Export Prometheus metrics from SQL queries

query-exporter is a Prometheus exporter which allows collecting metrics from database queries, at specified time intervals.

It uses SQLAlchemy to connect to different database engines, including PostgreSQL, MySQL, Oracle and Microsoft SQL Server.

Each query can be run on multiple databases, and update multiple metrics.

The application is simply run as:

query-exporter config.yaml

where the passed configuration file contains the definitions of the databases to connect and queries to perform to update metrics.

Configuration file format

A sample configuration file for the application looks like this:

databases:
  db1:
    dsn: sqlite://
    connect-sql:
      - PRAGMA application_id = 123
      - PRAGMA auto_vacuum = 1
    labels:
      region: us1
      app: app1
  db2:
    dsn: sqlite://
    keep-connected: false
    labels:
      region: us2
      app: app1

metrics:
  metric1:
    type: gauge
    description: A sample gauge
  metric2:
    type: summary
    description: A sample summary
    labels: [l1, l2]
    expiration: 24h
  metric3:
    type: histogram
    description: A sample histogram
    buckets: [10, 20, 50, 100, 1000]
  metric4:
    type: enum
    description: A sample enum
    states: [foo, bar, baz]

queries:
  query1:
    interval: 5
    databases: [db1]
    metrics: [metric1]
    sql: SELECT random() / 1000000000000000 AS metric1
  query2:
    interval: 20
    timeout: 0.5
    databases: [db1, db2]
    metrics: [metric2, metric3]
    sql: |
      SELECT abs(random() / 1000000000000000) AS metric2,
             abs(random() / 10000000000000000) AS metric3,
             "value1" AS l1,
             "value2" AS l2
  query3:
    schedule: "*/5 * * * *"
    databases: [db2]
    metrics: [metric4]
    sql: |
      SELECT value FROM (
        SELECT "foo" AS metric4 UNION
        SELECT "bar" AS metric3 UNION
        SELECT "baz" AS metric4
      )
      ORDER BY random()
      LIMIT 1

databases section

This section contains defintions for databases to connect to. Key names are arbitrary and only used to reference databases in the queries section.

Each database defintions can have the following keys:

dsn:

database connection details.

It can be provided as a string in the following format:

dialect[+driver]://[username:password][@host:port]/database[?option=value&...]

(see SQLAlchemy documentation for details on available engines and options), or as key/value pairs:

dialect: <dialect>[+driver]
user: <username>
password: <password>
host: <host>
port: <port>
database: <database>
options:
  <key1>: <value1>
  <key2>: <value2>

All entries are optional, except dialect.

Note that in the string form, username, password and options need to be URL-encoded, whereas this is done automatically for the key/value form.

See database-specific options page for some extra details on database configuration options.

It's also possible to get the connection string indirectly from other sources:

• from an environment variable (e.g. $CONNECTION_STRING) by setting dsn to:

  env:CONNECTION_STRING
  

• from a file, containing only the DSN value, by setting dsn to:

  file:/path/to/file
  

These forms only support specifying the actual DNS in the string form.

connect-sql:

An optional list of queries to run right after database connection. This can be used to set up connection-wise parameters and configurations.

keep-connected:

whether to keep the connection open for the database between queries, or disconnect after each one. If not specified, defaults to true. Setting this option to false might be useful if queries on a database are run with very long interval, to avoid holding idle connections.

autocommit:

whether to set autocommit for the database connection. If not specified, defaults to true. This should only be changed to false if specific queries require it.

labels:

an optional mapping of label names and values to tag metrics collected from each database. When labels are used, all databases must define the same set of labels.

metrics section

This section contains Prometheus metrics definitions. Keys are used as metric names, and must therefore be valid metric identifiers.

Each metric definition can have the following keys:

type:

the type of the metric, must be specified. The following metric types are supported:

• counter: value is incremented with each result from queries
• enum: value is set with each result from queries
• gauge: value is set with each result from queries
• histogram: each result from queries is added to observations
• summary: each result from queries is added to observations

description:

an optional description of the metric.

labels:

an optional list of label names to apply to the metric.

If specified, queries updating the metric must return rows that include values for each label in addition to the metric value. Column names must match metric and labels names.

buckets:

for histogram metrics, a list of buckets for the metrics.

If not specified, default buckets are applied.

states:

for enum metrics, a list of string values for possible states.

Queries for updating the enum must return valid states.

expiration:

the amount of time after which a series for the metric is cleared if no new value is collected.

Last report times are tracked independently for each set of label values for the metric.

This can be useful for metric series that only last for a certain amount of time, to avoid an ever-increasing collection of series.

The value is interpreted as seconds if no suffix is specified; valid suffixes are s, m, h, d. Only integer values are accepted.

queries section

This section contains definitions for queries to perform. Key names are arbitrary and only used to identify queries in logs.

Each query definition can have the following keys:

databases:

the list of databases to run the query on.

Names must match those defined in the databases section.

Metrics are automatically tagged with the database label so that indipendent series are generated for each database that a query is run on.

interval:

the time interval at which the query is run.

