relayd.conf(5) mentions the following:
Each table must contain at least one host address; multiple hosts are
separated by newline, comma, or whitespace.
So the addition of commas as separators should now cover all possible
cases.
Without this, hosts separated by commas in a table will show up with a
trailing comma in their host name and will probably not get their stats
correctly assigned when grabbing information from relayctl.
if a table is commented out, you currently get an entry for the hosts
that the entry contains but you'll get NaN for those hosts all the time
if they are not defined elsewhere.
to avoid that we anchor our match to the beginning of lines and permit
leading spaces since some folks might like to indent lines.
It is very common for PuppetDB installs to run on a different host than
the puppetmaster. In such cases, a certificate file is normally used to
establish an encrypted communication to the server.
The most common setup for this server certificate is to use a
certificate that was signed by the puppetmaster's CA, so one would want
to verify the server cert against this same CA (it should be present on
puppet clients).
Moreover, to ensure that *only* puppet clients can communicate with
PuppetDB, a pair of client certificat/key files are usually used to
authenticate clients.
PuppetDB requires a pretty high amount of memory allocated to it. If you
don't give enough memory space to the JVM, PuppetDB might crash when it
has fully used the available max memory. This plugin should be useful
for better tuning up your instance(s) of PuppetDB.
Currently, it's only compatible with PuppetDB 6.x since it's requesting
data from the API as described for that version:
https://puppet.com/docs/puppetdb/6.0/api/metrics/v1/mbeans.html#jvm-metrics
To add support for more versions, one would need to consult the same
documentation for different versions and to adapt URLs used and the code
that's obtaining the data from JSON if the layout of the responses is
different.
The pattern under which the data was saved was different than the pattern under the data was loaded on the next run. The reason being an undefined variable that evaluates to the empty string. The result was that the number of matching lines reported was for the whole logfile instead of the time since the last run.
