This page describes the enhanced/reloadable configuration feature, based on Apache Commons Configuration, which has been submitted for possible inclusion in DSpace 6.
- Ticket:
- PR: https://github.com/DSpace/DSpace/pull/1104
TESTERS NEEDED! While the basics of this functionality "work" (see PR above), this change literally changes how every configuration is read by DSpace (as Apache Commons Configuration has its own enhanced Property file syntax, see below for more on that).
This means it's likely that some specific features (especially optional ones) may need to have their configuration file/settings tweaked. I've done my best to already fix the configurations of out-of-the-box features, but have not yet tested all optional features.
Overview
In DSpace 5 or below, DSpace used it's own custom Property-based configuration scheme, along with a custom build.properties
which could tweak the build/compilation process in order to "override" some pre-selected configurations in the dspace.cfg
file. While this configuration scheme "worked" at a basic level, it required a lot of custom variable interpolation (i.e. filtering) to occur in both the Maven build process (mvn package
) and the Ant install process (ant fresh_install
or ant update
). The end result was that configuration files in your DSpace installation directory ([dspace.dir]
) contained the correct settings from your build.properties file, but all variables (${setting}
) were filled out. So, it was no longer possible to easily tweak certain key settings (like dspace.dir
or solr.server
) without having to either re-run the entire build process or make corrections to several files at once.
Enter Apache Commons Configuration.
The Enhanced Configuration Scheme feature uses Apache Commons Configuration (version 1.10) as the new configuration scheme for DSpace. This provides several key advantages over our old, custom configuration scheme:
- Apache Commons Configuration is a well-established Java library whose goal is to make configuration more flexible and easier to manage.
- It automatically interpolates all settings at runtime. This means we no longer need to replace variables (
${setting}
) within our configurations. They will be auto-determined at runtime based on the value of that variable within one of the configuration files For more on variable interpolation see its Basic Features documentation - It is a flexible configuration scheme. It can read configurations from several sources at once, including Properties files, XML config files and even database tables (see its Overview documentation). Currently, in the DSpace Enhanced Configuration Scheme we are still only using Properties files, similar to DSpace 5 and below. But, we now be able to easily move all or some configurations to XML configs or database config tables.
- The locations of the configuration sources can be easily customized by DSpace administrators in a new
config-definition.xml
file, which configures Apache Commons Configuration for DSpace. More on that below. - The
config-definition.xml
file itself is simply a "configuration definition" file as defined by Apache Commons Configuration. See the Configuration File Documentation for more details.
- The locations of the configuration sources can be easily customized by DSpace administrators in a new
- It allows for easy overriding of configuration values from other sources. How the overrides occur is up to how you've configured Apache Commons Configuration. For DSpace, we have a new
config-definition.xml
which defines the following override scheme (again, this can be easily tweaked for local needs):- If a setting is specified in Java System Properties (e.g.
-D[setting]=[value]
), it overrides the same setting found in any below location - If a setting is specified as an Environment Variable, it overrides the same setting found in any below location
- If a setting is specified in the new
local.cfg
configuration file, it overrides the default value in any below location - Default values for all settings are specified in the
dspace.cfg
and themodules/*.cfg
configuration files.
- If a setting is specified in Java System Properties (e.g.
- It supports enhanced Properties files. This means our
dspace.cfg
,local.cfg
and other configuration files can now immediately support some enhanced options, including:- The ability to easily include other configuration files via: "
include=[config-file-location]
" (See the end of the updateddspace.cfg
for examples) - The ability to provide lists of values to "array" configurations by specifying the setting multiple times (rather than creating a giant comma separated configuration spanning multiple lines). For example, enabling both LDAP and Password authentication can now be done via these two lines:
plugin.sequence.org.dspace.authenticate.AuthenticationMethod = org.dspace.authenticate.LDAPAuthentication
plugin.sequence.org.dspace.authenticate.AuthenticationMethod = org.dspace.authenticate.PasswordAuthentication
- For more information see the Commons Config Properties File documentation
- The ability to easily include other configuration files via: "
- More information/ features can also be found in the Apache Commons Configuration v1.10 User Guide
Building / Installing DSpace
With the Enhanced Configuration Scheme, the DSpace build process is slightly changed. The build.properties
file no longer exists and therefore has no effect on the build process.
