Skip to content

Latest commit

 

History

History
677 lines (413 loc) · 31.5 KB

configure.markdown

File metadata and controls

677 lines (413 loc) · 31.5 KB
title layout canonical
PuppetDB 2.3 » Configuration
default
/puppetdb/latest/configure.html

Summary

PuppetDB has three main groups of settings:

  • The init script's configuration file, which sets the JVM heap size and the location of PuppetDB's main config file
  • Logging settings, which go in the logback.xml file and can be changed without restarting PuppetDB
  • All other settings, which go in PuppetDB's configuration file(s) and take effect after the service is restarted

Init Script Config File

If you installed PuppetDB from packages or used the rake install installation method, an init script was created for PuppetDB. This script has its own configuration file, whose location varies by platform and by package:

OS and Package File
Redhat-like (open source) /etc/sysconfig/puppetdb
Redhat-like (PE) /etc/sysconfig/pe-puppetdb
Debian/Ubuntu (open source) /etc/default/puppetdb
Debian/Ubuntu (PE) /etc/default/pe-puppetdb

In this file, you can change the following settings:

  • JAVA_BIN --- The location of the Java binary.
  • JAVA_ARGS --- Command line options for the Java binary, most notably the -Xmx (max heap size) flag.
  • USER --- The user PuppetDB should be running as.
  • INSTALL_DIR --- The directory into which PuppetDB is installed.
  • CONFIG --- The location of the PuppetDB config file, which may be a single file or a directory of .ini files.

Configuring the Java Heap Size

To change the JVM heap size for PuppetDB, edit the init script config file by setting a new value for the -Xmx flag in the JAVA_ARGS variable.

For example, to cap PuppetDB at 192MB of memory:

JAVA_ARGS="-Xmx192m"

To use 1GB of memory:

JAVA_ARGS="-Xmx1g"

Configuring JMX Access

While all JMX metrics are exposed using the /metrics namespace, you can also expose direct JMX access using standard JVM means as documented here. This can be done using the JAVA_ARGS init script setting, similar to configuring the heap size.

For example, adding the following JVM options will open up a JMX socket on port 1099:

JAVA_ARGS="-Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.port=1099"

The logback Logging Config File

Logging is configured with a logback.xml file, whose location is defined with the logging-config setting. If you change the log settings while PuppetDB is running, it will apply the new settings without requiring a restart.

See the logback documentation for more information about logging options.

The PuppetDB Configuration File(s)

PuppetDB is configured using an INI-style config format with several [sections]. This is very similar to the format used by Puppet. All of the sections and settings described below belong in the PuppetDB config file(s).

Whenever you change PuppetDB's configuration settings, you must restart the service for the changes to take effect.

You can change the location of the main config file in the init script config file. This location can point to a single configuration file or a directory of .ini files. If you specify a directory (conf.d style), PuppetDB will merge the .ini files in alphabetical order.

If you've installed PuppetDB from a package, by default it will use the conf.d config style. The default config directory is /etc/puppetdb/conf.d (or /etc/puppetlabs/puppetdb/conf.d for Puppet Enterprise). If you're running from source, you may use the -c command-line argument to specify your config file or directory.

An example configuration file:

[global]
vardir = /var/lib/puppetdb
logging-config = /var/lib/puppetdb/logback.xml

[database]
classname = org.postgresql.Driver
subprotocol = postgresql
subname = //localhost:5432/puppetdb

[puppetdb]
certificate-whitelist = /path/to/file/containing/certnames
disable-update-checking = false

[jetty]
port = 8080

Playing Nice With the PuppetDB Module

If you installed PuppetDB with the puppetlabs-puppetdb module, PuppetDB's settings will be managed by Puppet. Most of the settings you care about can be configured with the module's class parameters; see the module's documentation for details.

If you do need to change rare settings that the module doesn't manage, you can do the following:

Create a new class in a new module (something like site::puppetdb::server::extra), declare any number of ini_setting resources as shown below, set the class to refresh the puppetdb::server class, and assign it to your PuppetDB server.

