edoras one Operator Guide

1. Overview

This document describes the operational aspects of edoras one and its integration with other systems.

It is intended for system administrators responsible for configuring an edoras one installation.

2. Server Requirements

2.1. Java

The minimum Java version to run edoras one is 7. If you use Java 8 then you will need to add the following system property such that it starts up without any errors: javax.xml.accessExternalSchema=all.

2.2. Database

If you use MySQL or MSSQL you need to set a case sensitive collation such that edoras one has the same functionality as on any other database. We recommend the following collation for MySQL and MSSQL:

MySQL collation: utf8_bin
Execute the following query to change it on your database: ALTER DATABASE {DATABASE-NAME} CHARACTER SET utf8 COLLATE utf8_bin;
MSSQL collation: Latin1_General_CS_AS
Execute the following query to change it on your database: ALTER DATABASE {DATABASE-NAME} COLLATE Latin1_General_CS_AS;

2.3. Fonts

The edoras one server requires a number of Microsoft TrueType fonts to be available to run correctly. On many systems the required fonts will already be installed by default, but in some cases this may not be the case. If the fonts are not available, the result may be a long wait while the server attempts to find the fonts somewhere in the filesystem and it may not be possible to start the server at all.

On Linux systems (e.g. CentOS) the required fonts can be obtained by installing the msttcorefonts package.

3. Tenant configuration

A new tenant in edoras one is created by adding a tenant definition file to the tenant data folder and restarting the server. The tenant data may either be built into the edoras one WAR file (see the :developer-guide: for more information) or placed in an external folder. The tenant data location can be configured using the tenant.data.location property. To use an external folder, the value of this property should have a file: prefix, for example:

file:${edoras-one.home}/tenants

3.1. Overall structure

The tenant definition is a JSON object definition, something like the following:

{
    "id": "tenantId",
    "name": "acme",
    "adminUserLogin": "testAdmin1",
    "adminUserEmail": "testAdmin@test.com",
    "accounts": [
        {
            "name": "account",
            "domain": "test.com",
            "cssClass": "account-blue",
            "groups": [ "group" ],
            "users": [
                {
                    "displayName": "John Smith",
                    "firstName": "John",
                    "lastName": "Smith",
                    "login": "john.smith",
                    "email": "john.smith@email.es",
                    "language": "en",
                    "memberGroups": ["group"]
                }
            ]
        }
    ]
}

The following sections describe the available attributes for each part of the definition.

3.2. Tenant information

The top level of the JSON contains information about the tenant. The following attributes are supported:

Table 1. Tenant attributes
Attribute name	Description
accounts	a list of the initial accounts
adminUserLogin	admin user’s login
adminUserEmail	admin user’s email address
name	tenant name

3.3. Account information

An account entry defines an initial account within the tenant. The following attributes are supported:

Table 2. Account attributes
Attribute name	Description
domain	domain name (used to create user email addresses automatically)
cssClass	account css class added to the html
groups	a list of group names
name	account name
users	a list of the initial account users

For each account entry, an account will be created with the given name, groups and users.

The CSS class will be used when any user that belongs to the account is logged in. Then this CSS class will be applied to the app div (.edoras-one).

For "Cloud" solutions, there are some predefined account CSS classes:

Table 3. Account predefined "Cloud" css classes and colours
CSS Class	Main Colour	Main Dark Colour
account-brown
account-green
account-teal
account-blue
account-purple
account-pink
(empty/default)

Leave CSS class empty to turn off UI customisation for the Account and use default colours.

3.4. User information

A user entry defines an initial user within the account. The following attributes are supported:

Table 4. User attributes
Attribute name	Description
displayName	user’s display name
firstName	user’s first name
lastName	user’s last name
login	user’s login
email	user’s email address
language	user’s language
memberGroups	groups that the user belongs to

For each user entry, a user will be created with the given information.

If no email address is provided and the account domain is set, then an email address will be created from the domain and user’s first and last names.

When defining a user’s group membership, both the default edoras one group names (edoras one Modeler etc.) and the group names explicitly defined in the account may be used.

4. Integrating edoras one with other systems

4.1. Mail integration

4.1.1. Outgoing mail

To send outgoing mails, edoras one uses a org.springframework.mail.javamail.JavaMailSender instance.