The value is interpreted as seconds if no suffix is specified; valid suffixes are s, m, h, d. Only integer values are accepted.

If a value is specified for interval, a schedule can't be specified.

If no value is specified (or specified as null), the query is only executed upon HTTP requests.

metrics:

the list of metrics that the query updates.

Names must match those defined in the metrics section.

parameters:

an optional list of parameters sets to run the query with.

If a query is specified with parameters in its sql, it will be run once for every set of parameters specified in this list, for every interval.

Each parameter set must be a dictionary where keys must match parameters names from the query SQL (e.g. :param).

As an example:

query:
  databases: [db]
  metrics: [metric]
  sql: |
    SELECT COUNT(*) AS metric FROM table
    WHERE id > :param1 AND id < :param2
  parameters:
    - param1: 10
      param2: 20
    - param1: 30
      param2: 40

schedule:

a schedule for executing queries at specific times.

This is expressed as a Cron-like format string (e.g. */5 * * * * to run every five minutes).

If a value is specified for schedule, an interval can't be specified.

If no value is specified (or specified as null), the query is only executed upon HTTP requests.

sql:

the SQL text of the query.

The query must return columns with names that match those of the metrics defined in metrics, plus those of labels (if any) for all these metrics.

query:
  databases: [db]
  metrics: [metric1, metric2]
  sql: SELECT 10.0 AS metric1, 20.0 AS metric2

will update metric1 to 10.0 and metric2 to 20.0.

Note:

since : is used for parameter markers (see parameters above), literal single : at the beginning of a word must be escaped with backslash (e.g. SELECT '\:bar' FROM table). There's no need to escape when the colon occurs inside a word (e.g. SELECT 'foo:bar' FROM table).

timeout:

a value in seconds after which the query is timed out.

If specified, it must be a multiple of 0.1.

Metrics endpoint

The exporter listens on port 9560 providing the standard /metrics endpoint.

By default, the port is bound on localhost. Note that if the name resolves both IPv4 and IPv6 addressses, the exporter will bind on both.

For the configuration above, the endpoint would return something like this:

# HELP database_errors_total Number of database errors
# TYPE database_errors_total counter
# HELP queries_total Number of database queries
# TYPE queries_total counter
queries_total{app="app1",database="db1",query="query1",region="us1",status="success"} 50.0
queries_total{app="app1",database="db2",query="query2",region="us2",status="success"} 13.0
queries_total{app="app1",database="db1",query="query2",region="us1",status="success"} 13.0
queries_total{app="app1",database="db2",query="query3",region="us2",status="error"} 1.0
# HELP queries_created Number of database queries
# TYPE queries_created gauge
queries_created{app="app1",database="db1",query="query1",region="us1",status="success"} 1.5945442444463024e+09
queries_created{app="app1",database="db2",query="query2",region="us2",status="success"} 1.5945442444471517e+09
queries_created{app="app1",database="db1",query="query2",region="us1",status="success"} 1.5945442444477117e+09
queries_created{app="app1",database="db2",query="query3",region="us2",status="error"} 1.5945444000140696e+09
# HELP query_latency Query execution latency
# TYPE query_latency histogram
query_latency_bucket{app="app1",database="db1",le="0.005",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="0.01",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="0.025",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="0.05",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="0.075",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="0.1",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="0.25",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="0.5",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="0.75",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="1.0",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="2.5",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="5.0",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="7.5",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="10.0",query="query1",region="us1"} 50.0
query_latency_bucket{app="app1",database="db1",le="+Inf",query="query1",region="us1"} 50.0
query_latency_count{app="app1",database="db1",query="query1",region="us1"} 50.0
query_latency_sum{app="app1",database="db1",query="query1",region="us1"} 0.004666365042794496
query_latency_bucket{app="app1",database="db2",le="0.005",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="0.01",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="0.025",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="0.05",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="0.075",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="0.1",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="0.25",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="0.5",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="0.75",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="1.0",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="2.5",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="5.0",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="7.5",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="10.0",query="query2",region="us2"} 13.0
query_latency_bucket{app="app1",database="db2",le="+Inf",query="query2",region="us2"} 13.0
query_latency_count{app="app1",database="db2",query="query2",region="us2"} 13.0
query_latency_sum{app="app1",database="db2",query="query2",region="us2"} 0.012369773990940303
query_latency_bucket{app="app1",database="db1",le="0.005",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="0.01",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="0.025",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="0.05",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="0.075",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="0.1",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="0.25",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="0.5",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="0.75",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="1.0",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="2.5",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="5.0",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="7.5",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="10.0",query="query2",region="us1"} 13.0
query_latency_bucket{app="app1",database="db1",le="+Inf",query="query2",region="us1"} 13.0
query_latency_count{app="app1",database="db1",query="query2",region="us1"} 13.0
query_latency_sum{app="app1",database="db1",query="query2",region="us1"} 0.004745393933262676
# HELP query_latency_created Query execution latency
# TYPE query_latency_created gauge
query_latency_created{app="app1",database="db1",query="query1",region="us1"} 1.594544244446163e+09
query_latency_created{app="app1",database="db2",query="query2",region="us2"} 1.5945442444470239e+09
query_latency_created{app="app1",database="db1",query="query2",region="us1"} 1.594544244447551e+09
# HELP metric1 A sample gauge
# TYPE metric1 gauge
metric1{app="app1",database="db1",region="us1"} -3561.0
# HELP metric2 A sample summary
# TYPE metric2 summary
metric2_count{app="app1",database="db2",l1="value1",l2="value2",region="us2"} 13.0
metric2_sum{app="app1",database="db2",l1="value1",l2="value2",region="us2"} 58504.0
metric2_count{app="app1",database="db1",l1="value1",l2="value2",region="us1"} 13.0
metric2_sum{app="app1",database="db1",l1="value1",l2="value2",region="us1"} 75262.0
# HELP metric2_created A sample summary
# TYPE metric2_created gauge
metric2_created{app="app1",database="db2",l1="value1",l2="value2",region="us2"} 1.594544244446819e+09
metric2_created{app="app1",database="db1",l1="value1",l2="value2",region="us1"} 1.594544244447339e+09
# HELP metric3 A sample histogram
# TYPE metric3 histogram
metric3_bucket{app="app1",database="db2",le="10.0",region="us2"} 1.0
metric3_bucket{app="app1",database="db2",le="20.0",region="us2"} 1.0
metric3_bucket{app="app1",database="db2",le="50.0",region="us2"} 2.0
metric3_bucket{app="app1",database="db2",le="100.0",region="us2"} 3.0
metric3_bucket{app="app1",database="db2",le="1000.0",region="us2"} 13.0
metric3_bucket{app="app1",database="db2",le="+Inf",region="us2"} 13.0
metric3_count{app="app1",database="db2",region="us2"} 13.0
metric3_sum{app="app1",database="db2",region="us2"} 5016.0
metric3_bucket{app="app1",database="db1",le="10.0",region="us1"} 0.0
metric3_bucket{app="app1",database="db1",le="20.0",region="us1"} 0.0
metric3_bucket{app="app1",database="db1",le="50.0",region="us1"} 0.0
metric3_bucket{app="app1",database="db1",le="100.0",region="us1"} 0.0
metric3_bucket{app="app1",database="db1",le="1000.0",region="us1"} 13.0
metric3_bucket{app="app1",database="db1",le="+Inf",region="us1"} 13.0
metric3_count{app="app1",database="db1",region="us1"} 13.0
metric3_sum{app="app1",database="db1",region="us1"} 5358.0
# HELP metric3_created A sample histogram
# TYPE metric3_created gauge
metric3_created{app="app1",database="db2",region="us2"} 1.5945442444469101e+09
metric3_created{app="app1",database="db1",region="us1"} 1.5945442444474254e+09
# HELP metric4 A sample enum
# TYPE metric4 gauge
metric4{app="app1",database="db2",metric4="foo",region="us2"} 0.0
metric4{app="app1",database="db2",metric4="bar",region="us2"} 0.0
metric4{app="app1",database="db2",metric4="baz",region="us2"} 1.0