Here's how the basics of building/installing DSpace:
- Download DSpace (as normal)
cd [dspace-source]
- Create your own initial local.cfg configuration file
cp local.cfg.EXAMPLE local.cfg
- The following fields MUST be specified in your local.cfg in order to install DSpace:
- dspace.dir
- database connection information (db.url, etc.)
- Build/Compile/Install as normal
mvn clean package
ant fresh_install
(orant update
)
- Once DSpace is installed, your local.cfg will be copied over to your
[dspace.dir]/config/
location. At that time you can optionally tweak it further (see local.cfg documentation below)
Unlike the old build.properties
, the new local.cfg
has NO effect on the Maven build process.
It is ONLY used by Ant to determine the location where DSpace should be installed/updated (using dspace.dir
), and also to initialize/update the database (using db.*
settings).
Many configuration names/keys have changed!
If you are upgrading from an earlier version of DSpace, you will need to be aware that many configuration names/keys have changed. Because Apache Commons Configuration allows for auto-overriding of configurations, all configuration names/keys in different *.cfg
files MUST be uniquely named (otherwise accidental, unintended overriding may occur).
In order to compensate for this, all modules/*.cfg
files had their configurations renamed to be prepended with the module name. As a basic example, all the configuration settings within the modules/oai.cfg
configuration now start with "oai.
".
Additionally, while the local.cfg
may look similar to the old build.properties
, many of its configurations have slightly different names. So, simply copying your build.properties into a local.cfg will NOT work.
This means that DSpace 5.x (or below) configurations are NOT compatible with the Enhanced Configuration Scheme. While you obviously can use your old configurations as a reference, you will need to start with fresh copy of all configuration files, and reapply any necessary configuration changes. However, as you'll see in the next section, you'll likely want to do that anyways in order to take full advantage of the new local.cfg
file.
local.cfg
The [dspace.dir]/config/local.cfg
file is the new way to customize your DSpace configuration based on your local needs.
There are a few key things to note about this configuration file:
- Any setting in your
local.cfg
will automatically OVERRIDE a setting of the same name in thedspace.cfg
or anymodules/*.cfg
file. This also means that you can copy ANY configuration (fromdspace.cfg
or anymodules/*.cfg
file) into your local.cfg
to specify a new value.- For example, specifying
dspace.url
inlocal.cfg
will override the default value ofdspace.url
indspace.cfg
. - Also, specifying
oai.solr.url
inlocal.cfg
will override the default value ofoai.solr.url
inconfig/modules/oai.cfg
- For example, specifying
- The
local.cfg
file is an Apache Commons Configuration Property file. For more information see the Commons Config Properties File documentation- This means it has enhanced features like the ability to include other config files (via "
include=
" statements).
- This means it has enhanced features like the ability to include other config files (via "
- As needed, you also are able to OVERRIDE settings in your
local.cfg
by specifying them as System Properties or Environment Variables.- For example, if you wanted to change your
dspace.dir
in development/staging environment, you could specify it as a System Property (e.g.-Ddspace.dir=[new-location]
). This new value will override any value in bothlocal.cfg
anddspace.cfg
.
- For example, if you wanted to change your
An example local.cfg is provided at [dspace-source]
/local.cfg.EXAMPLE. The example only provides a few key configurations which all DSpace sites are likely to need to customize. However, you may add (or remove) any other configuration to your local.cfg
to customize it as you see fit.
