Frequently asked questions

Yes, but you must use PuppetDB 3.x to do so. Please consult the PuppetDB 3.x documentiation for more details.

Yes, but you must use PuppetDB 3.x to do so. Please consult the Migrating Data for more information.

There are two common error cases with the dashboard:

• You're trying to talk over plain text (8080) and PuppetDB's not listening.

By default, PuppetDB only listens for plain text connections on localhost, for security reasons. In order to talk to it this way, you'll need to either forward the plain text port or change the interface PuppetDB binds on to one that is accessible from the outside world. In the latter case, you'll want to use some other means to secure PuppetDB (for instance, by restricting which hosts are allowed to talk to PuppetDB through a firewall).

• You're trying to talk over SSL and nobody trusts anybody else.

Because PuppetDB uses the certificate authority (CA) of your Puppet infrastructure, and a certificate signed by it, PuppetDB doesn't trust your browser, and your browser doesn't trust PuppetDB. In this case, you'll need to give your browser a certificate signed by your Puppet CA. Support for client certificates is varied between browsers, so it's preferred to connect over plain text, as outlined above.

Yes. However, the setup is quite different from the normal Puppet Server based setup, so consult the documentation for more details.

Actually, PuppetDB isn't written in Java at all! It's written in a language called Clojure, which is a dialect of Lisp that runs on the Java Virtual Machine (JVM). Several other languages were prototyped, including Ruby and JRuby, but they lacked the necessary performance. We chose to use a JVM language because of its excellent libraries and high performance. Of the available JVM languages, we used Clojure because of its expressiveness, performance, and our team's previous experience with the language.

JDK 11 is officially supported. Other versions may work, and issues will be addressed on a best-effort basis, but support is not guaranteed.

PostgreSQL is the recommended database for production use.

As with our choice of language, we prototyped several databases before settling on PostgreSQL. These included Neo4j, Riak, and MySQL with ActiveRecord in Ruby. We have no plans to support any other databases, including MySQL, which lacks important features such as array columns and recursive queries.

The expected behavior is a warning when the pg_trgm extension is not installed. If you are seeing a message that asks you to run CREATE EXTENSION pg_trgm; then it's not erroring, but we do suggest you install the pg_trgm extension. The error can be seen in your PostgreSQL log and the PuppetDB log and looks like the output below.

PostgreSQL:

  < 2016-08-10 14:03:04.523 PDT >ERROR: could not access file "$libdir/pg_trgm": No such file or directory
  < 2016-08-10 14:03:04.523 PDT >STATEMENT: CREATE INDEX fact_values_string_trgm ON fact_values USING gin (value_string gin_trgm_ops)

PuppetDB:

  2018-08-01 18:32:31,433 INFO  [async-dispatch-2] [p.p.s.migrate] Creating additional index `fact_paths_path_trgm`
  2018-08-01 18:32:31,513 ERROR [async-dispatch-2] [p.t.internal] Error during service start!!!
  java.sql.BatchUpdateException: Batch entry 0 CREATE INDEX fact_paths_path_trgm ON fact_paths USING gist (path gist_trgm_ops) was aborted.

This error occurs when the database believes that pg_trgm has been installed, but for some reason the extension has been uninstalled or removed. Ensure the PostgreSQL extension pg_trgm has been installed. Depending on your operating system, you may be able to install this extension by installing the postgresql-contrib package.

FAILED org.eclipse.jetty.server.Server@6b2c636d: java.net.BindException: Cannot assign requested address
java.net.BindException: Cannot assign requested address

PuppetDB will error with this message if the IP address associated with the ssl-host parameter in the jetty.ini isn't linked to a known interface or resolvable.

There are numerous possible contributing factors to high CPU usage by PuppetDB, both on the application server and (if different) the database. Examples include the total number of nodes managed by Puppet, the frequency of the agent runs, and the number of changes to the nodes on each run. For more information on possible causes and ways to mitigate them, refer to the support and troubleshooting guide.

Puppet 3.x introduced a new profiling capability that we leveraged in the puppetdb-termini client code. By simply adding profile=true to your puppet.conf, you can enable detailed profiling of all aspects of Puppet, including puppetdb-termini. For this to work, you must enable debugging on your Puppet Server instance as well.

Note: We encourage all users to use common sense when working with profiling mechanisms. Using these tools will add more load, which can increase speed problems in a limited capacity environment. Enabling profiling in production environments should only be done with care and for a very short period of time.

To enable easy searching, all PuppetDB profiling events are prefixed with PuppetDB:. This information is also helpful to our developers, so feel free to include these details when reporting issues with PuppetDB.

When PuppetDB is running in a "steady state", it should have a very low queue depth (ideally 0). Something like a database outage can cause a temporary spike in queue depth. Having a queue depth without an outage or other significant event likely means that PuppetDB can't keep up with the work that is being enqueued. This is a good indication that some performance tuning needs to take place. There are several areas to consider when performance tuning. PuppetDB is sensitive to PostgreSQL performance issues, so usually that is a good place to start. Assuming that the PostgreSQL instance isn't under a heavy load, the focus can shift to tuning PuppetDB itself.

The threads setting indicates how many commands can be processed concurrently. If PuppetDB is consuming too many resources on a shared system, this number can be reduced. For servers that are dedicated PuppetDB instances, setting this value to the number of logical cores could significantly improve command throughput. Increasing the number of threads should also be paired with increasing the amount of memory allocated to PuppetDB.

The concurrent-writes setting indicates how many threads can write to the disk at one time. Faster enqueuing will result in faster puppet runs as the PuppetDB terminus enqueues the message as part of the puppet run. The impact of this setting is heavily related to disk performance on the system. On a system with an SSD, this setting will have very little impact on performance or load on the system. On a system with a spinning disk, this setting can heavily impact load average and command throughput. Having this setting higher than the default (i.e. 16 or 32) could result in faster enqueuing, but will also result in a significant spike in load average as the kernel will have I/O write requests "backed up". Changing this setting to lower than the default should reduce the load on the system but will reduce the throughput on the PuppetDB instance. That could potentially increase the time it takes to enqueue a command and thus slow the puppet runs.