00
variableMap map [ string ] MetricMapNamespace
metricMap map [ string ] MetricMapNamespace
2015-08-24 17:27:14 +00:00
}
2016-01-06 19:19:41 +00:00
// NewExporter returns a new PostgreSQL exporter for the provided DSN.
2015-08-24 17:27:14 +00:00
func NewExporter ( dsn string ) * Exporter {
return & Exporter {
dsn : dsn ,
duration : prometheus . NewGauge ( prometheus . GaugeOpts {
Namespace : namespace ,
Subsystem : exporter ,
Name : "last_scrape_duration_seconds" ,
Help : "Duration of the last scrape of metrics from PostgresSQL." ,
} ) ,
totalScrapes : prometheus . NewCounter ( prometheus . CounterOpts {
Namespace : namespace ,
Subsystem : exporter ,
Name : "scrapes_total" ,
Help : "Total number of times PostgresSQL was scraped for metrics." ,
} ) ,
error : prometheus . NewGauge ( prometheus . GaugeOpts {
Namespace : namespace ,
Subsystem : exporter ,
Name : "last_scrape_error" ,
Help : "Whether the last scrape of metrics from PostgreSQL resulted in an error (1 for error, 0 for success)." ,
} ) ,
2016-01-06 19:19:41 +00:00
variableMap : makeDescMap ( variableMaps ) ,
metricMap : makeDescMap ( metricMaps ) ,
2015-08-24 17:27:14 +00:00
}
}
// Describe implements prometheus.Collector.
func ( e * Exporter ) Describe ( ch chan <- * prometheus . Desc ) {
// We cannot know in advance what metrics the exporter will generate
// from Postgres. So we use the poor man's describe method: Run a collect
// and send the descriptors of all the collected metrics. The problem
// here is that we need to connect to the Postgres DB. If it is currently
// unavailable, the descriptors will be incomplete. Since this is a
// stand-alone exporter and not used as a library within other code
// implementing additional metrics, the worst that can happen is that we
// don't detect inconsistent metrics created by this exporter
// itself. Also, a change in the monitored Postgres instance may change the
// exported metrics during the runtime of the exporter.
metricCh := make ( chan prometheus . Metric )
doneCh := make ( chan struct { } )
go func ( ) {
for m := range metricCh {
ch <- m . Desc ( )
}
close ( doneCh )
} ( )
e . Collect ( metricCh )
close ( metricCh )
<- doneCh
}
// Collect implements prometheus.Collector.
func ( e * Exporter ) Collect ( ch chan <- prometheus . Metric ) {
e . scrape ( ch )
ch <- e . duration
ch <- e . totalScrapes
ch <- e . error
}
func newDesc ( subsystem , name , help string ) * prometheus . Desc {
return prometheus . NewDesc (
prometheus . BuildFQName ( namespace , subsystem , name ) ,
help , nil , nil ,
)
}
2016-11-16 15:16:45 +00:00
// Query the SHOW variables from the query map
// TODO: make this more functional
2016-11-16 16:09:20 +00:00
func queryShowVariables ( ch chan <- prometheus . Metric , db * sql . DB , variableMap map [ string ] MetricMapNamespace ) [ ] error {
2015-12-09 05:41:28 +00:00
log . Debugln ( "Querying SHOW variables" )
2016-11-16 16:09:20 +00:00
nonFatalErrors := [ ] error { }
for _ , mapping := range variableMap {
2015-12-09 05:41:28 +00:00
for columnName , columnMapping := range mapping . columnMappings {
// Check for a discard request on this value
if columnMapping . discard {
continue
}
// Use SHOW to get the value
row := db . QueryRow ( fmt . Sprintf ( "SHOW %s;" , columnName ) )
2016-01-06 19:19:41 +00:00
var val interface { }
2015-12-09 05:41:28 +00:00
err := row . Scan ( & val )
if err != nil {
2016-11-16 16:09:20 +00:00
nonFatalErrors = append ( nonFatalErrors , errors . New ( fmt . Sprintln ( "Error scanning runtime variable:" , columnName , err ) ) )
2015-12-09 05:41:28 +00:00
continue
}
fval , ok := columnMapping . conversion ( val )
2016-01-06 19:19:41 +00:00
if ! ok {
2016-11-16 16:09:20 +00:00
nonFatalErrors = append ( nonFatalErrors , errors . New ( fmt . Sprintln ( "Unexpected error parsing column: " , namespace , columnName , val ) ) )
2015-12-09 05:41:28 +00:00
continue
}
ch <- prometheus . MustNewConstMetric ( columnMapping . desc , columnMapping . vtype , fval )
}
}
2016-11-16 16:09:20 +00:00
return nonFatalErrors
2016-11-16 15:16:45 +00:00
}
2015-12-09 05:41:28 +00:00
2016-11-16 16:09:20 +00:00
// Query within a namespace mapping and emit metrics. Returns fatal errors if
// the scrape fails, and a slice of errors if they were non-fatal.