{% highlight ruby %} # Site-specific PuppetDB settings. Declare this class on any node that gets the puppetdb::server class. class site::puppetdb::server::extra {

  # Restart the PuppetDB service if settings change
  Class[site::puppetdb::server::extra] ~> Class[puppetdb::server]

  # Get PuppetDB confdir
  include puppetdb::params
  $confdir = $puppetdb::params::confdir

  # Set resource defaults
  Ini_setting {
    ensure  => present,
    require => Class['puppetdb::server::validate_db'],
  }

  ini_setting {'puppetdb-catalog-hash-conflict-debugging':
    path    => "${confdir}/global.ini",
    section => 'global',
    setting => 'catalog-hash-conflict-debugging',
    value   => 'true',
  }
}

{% endhighlight %}

[global] Settings

The [global] section is used to configure application-wide behavior.

vardir

This defines the parent directory for the MQ's data directory. Also, if a database isn't specified, the default database's files will be stored in <vardir>/db. The directory must exist and be writable by the PuppetDB user in order for the application to run.

logging-config

This describes the full path to a logback.xml file. Covering all the options available for configuring logback is outside the scope of this document; see the aforementioned link for exhaustive information.

If this setting isn't provided, PuppetDB defaults to logging at INFO level to standard out.

If you installed from packages, PuppetDB will use the logback.xml file in the /etc/puppetdb/ or /etc/puppetlabs/puppetdb directory. Otherwise, you can find an example file in the ext directory of the source.

You can edit the logging configuration file while PuppetDB is running, and it will automatically react to changes after a few seconds.

update-server

The URL to query when checking for newer versions; defaults to http://updates.puppetlabs.com/check-for-updates. Overriding this setting may be useful if your PuppetDB server is firewalled and can't make external HTTP requests, in which case you could configure a proxy server to send requests to the updates.puppetlabs.com URL and override this setting to point to your proxy server.

catalog-hash-conflict-debugging

When this is set to true, debugging information will be written to <vardir>/debug/catalog-hashes every time a catalog is received with a hash that is different than the previously received catalog for that host. Note that this should only be enabled when troubleshooting performance related issues with PuppetDB and the database server. This will output many files and could potentially slow down a production PuppetDB instance. See the Troubleshooting Low Catalog Duplication guide for more information on the outputted files and debugging this problem.

url-prefix

Deprecated: This setting will be removed in the future and url-prefixes should be configured using the web-router-service.

This optional setting may be used to mount the PuppetDB web application at a URL other than "/". This should not be necessary unless you intend to run additional web applications in the same server with your PuppetDB instance. NOTE: if you change this setting, you must also set the corresponding setting in your Puppet Master's puppetdb.conf file.

[puppetdb] Settings

The [puppetdb] section is used to configure PuppetDB application-specific behavior.

certificate-whitelist

Optional. This describes the path to a file that contains a list of certificate names, one per line. Incoming HTTPS requests will have their certificates validated against this list of names and only those with an exact matching entry will be allowed through. (For a puppet master, this compares against the value of the certname setting, rather than the dns_alt_names setting.)

If not supplied, PuppetDB uses standard HTTPS without any additional authorization. All HTTPS clients must still supply valid, verifiable SSL client certificates.

disable-update-checking

Optional. Setting this to true disables checking for updated versions of PuppetDB. Defaults to false.

[database] Settings

The [database] section configures PuppetDB's database settings.

PuppetDB can use either a built-in HSQLDB database or a PostgreSQL database. If no database information is supplied, an HSQLDB database at <vardir>/db will be used.

FAQ: Why no MySQL or Oracle support?

MySQL lacks several features that PuppetDB relies on; the most notable is recursive queries. We have no plans to ever support MySQL.

Depending on demand, Oracle support may be forthcoming in a future version of PuppetDB. This hasn't been decided yet.

Using Built-in HSQLDB

To use an HSQLDB database at the default <vardir>/db, you can simply remove all database settings. To configure the DB for a different location, put the following in the [database] section:

classname = org.hsqldb.jdbcDriver
subprotocol = hsqldb
subname = file:</PATH/TO/DB>;hsqldb.tx=mvcc;sql.syntax_pgs=true

Replace </PATH/TO/DB> with the filesystem location in which you'd like to persist the database.

Do not use the username or password settings.