Link to local.cfg.EXAMPLE on Tim's DS-2654 branch: https://github.com/tdonohue/DSpace/blob/DS-2654-common-config/local.cfg.EXAMPLE
config-definition.xml
Link to config-definition.xml on Tim's DS-2654 branch: https://github.com/tdonohue/DSpace/blob/DS-2654-common-config/dspace/config/config-definition.xml
The [dspace.dir]/config/config-definition.xml
file defines the Apache Commons Configuration settings that DSpace utilizes by default. It is a valid "configuration definition" file as defined by Apache Commons Configuration. See the Configuration File Documentation for more details.
Link to config-definition.xml on Tim's DS-2654 branch: https://github.com/tdonohue/DSpace/blob/DS-2654-common-config/local.cfg.EXAMPLE
You are welcome to customize the config-definition.xml
to customize your local configuration scheme as you see fit. Any customizations to this file will require restarting your servlet container (e.g. Tomcat).
By default, the DSpace config-definition.xml
file defines the following configuration:
- All DSpace configurations are loaded via Properties files
- Note: Apache Commons Configuration does support other configuration sources such as XML configurations or database configurations, see its Overview documentation)
- Configuration Files/Sources: By default, only two configuration files are loaded into Apache Commons Configuration:
local.cfg
(see documentation on local.cfg above)dspace.cfg
(NOTE: however that allmodules/*.cfg
are loaded bydspace.cfg
via "include=
" statements at the end of that configuration file)
- Configuration Override Scheme: The configuration override scheme is defined as follows. Configurations specified in earlier locations will automatically override any later values:
- System Properties (-D[setting]=[value]) override all other options
- Environment Variables
local.cfg
dspace.cfg
(and allmodules/*.cfg
files) contain the default values for all settings
- Configuration Auto-Reload: By default, all configuration files are automatically checked each minute for changes. If they have changed, they are automatically reloaded.
Configuration Reloading and Caching
As noted above, by default, DSpace will now automatically reload any modified configuration file (local.cfg
, dspace.cfg
or modules/*.cfg
) within one minute.
While the new values are immediately available within the DSpace ConfigurationService, some configurations may still be "cached" within UI-specific code. This often occurs when a UI (or API) loads a configuration value into a static
variable, or otherwise implements/provides its own object caching mechanism.
The Enhanced Configuration Scheme codebase does NOT attempt to correct all these instances of caching within UIs or APIs. This would require individual configurations to be tested and any caching mechanisms to be removed.
FAQs
Can I have different local.cfg files for different environments (e.g. development/testing/staging/production)?
Yes and No. By default, DSpace does NOT allow you to have multiple local.cfg
files (one per environment). However, with some minimal tweaks to your configuration scheme, you likely (untested) could achieve this in one of two ways:
Change your
config-definition.xml
to use a system property (of your choice) instead of the hardcoded name "local.cfg". The Configuration Definition file itself does allow for variables to be included, but they must be specified in a previous configuration source (in that config-definition.xml) or via a system property. See the Configuration File Documentation for more details. So, you could simply change your config-definition.xml to use a "dspace.env
" system property, and pass "-Ddspace.env=dev
" to have it use a[dspace.dir]/config/dev.cfg
:<!-- Change local.cfg to be ${dspace.env} in your config-definition.xml --> <properties fileName="${dspace.env}.cfg" throwExceptionOnMissing="false" config-name="local" config-optional="true"> ... </properties> <!-- OPTIONALLY: If you wanted to have some default local configs shared among *all* environments, you could add a NEW "properties" file to always load those defaults. In this example, default.cfg would be loaded for ALL environments. Configs in the environment-specific ${dspace.env}.cfg would override default.cfg, and both would override dspace.cfg (and other *.cfg). --> <properties fileName="default.cfg" throwExceptionOnMissing="false" config-name="default" config-optional="true"> ... </properties>
Alternatively, you could use the "
include=
" option (of Apache Commons Configuration Properties Files) within yourlocal.cfg
file to load a different configuration file, again based on a setting specified as a system property. For example, yourlocal.cfg
file would ONLY consist of "include=
" statement(s), which would load whichever configuration file was specified as the "dspace.env" system property:# This is the ENTIRE local.cfg (all settings would instead be located in environment-specific config files) # Its job is just to load up the configuration for the environment specified by "dspace.env" # For example, -Ddspace.env=dev would load [dspace.dir]/config/dev.cfg # and -Ddspace.env=prod would load [dspace.dir]/config/prod.cfg # Load the environment-specific file include = ${dspace.env}.cfg # OPTIONALLY: If you wanted to have some default local configs shared among *all* environments, you could add # a second "include=" statement to always load those defaults from a file of your choice. In this example, # a default.cfg would be loaded for ALL environments. Configs in the environment-specific ${dspace.env}.cfg # would override default.cfg, and both would override dspace.cfg (and other *.cfg). include = default.cfg
While the above examples both use a property named ${dspace.env}
, you can use whatever property you want. The name itself doesn't matter. Additionally, both show examples of using a "default.cfg
" to specify properties which are shared between several environments. This file can also be named whatever you want. Just tweak the name(s) in the examples above to meet your local needs.