func queryNamespaceMapping ( ch chan <- prometheus . Metric , db * sql . DB , namespace string , mapping MetricMapNamespace ) ( [ ] error , error ) {
2016-11-16 15:16:45 +00:00
query , er := queryOverrides [ namespace ]
if er == false {
query = fmt . Sprintf ( "SELECT * FROM %s;" , namespace )
}
2016-01-06 19:19:41 +00:00
2016-11-16 15:16:45 +00:00
// Don't fail on a bad scrape of one metric
rows , err := db . Query ( query )
if err != nil {
2016-11-16 16:09:20 +00:00
return [ ] error { } , errors . New ( fmt . Sprintln ( "Error running query on database: " , namespace , err ) )
2016-11-16 15:16:45 +00:00
}
defer rows . Close ( )
2016-01-06 19:19:41 +00:00
2016-11-16 15:16:45 +00:00
var columnNames [ ] string
columnNames , err = rows . Columns ( )
if err != nil {
2016-11-16 16:09:20 +00:00
return [ ] error { } , errors . New ( fmt . Sprintln ( "Error retrieving column list for: " , namespace , err ) )
2016-11-16 15:16:45 +00:00
}
2016-01-06 19:19:41 +00:00
2016-11-16 15:16:45 +00:00
// Make a lookup map for the column indices
var columnIdx = make ( map [ string ] int , len ( columnNames ) )
for i , n := range columnNames {
columnIdx [ n ] = i
}
2016-01-06 19:19:41 +00:00
2016-11-16 15:16:45 +00:00
var columnData = make ( [ ] interface { } , len ( columnNames ) )
var scanArgs = make ( [ ] interface { } , len ( columnNames ) )
for i := range columnData {
scanArgs [ i ] = & columnData [ i ]
}
2016-11-16 16:09:20 +00:00
nonfatalErrors := [ ] error { }
2016-11-16 15:16:45 +00:00
for rows . Next ( ) {
err = rows . Scan ( scanArgs ... )
if err != nil {
2016-11-16 16:09:20 +00:00
return [ ] error { } , errors . New ( fmt . Sprintln ( "Error retrieving rows:" , namespace , err ) )
2016-11-16 15:16:45 +00:00
}
2016-01-06 19:19:41 +00:00
2016-11-16 15:16:45 +00:00
// Get the label values for this row
var labels = make ( [ ] string , len ( mapping . labels ) )
for idx , columnName := range mapping . labels {
labels [ idx ] , _ = dbToString ( columnData [ columnIdx [ columnName ] ] )
}
// Loop over column names, and match to scan data. Unknown columns
// will be filled with an untyped metric number *if* they can be
// converted to float64s. NULLs are allowed and treated as NaN.
for idx , columnName := range columnNames {
if metricMapping , ok := mapping . columnMappings [ columnName ] ; ok {
// Is this a metricy metric?
if metricMapping . discard {
continue
2015-08-24 17:27:14 +00:00
}
2016-01-06 19:19:41 +00:00
2016-11-16 15:16:45 +00:00
value , ok := dbToFloat64 ( columnData [ idx ] )
if ! ok {
2016-11-16 16:09:20 +00:00
nonfatalErrors = append ( nonfatalErrors , errors . New ( fmt . Sprintln ( "Unexpected error parsing column: " , namespace , columnName , columnData [ idx ] ) ) )
2016-11-16 15:16:45 +00:00
continue
}
2015-08-27 20:58:54 +00:00
2016-11-16 15:16:45 +00:00
// Generate the metric
ch <- prometheus . MustNewConstMetric ( metricMapping . desc , metricMapping . vtype , value , labels ... )
} else {
// Unknown metric. Report as untyped if scan to float64 works, else note an error too.
desc := prometheus . NewDesc ( fmt . Sprintf ( "%s_%s" , namespace , columnName ) , fmt . Sprintf ( "Unknown metric from %s" , namespace ) , nil , nil )
// Its not an error to fail here, since the values are
// unexpected anyway.