Using PostgreSQL

Before using the PostgreSQL backend, you must set up a PostgreSQL server, ensure that it will accept incoming connections, create a user for PuppetDB to use when connecting, and create a database for PuppetDB. Completely configuring PostgreSQL is beyond the scope of this manual, but if you are logged in as root on a running Postgres server, you can create a user and database as follows:

$ sudo -u postgres sh
$ createuser -DRSP puppetdb
$ createdb -E UTF8 -O puppetdb puppetdb
$ exit

If you are running PostgreSQL 9.3 or above you should install the regexp optimized index extension pg_trgm. This may require installing the postgresql-contrib (or equivalent) package, depending on your distribution:

$ sudo -u postgres sh
$ psql puppetdb -c 'create extension pg_trgm'
$ exit

Next you will most likely need to modify the pg_hba.conf file to allow for md5 authentication from at least localhost. To locate the file you can either issue a locate pg_hba.conf command (if your distribution supports it) or consult your distribution's documentation for the PostgreSQL confdir.

The following example pg_hba.conf file allows md5 authentication from localhost for both IPv4 and IPv6 connections:

# TYPE  DATABASE   USER   CIDR-ADDRESS  METHOD
local   all        all                  md5
host    all        all    127.0.0.1/32  md5
host    all        all    ::1/128       md5

Restart PostgreSQL and ensure you can log in by running:

$ sudo service postgresql restart
$ psql -h localhost puppetdb puppetdb

To configure PuppetDB to use this database, put the following in the [database] section:

classname = org.postgresql.Driver
subprotocol = postgresql
subname = //<HOST>:<PORT>/<DATABASE>
username = <USERNAME>
password = <PASSWORD>

Replace <HOST> with the DB server's hostname. Replace <PORT> with the port on which PostgreSQL is listening. Replace <DATABASE> with the name of the database you've created for use with PuppetDB.

Using SSL With PostgreSQL

It's possible to use SSL to protect connections to the database. There are several extra steps and considerations when doing so; see the PostgreSQL SSL setup page for complete details.

The main difference in the config file is that you must be sure to add ?ssl=true to the subname setting:

subname = //<HOST>:<PORT>/<DATABASE>?ssl=true

gc-interval

This controls how often, in minutes, to compact the database. The compaction process reclaims space and deletes unnecessary rows. If not supplied, the default is every 60 minutes.

node-ttl

Auto-deactivate nodes that haven't seen any activity (no new catalogs, facts, etc) in the specified amount of time. You may specify the time as a string using any of the following suffixes:

`d`  - days
`h`  - hours
`m`  - minutes
`s`  - seconds
`ms` - milliseconds

So, e.g., a value of 30d would set the time-to-live to 30 days, and a value of 48h would set the time-to-live to 48 hours.

Nodes will be checked for staleness every gc-interval minutes. Manual deactivation will continue to work as always.

If unset, auto-deactivation of nodes is disabled.

node-purge-ttl

Automatically delete nodes that have been deactivated for the specified amount of time. This will also delete all facts, catalogs and reports for the node. This ttl may be specified the same way as node-ttl above.

If unset, auto-deletion of nodes is disabled.

report-ttl

Automatically delete reports that are older than the specified amount of time. You may specify the time as a string using any of the suffixes described in the node-ttl section above.

Outdated reports will be deleted during the database garbage collection, which runs every gc-interval minutes.

If unset, the default value is 14 days.

classname

This sets the JDBC class to use. Set this to:

  • org.hsqldb.jdbcDriver when using the embedded database
  • org.postgresql.Driver when using PostgreSQL

subprotocol

Set this to:

  • hsqldb when using the embedded database
  • postgresql when using PostgreSQL

subname

This describes where to find the database. Set this to:

  • file:</PATH/TO/DB>;hsqldb.tx=mvcc;sql.syntax_pgs=true when using the embedded database, replacing </PATH/TO/DB> with a local filesystem path
  • //<HOST>:<PORT>/<DATABASE> when using PostgreSQL, replacing <HOST> with the DB server's hostname, <PORT> with the port on which PostgreSQL is listening, and <DATABASE> with the name of the database
    • Append ?ssl=true to this if your PostgreSQL server is using SSL.

username

This is the username to use when connecting. Only used with PostgreSQL.

password

This is the password to use when connecting. Only used with PostgreSQL.

log-slow-statements

This sets the number of seconds before an SQL query is considered "slow." Slow SQL queries are logged as warnings, to assist in debugging and tuning. Note PuppetDB does not interrupt slow queries; it simply reports them after they complete.