The option you choose above would likely depend on your own local practices/needs. Either of these options should work, provided that you place your environment-specific configuration files within the [dspace.dir]/config
directory alongside the local.cfg
file.
Advanced Topics
Configuration Interpolation
This is less important to normal users of DSpace, but may be of high interest to developers and some system administrators.
It's important to be aware of the fact that variables within the following types of configurations are now AUTOMATICALLY interpolated at runtime using Apache Commons Configuration (and our ConfigurationService). This means that variables (${setting}) are no longer filtered by Maven or Ant for any of the following configuration types:
- Configuration files (namely
local.cfg
,dspace.cfg
andmodules/*.cfg
) - Log4j settings (namely
log4j.properties
) - Spring XML configs (namely
[dspace.dir]/config/spring/api/*.xml
)
There is only one remaining file type which still requires its configurations/settings to be filtered/interpolated manually:
- All
web.xml
files unfortunately still need to have their${dspace.dir}
variable filtered (by Ant). This is because thedspace.dir
context parameter in theseweb.xml
files is used to initialize the DSpace Kernel (and tell the webapp where the DSpace home directory is). Unfortunately, there's no way to interpolate this value at runtime as thedspace.dir
value does not exist until the Kernel and the ConfigurationService have initialized.- The only way we'd get around this problem would be to REQUIRE a
dspace.dir
ALWAYS be specified to the servlet container (as a Context parameter and/or system property). - In other words, the DSpace webapps cannot function/initialize without a
dspace.dir
. We either need to filter a value for it (during ant update/fresh_install), or we need to REQUIRE that it be specified by other means.
- The only way we'd get around this problem would be to REQUIRE a
Java API Changes
ConfigurationManager vs ConfigurationService
In the DSpace 5 Java API, we had two types of Configuration objects: org.dspace.coreConfigurationManager
and org.dspace.services.ConfigurationService
.
While the the ConfigurationManager
still exists in the API (and is still called by some areas of the codebase), it is now a "wrapper" object. It simply wraps calls to the configured ConfigurationService
.
As before, the default ConfigurationService is the org.dspace.servicemanager.config.DSpaceConfigurationService
(in dspace-services).
The DSpaceConfigurationService
has been updated/enhanced to utilize Apache Commons Configuration, and to better align its methods with the old ConfigurationManager
class. It also has added a new reloadConfig()
method which can be called on demand to automatically reload all configurations.
PluginManager vs PluginService
In DSpace 5, the org.dspace.core.PluginManager
class managed all DSpace "plugin" definitions (i.e. plugin.*
settings in dspace.cfg
). (SIDENOTE: these DSpace "plugin" definitions are simply Java interfaces, which are then mapped to classes which implement that plugin interface).
While this concept still exists, the PluginManager
itself has been entirely replaced by a new org.dspace.core.service.PluginService.
The default PluginService is a new org.dspace.core.LegacyPluginServiceImpl
class, which implements the functionality of the old PluginManager
.