value , ok := dbToFloat64 ( columnData [ idx ] )
if ! ok {
2016-11-16 16:09:20 +00:00
nonfatalErrors = append ( nonfatalErrors , errors . New ( fmt . Sprintln ( "Unparseable column type - discarding: " , namespace , columnName , err ) ) )
2016-11-16 15:16:45 +00:00
continue
2015-08-24 17:27:14 +00:00
}
2016-01-06 19:19:41 +00:00
2016-11-16 15:16:45 +00:00
ch <- prometheus . MustNewConstMetric ( desc , prometheus . UntypedValue , value , labels ... )
}
}
}
2016-11-16 16:09:20 +00:00
return nonfatalErrors , nil
}
// Iterate through all the namespace mappings in the exporter and run their
// queries.
func queryNamespaceMappings ( ch chan <- prometheus . Metric , db * sql . DB , metricMap map [ string ] MetricMapNamespace ) map [ string ] error {
// Return a map of namespace -> errors
namespaceErrors := make ( map [ string ] error )
for namespace , mapping := range metricMap {
log . Debugln ( "Querying namespace: " , namespace )
nonFatalErrors , err := queryNamespaceMapping ( ch , db , namespace , mapping )
// Serious error - a namespace disappeard
if err != nil {
namespaceErrors [ namespace ] = err
log . Infoln ( err )
}
// Non-serious errors - likely version or parsing problems.
if len ( nonFatalErrors ) > 0 {
for _ , err := range nonFatalErrors {
log . Infoln ( err . Error ( ) )
}
}
}
return namespaceErrors
2016-11-16 15:16:45 +00:00
}
2015-08-27 18:26:52 +00:00
2016-11-16 15:16:45 +00:00
func ( e * Exporter ) scrape ( ch chan <- prometheus . Metric ) {
defer func ( begun time . Time ) {
e . duration . Set ( time . Since ( begun ) . Seconds ( ) )
} ( time . Now ( ) )
2016-01-06 19:19:41 +00:00
2016-11-16 15:16:45 +00:00
e . error . Set ( 0 )
e . totalScrapes . Inc ( )
2016-01-06 19:19:41 +00:00
2016-11-16 15:16:45 +00:00
db , err := sql . Open ( "postgres" , e . dsn )
if err != nil {
log . Infoln ( "Error opening connection to database:" , err )
e . error . Set ( 1 )
return
}
defer db . Close ( )
2016-01-06 19:19:41 +00:00
2016-11-16 15:16:45 +00:00
log . Debugln ( "Querying Postgres Version" )
versionRow := db . QueryRow ( "SELECT version();" )
var versionString string
err = versionRow . Scan ( & versionString )
if err != nil {
log . Errorln ( "Error scanning version string:" , err )
e . error . Set ( 1 )
return
}
shortVersion := parseVersion ( versionString )
// Output the version as a special metric
versionDesc := prometheus . NewDesc ( fmt . Sprintf ( "%s_%s" , namespace , staticLabelName ) , "Version string as reported by postgres" , [ ] string { "version" , "short_version" } , nil )
ch <- prometheus . MustNewConstMetric ( versionDesc , prometheus . UntypedValue , 1 , versionString , shortVersion )
2016-01-06 19:19:41 +00:00
2016-11-16 15:16:45 +00:00
// Handle querying the show variables
2016-11-16 16:09:20 +00:00
nonFatalErrors := queryShowVariables ( ch , db , e . variableMap )
if len ( nonFatalErrors ) > 0 {
e . error . Set ( 1 )
}
2016-11-16 15:16:45 +00:00
2016-11-16 16:09:20 +00:00
errMap := queryNamespaceMappings ( ch , db , e . metricMap )
if len ( errMap ) > 0 {
e . error . Set ( 1 )
2015-08-24 17:27:14 +00:00
}
}
func main ( ) {
flag . Parse ( )
2016-06-29 00:32:47 +00:00
if * queriesPath != "" {
2016-06-29 22:09:42 +00:00
err := addQueries ( * queriesPath )
2016-06-29 00:32:47 +00:00
if err != nil {
log . Warnln ( "Unparseable queries file - discarding merge: " , * queriesPath , err )
}
}
2016-08-20 16:32:08 +00:00
if * onlyDumpMaps {
dumpMaps ( )
return
}
2015-08-24 17:27:14 +00:00
dsn := os . Getenv ( "DATA_SOURCE_NAME" )
if len ( dsn ) == 0 {
log . Fatal ( "couldn't find environment variable DATA_SOURCE_NAME" )
}
exporter := NewExporter ( dsn )
prometheus . MustRegister ( exporter )
http . Handle ( * metricPath , prometheus . Handler ( ) )
http . HandleFunc ( "/" , func ( w http . ResponseWriter , r * http . Request ) {
w . Write ( landingPage )
} )
log . Infof ( "Starting Server: %s" , * listenAddress )
log . Fatal ( http . ListenAndServe ( * listenAddress , nil ) )
}