The default value is 10 seconds. A value of 0 will disable logging of slow queries.

conn-max-age

The maximum time (in minutes), for a pooled connection to remain unused before it is closed off.

If not supplied, we default to 60 minutes.

conn-keep-alive

This sets the time (in minutes), for a connection to remain idle before sending a test query to the DB. This is useful to prevent a DB from timing out connections on its end.

If not supplied, we default to 45 minutes.

conn-lifetime

The maximum time (in minutes) a pooled connection should remain open. Any connections older than this setting will be closed off. Connections currently in use will not be affected until they are returned to the pool.

If not supplied, we won't terminate connections based on their age alone.

###statements-cache-size

This setting defines how many prepared statements are cached automatically. For a large amount of dynamic queries this number could be increased to increase performance, at the cost of memory consumption and database resources.

If not supplied, we default to 1000.

[read-database] Settings

The [read-database] section configures PuppetDB's read-database settings, useful when running a PostgreSQL Hot Standby cluster. Currently only configuring a Postgres read-database is supported. See the Postgres docs here for details on configuring the cluster. The [read-database] portion of the configuration is in addition to the [database] settings. If [read-database] is specified, [database] must also be specified.

To configure PuppetDB to use a read-only database from the cluster, put the following in the [read-database] section:

classname = org.postgresql.Driver
subprotocol = postgresql
subname = //<HOST>:<PORT>/<DATABASE>
username = <USERNAME>
password = <PASSWORD>

Replace <HOST> with the DB server's hostname. Replace <PORT> with the port on which PostgreSQL is listening. Replace <DATABASE> with the name of the database you've created for use with PuppetDB.

Using SSL With PostgreSQL

It's possible to use SSL to protect connections to the database. There are several extra steps and considerations when doing so; see the PostgreSQL SSL setup page for complete details.

The main difference in the config file is that you must be sure to add ?ssl=true to the subname setting:

subname = //<HOST>:<PORT>/<DATABASE>?ssl=true

classname

This sets the JDBC class to use. This should be org.postgresql.Driver.

subprotocol

Set this to postgresql.

subname

This describes where to find the database. Set this to:

  • //<HOST>:<PORT>/<DATABASE> when using PostgreSQL, replacing <HOST> with the DB server's hostname, <PORT> with the port on which PostgreSQL is listening, and <DATABASE> with the name of the database
    • Append ?ssl=true to this if your PostgreSQL server is using SSL.

username

This is the username to use when connecting.

password

This is the password to use when connecting.

log-slow-statements

This sets the number of seconds before an SQL query is considered "slow." Slow SQL queries are logged as warnings, to assist in debugging and tuning. Note PuppetDB does not interrupt slow queries; it simply reports them after they complete.

The default value is 10 seconds. A value of 0 will disable logging of slow queries.

conn-max-age

The maximum time (in minutes), for a pooled connection to remain unused before it is closed off.

If not supplied, we default to 60 minutes.

conn-keep-alive

This sets the time (in minutes), for a connection to remain idle before sending a test query to the DB. This is useful to prevent a DB from timing out connections on its end.

If not supplied, we default to 45 minutes.

conn-lifetime

The maximum time (in minutes) a pooled connection should remain open. Any connections older than this setting will be closed off. Connections currently in use will not be affected until they are returned to the pool.

If not supplied, we won't terminate connections based on their age alone.

[command-processing] Settings

The [command-processing] section configures the command-processing subsystem.

Every change to PuppetDB's data stores arrives via commands that are inserted into a message queue (MQ). Command processor threads pull items off of that queue, persisting those changes.

threads

This defines how many command processing threads to use. Each thread can process a single command at a time. The number of threads can be tuned based on what you see in the performance dashboard.

