- Can I migrate my data from ActiveRecord storeconfigs?
- Can I migrate from an HSQL PuppetDB to PostgreSQL PuppetDB instance?
- The PuppetDB dashboard gives me a weird SSL error when I visit it. What gives?
- Does PuppetDB support Puppet apply?
- Why is PuppetDB written in Java?
- Which versions of Java are supported?
- Which databases are supported?
- Why does PDB crash on startup while upgrading?
- The PuppetDB daemon shuts down with a “Cannot assign requested address” error. What does this mean, and how do I fix it?
- Why is PuppetDB using so much CPU?
- My Puppet master is running slower since I enabled PuppetDB. How can I profile it?
- How can I improve command processing performance?
Can I migrate my data from ActiveRecord storeconfigs?
Yes, but you must use PuppetDB 3.x to do so. Please consult the PuppetDB 3.x documentiation for more details.
Can I migrate from an HSQL PuppetDB to PostgreSQL PuppetDB instance?
Yes, but you must use PuppetDB 3.x to do so. Please consult the Migrating Data for more information.
The PuppetDB dashboard gives me a weird SSL error when I visit it. What gives?
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.
Does PuppetDB support Puppet apply?
Yes. However, the setup is quite different from the normal master-based setup, so consult the documentation for more details.
Why is PuppetDB written in Java?
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.
Which versions of Java are supported?
The officially supported versions are OpenJDK and Oracle JDK 1.7 on Debian-derived distros and 1.8 on RHEL-derived distros. Other versions may work, and issues will be addressed on a best-effort basis, but support is not guaranteed.
Which databases are supported?
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.
Why does PDB crash on startup while upgrading?
If you’re upgrading from an older version of PDB that stored incoming commands in ActiveMQ, it’s possible that you have a currupted KahaDB. If so, you might see an exception that mentions the KahaDB libraries (org.apache.activemq.store.kahadb). For example:
java.io.EOFException at java.io.RandomAccessFile.readInt(RandomAccessFile.java:776) at org.apache.activemq.store.kahadb.disk.journal.DataFileAccessor.readRecord(DataFileAccessor.java:81)
This problem may be fixed by stopping PDB, moving the mq directory somewhere else, and restarting PDB, but any messages that were in the mq directory will be lost.
Note that this ActiveMQ message:
2016-02-29 15:20:53,571 WARN [o.a.a.b.BrokerService] Store limit is 102400 mb (current store usage is 1 mb). The data directory: /home/wyatt/work/puppetdb-mq-tmp/mq/localhost/KahaDB only has 71730 mb of usable space. - resetting to maximum available disk space: 71730 mb
is harmless and does not indicate an issue.
The PuppetDB daemon shuts down with a “Cannot assign requested address” error. What does this mean, and how do I fix it?
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.
Why is PuppetDB using so much CPU?
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.
My Puppet master is running slower since I enabled PuppetDB. How can I profile it?
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
master 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.
How can I improve command processing performance?
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 once. 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.