Builtin metrics

The exporter provides a few builtin metrics which can be useful to track query execution:

database_errors{database="db"}:

a counter used to report number of errors, per database.

queries{database="db",query="q",status="[success|error|timeout]"}:

a counter with number of executed queries, per database, query and status.

query_latency{database="db",query="q"}:

a histogram with query latencies, per database and query.

In addition, metrics for resources usage for the exporter procecss can be included by passing --process-stats in the command line.

Debugging / Logs

You can enable extended logging using the -L commandline switch. Possible log levels are CRITICAL, ERROR, WARNING, INFO, DEBUG.

Database engines

SQLAlchemy doesn't depend on specific Python database modules at installation. This means additional modules might need to be installed for engines in use. These can be installed as follows:

pip install SQLAlchemy[postgresql] SQLAlchemy[mysql] ...

based on which database engines are needed.

See supported databases for details.

Install from Snap

query-exporter can be installed from Snap Store on systems where Snaps are supported, via:

sudo snap install query-exporter

The snap provides both the query-exporter command and a deamon instance of the command, managed via a Systemd service.

To configure the daemon:

• create or edit /var/snap/query-exporter/current/config.yaml with the configuration
• run sudo snap restart query-exporter

The snap has support for connecting the following databases:

• PostgreSQL (postgresql://)
• MySQL (mysql://)
• SQLite (sqlite://)
• Microsoft SQL Server (mssql://)
• IBM DB2 (db2://) on supported architectures (x86_64, ppc64le and s390x)

Run in Docker

query-exporter can be run inside Docker containers, and is availble from the Docker Hub:

docker run -p 9560:9560/tcp -v "$CONFIG_FILE:/config.yaml" --rm -it adonato/query-exporter:latest

where $CONFIG_FILE is the absolute path of the configuration file to use. Note that the image expects the file to be available as /config.yaml in the container.

The image has support for connecting the following databases:

• PostgreSQL (postgresql://)
• MySQL (mysql://)
• SQLite (sqlite://)
• Microsoft SQL Server (mssql://)
• IBM DB2 (db2://)
• Oracle (oracle://)