This setting defaults to half the number of cores in your system.

dlo-compression-threshold

This setting specifies the maximum duration to keep messages in the dead-letter office before archiving them. This process will check for compressible messages on startup and after every gc-interval, but will only perform the archive once per dlo-compression-threshold. The same format can be used as for the node-ttl setting above. If set to 0 seconds, this behavior will be disabled. The default value is 1 day.

store-usage

This setting sets the maximum amount of space in megabytes that PuppetDB's ActiveMQ can use for persistent message storage.

temp-usage

This setting sets the maximum amount of space in megabytes that PuppetDB's ActiveMQ can use for temporary message storage.

max-frame-size

This setting sets the maximum frame size for persisted activemq messages supplied in bytes. Default value is 209715200 (i.e 200 MB).

[jetty] (HTTP) Settings

The [jetty] section configures HTTP for PuppetDB.

Note: If you are using Puppet Enterprise and want to enable the PuppetDB dashboard from the PE console, refer to Changing PuppetDB's Parameters for more information. PE users should not edit jetty.ini.

host

This sets the IP interface to listen on for unencrypted HTTP traffic. If not supplied, we bind to localhost, which will reject connections from anywhere but the PuppetDB server itself. To listen on all available interfaces, use 0.0.0.0.

To avoid DNS resolution confusion, it is recommended that if you wish to set this to something other than localhost an IP address should be used instead of a hostname.

Note: Unencrypted HTTP is the only way to view the performance dashboard, since PuppetDB uses host verification for SSL. However, it can also be used to make any call to PuppetDB's API, including inserting exported resources and retrieving arbitrary data about your Puppet-managed nodes. If you enable cleartext HTTP, you MUST configure your firewall to protect unverified access to PuppetDB.

port

This sets what port to use for unencrypted HTTP traffic. If not supplied, we won't listen for unencrypted traffic at all.

max-threads

This sets the maximum number of threads assigned to responding to HTTP and HTTPS requests, effectively changing how many concurrent requests can be made at one time. Defaults to 50.

Note: Due to how our web-server (Jetty 7) behaves, this setting must be higher then the number of CPU's on your system or it will stop processing any HTTP requests.

ssl-host

This sets IP interface to listen on for encrypted HTTPS traffic. If not supplied, we bind to localhost. To listen on all available interfaces, use 0.0.0.0.

To avoid DNS resolution confusion, it is recommended that if you wish to set this to something other than localhost an IP address should be used instead of a hostname.

ssl-port

This sets the port to use for encrypted HTTPS traffic. If not supplied, we won't listen for encrypted traffic at all.

ssl-cert

This sets the path to the server certificate PEM file used by the PuppetDB web service for HTTPS. During the SSL handshake for a connection, certificates extracted from this file are presented to the client for the client's use in validating the server. This file may contain a single certificate or a chain of certificates ordered from the end certificate first to the most-root certificate last. For example, a certificate chain could contain:

  • An end certificate
  • An intermediate CA certificate with which the end certificate was issued
  • A root CA certificate with which the intermediate CA certificate was issued

In the PEM file, the end certificate should appear first, the intermediate CA certificate should appear second, and the root CA certificate should appear last.

If a chain is present, it is not required to be complete. If a path has been specified for the ssl-cert-chain setting, the server will construct the cert chain starting with the first certificate found in the ssl-cert PEM and followed by any certificates in the ssl-cert-chain PEM. In the latter case, any certificates in the ssl-cert PEM beyond the first one would be ignored.

Note: This setting overrides the alternate configuration settings keystore and key-password.

ssl-key

This sets the path to the private key PEM file that corresponds with the ssl-cert, it used by the PuppetDB web service for HTTPS.

Note: This setting overrides the alternate configuration settings keystore and key-password.

ssl-ca-cert

This sets the path to the CA certificate PEM file used for client authentication. Authorized clients must be signed by the CA that that corresponds to this certificate.

Note: This setting overrides the alternate configuration settings truststore and trust-password.

keystore

This sets the path to a Java keystore file containing the key and certificate to be used for HTTPS.

key-password

This sets the passphrase to use for unlocking the keystore file.

truststore

This describes the path to a Java keystore file containing the CA certificate(s) for your puppet infrastructure.

trust-password

This sets the passphrase to use for unlocking the truststore file.

cipher-suites

Optional. A comma-separated list of cryptographic ciphers to allow for incoming SSL connections. Valid names are listed in the official JDK cryptographic providers documentation; you'll need to use the all-caps cipher suite name.

