-
Notifications
You must be signed in to change notification settings - Fork 36
Class[riak]: Parameters and Configuration
The module is centered around the riak class. The all-defaults configuration looks like this:
class { 'riak': }
It's 'anti-class' that you use when you want to remove Riak from a node, looks like this:
class { 'riak':
ensure => absent,
}
The module has some version hard-coded (at the time of writing it's 1.2.0), but you can upgrade by setting the version
parameter:
class { 'riak':
version => '1.3.0',
use_repos => false,
}
When you set use_repos = false, the module will download the rpm/deb rather than going through Basho's repositories.
The module has a hard-coded download url, but you can change it by setting the download
parameter:
class { 'riak':
download => 'https://internal-ci:8888/riak.custom-1.2.1-lol1.rpm',
download_hash => 'https:///internal-ci:8888/riak.custom-1.2.1.lol1.rpm.sha' #sha2-file
}
The hash-file should follow the format from Basho's own download formats.
class { 'riak':
source => [ "puppet:///modules/site/riak-app.config.${$environment}",
'puppet:///modules/site/riak-app.config' ]
}
This allows you to simply customize your App.config file. If you set this parameter, it overrides the default code-generation way of specifying the App.config file, and the parameter is mutually exclusive with template
.
class { 'riak':
template => 'puppet:///modules/site/App.config.erb'
}
In your erb-template, you can access the variable appcfg
, which is a hash. So you could e.g. write:
[{ riak_core, [ { mysetting, "<%= @appcfg[:riak_search][:enabled] ? 'yey' : 'of noes' %>" } ] }].
The template
parameter is mutually exclusive with the source
parameter. If you set this parameter, it overrides the default code-generation way of specifying the App.config file.
Sets the directory for logs for riak.
Sets the directory for erlang's virtual machine logs for riak. This is by default the same as log_dir
.
Where you wish to store the data of riak.
Default: true
Whether to restart the node (i.e. the riak
service) when you change the configuration.
Default: false
Whether to disable the riak service. If you set this to true, the service will progress into the stopped state. Different from absent
in that it doesn't remove riak from the node.
Default: false
Whether to disable the service from booting up. If also disable
is set, the service is also stopped right away.
Default: 4096
The max no. of open file descriptors, see the docs.
Default: riak/limits.conf.erb
The template to evaluate and use for /etc/security/limits.conf
.
You can modify App.config
by passing a hash to either cfg
or to vmargs_cfg
, which feeds vm.args
. Here's an example of adding a https binding with a custom certfile and keyfile.
class { 'riak':
cfg => {
riak_core => {
https => {
"__string_${$::ipaddress}" => 8443
},
ssl => {
certfile => "${etc_dir}/cert.pem",
keyfile => "${etc_dir}/key.pem"
}
}
}
}
The configuration options are equal to any that Riak accepts. Have a look at Basho's wiki for a complete listing, or at the rebar-file in their repo.
The prefix __string_
is one of many. They work like this:
Create an Erlang tuple:
['__tuple_', 46, $::ipaddress, 'whatever']
Create an Erlang list:
['__list_', 46, $::ipaddress, 'whatever']
Create an Erlang atom:
'__atom_certfile'
Create an Erlang binary string:
'__binary_mystringcomeshere'
All keys in a puppet hash are translated by default to Erlang atoms, because that's what the config uses most of the time. In the case that you need a key to be translated to a string, e.g. like in this code:
class { riak:
cfg => {
# ...
riak_core => {
http => {
"__string_${$::ipaddress}" => 8098
}
# ...
}
}
you simply prefix the key with __string_
.
All hashes will be sorted on keys, in each layer, to make sure that the files change as little as possible from puppet run to puppet run.