If not supplied, PuppetDB uses the default cipher suites for your local system on JDK versions older than 1.7.0u6. On newer JDK versions, PuppetDB will use only non-DHE cipher suites.

ssl-protocols

Optional. A comma-separated list of protocols to allow for incoming SSL connections. Valid names are listed in the official JDK cryptographic protocol documentation; you'll need to use the names with verbatim capitalization. For example: TLSv1, TLSv1.1, TLSv1.2.

If not supplied, PuppetDB uses a default of TLSv1, TLSv1.1, TLSv1.2. By default SSLv3 is not included in that list due to known vulnerabilities. Users wanting to use SSLv3 need to specify it in their list explicitly.

ssl-crl-path

Optional. This describes a path to a Certificate Revocation List file. Incoming SSL connections will be rejected if the client certificate matches a revocation entry in the file.

ssl-cert-chain

This sets the path to a PEM with CA certificates for use in presenting a client with the server's chain of trust. Certs found in this PEM file are appended after the first certificate from the ssl-cert PEM in the construction of the certificate chain. This is an optional setting. The certificates in the ssl-cert-chain PEM file should be ordered from the least-root CA certificate first to the most-root CA certificate last. For example, a certificate chain could contain:

  • An end certificate
  • An intermediate CA certificate with which the end certificate was issued
  • A root CA certificate with which the intermediate CA certificate was issued

The end certificate should appear in the ssl-cert PEM file. In the ssl-cert-chain PEM file, the intermediate CA certificate should appear first and the root CA certificate should appear last.

The chain is not required to be complete.

Note: This setting overrides the alternate configuration settings keystore and key-password.

access-log-config

Optional. This is a path to an XML file containing configuration information for the Logback-access module. If present, a logger will be set up to log information about any HTTP requests Jetty receives according to the logging configuration, as long as the XML file pointed to exists and is valid. Information on configuring the Logback-access module is available here.

A configuration file may resemble the following:

<configuration debug="false">
  <appender name="FILE" class="ch.qos.logback.core.FileAppender">
    <file>./dev-resources/access.log</file>
      <encoder>
        <pattern>%h %l %u %user %date "%r" %s %b</pattern>
      </encoder>
    </appender>
    <appender-ref ref="FILE" />
</configuration>

This example configures a FileAppender that outputs to a file, access.log, in the dev-resources directory. It will log the remote host making the request, the log name, the remote user making the request, the date/time of the request, the URL and method of the request, the status of the response, and the size in bytes of the response.

graceful-shutdown-timeout

After receiving a shut down, this is the number of milliseconds the server will wait for in-flight requests to complete before actually shutting down. New requests will be blocked during this time. Defaults to 30000.

request-header-max-size

This sets the maximum size of an HTTP Request Header. If a header is sent that exceeds this value, Jetty will return an HTTP 413 Error response. This defaults to 8192 bytes, and only needs to be configured if an exceedingly large header is being sent in an HTTP Request.

[nrepl] Settings

The [nrepl] section configures remote runtime modification. For more detailed info, see Debugging with the Remote REPL.

Enabling a remote REPL allows you to manipulate the behavior of PuppetDB at runtime. This should only be done for debugging purposes, and is thus disabled by default. An example configuration stanza:

[nrepl]
type = nrepl
port = 8081
host = 127.0.0.1

enabled

Set to true to enable the REPL. Defaults to false.

port

The port to use for the REPL.

host

Specifies the host or IP address for the repl service to listen on. By default this is 127.0.0.1 only, as this is an insecure channel this is the only recommended setting for production environments. Although this is generally not recommended for production, you can listen on all interfaces, you can specify 0.0.0.0 for example.

:web-router-service Settings

The :web-router-service section is used to configure the routes at which applications running with your PuppetDB instance are mounted. This configuration section must be done in a .conf file (this is the Human-Optimized Config Object Notation format; a flexible superset of JSON defined by the typesafe config library). For more information on configuring the web-router-service see the trapperkeeper-webserver-jetty9 docs.

:puppetlabs.puppetdb.cli.services/puppetdb-service

This setting may be used to mount the PuppetDB web application at a URL other than "/". This should not be necessary unless you intend to run additional web applications in the same server with your PuppetDB instance. NOTE: if you change this setting, you must also set the corresponding setting in your Puppet Master's puppetdb.conf file.