A suitable bean definition is therefore required in the installation-specific Spring configuration. For debugging purposes in a local development environment, a simple logging implementation is provided that simply logs all outgoing mails to the server log (no mail is actually sent):

    <!-- during development, just log outgoing emails -->
    <bean id="mailSender" class="com.edorasware.cloud.core.mail.LoggingMailSender"/>

For real servers, a full bean configuration will be required:

    <bean name="mailSender" class="org.springframework.mail.javamail.JavaMailSenderImpl">
        <property name="defaultEncoding" value="UTF-8"/>
        <property name="host" value="smtp.gmail.com"/>
        <property name="port" value="465"/>
        <property name="username" value="${mail.smtp.username}"/>
        <property name="password" value="${mail.smtp.password}"/>
        <property name="javaMailProperties">
            <props>
                <prop key="mail.debug">${mail.debug}</prop>
                <prop key="mail.transport.protocol">smtp</prop>
                <prop key="mail.smtp.auth">true</prop>
                <prop key="mail.smtp.socketFactory.port">465</prop>
                <prop key="mail.smtp.socketFactory.class">javax.net.ssl.SSLSocketFactory</prop>
                <prop key="mail.smtp.socketFactory.fallback">false</prop>
                <prop key="mail.smtp.quitwait">false</prop>
            </props>
        </property>
    </bean>

In a typical installation where the bean configuration is stored on the server where it will be used, the property placeholders shown in this example can be directly replaced with the appropriate values.

For details about mail sender configuration please refer to the Spring documentation.

5. Standard properties

edoras one provides a number of configuration properties that can be used to control the behaviour of the application. These properties can be set in a number of different ways:

in a property file built in to the on-premise application (this can only be changed by a developer)
in an external property file: one.properties
as a system property, for example on the application command line

These are listed in order of increasing precedence, so a property value defined by the application can be overwritten by an external property file, for example.

Properties are used to configure individual settings within edoras one’s Spring configuration. As an on-premise project may replace some or all of this Spring configuration, the properties that may be used in any given installation may also vary. The remainder of this section assumes that the installation uses the standard edoras one Spring configuration files.

5.1. Locating the external property file

edoras one looks for external property file in the edoras one configuration folder. By default this is the folder .edoras-one in the user’s home directory. An alternative configuration folder can be selected by setting the system property edoras-one.home.

If no external property file can be found then a message to that effect will be produced by edoras one during the application startup.

5.2. Basic configuration properties

Table 5. Basic properties
Property	Default value	Description
`edoras-one.home`	`${user.home}/.edoras-one`	the location of the edoras one configuration folder. This can only be set as a system property.
`application.endpoint`	`http://localhost:8080/one`	the base URL where the application is running

5.2.1. The `application.endpoint` property

The application.endpoint is used in several features of edoras one including the automatic email sender which sends an assigned task URL or the linked resource filter which is used to map constant front end resources names to hashed/versioned versions of these. This endpoint needs to be set to the endpoint where the clients are able to access edoras one. If the endpoint is set to https://edoras-one.com then the clients need to access it through this URL.

There are two setup methods on how you are able to access edoras one: first directly through the servlet container and second via a proxy. If you use the first option everything should work properly when you set the correct application.endpoint, but when you use the second option (behind a proxy) then you need to set the application .endpoint property to the endpoint of the proxy and not the one of the proxied servlet container.

If you have several endpoints accessing the same edoras one instance, then you need to ensure that the application.endpoint is set to one of these endpoints which is working correctly, but keep in mind that all generated URL’s will have this endpoint as prefix.

5.3. Content manager properties

Table 6. Content manager properties
Property	Default value	Description
`content.enableVersioning`	`false`	when set to `true`, the content manager will keep all the versions of the content on every update
`content.conversionCacheDir`	`${edoras-one.home}/temp/converted-content`	the location for the on-disk content conversion cache

5.4. Query export properties

Table 7. Query export properties
Property	Default value	Description
`export.date.format`	`yyyy-MM-dd’T’HH:mm:ss.SSS’Z'`	the format to be used for exported dates (see SimpleDateFormat)
`export.date.timezone`	`UTC`	the timezone to be used for exported dates (see TimeZone)

5.5. Elasticsearch integration properties

The following properties can be used to configure the elasticsearch integration so that it can connect to a suitable elasticsearch cluster. When a property is not defined then the default value will be used. The settings currently in use are also shown in the server log on system startup.

Table 8. Elasticsearch integration properties
Property	Default value	Description
`elasticsearch.cluster.name`	`edorasone`	the elasticsearch cluster name
`elasticsearch.configuration`	(default configuration file)	a URI for the local elasticsearch configuration file (e.g. `file:${edoras-one.home}/elasticsearch.yml`)
`elasticsearch.root.dir`	`tmpdir`	the root directory for elasticsearch data storage (only needed if data is stored locally). The default is a 'magic' value that uses a folder in Java temporary storage (useful only for evaluation or testing).
`elasticsearch.type`	`local`	the elasticsearch integration type. See below for more information.
`elasticsearch.addresses`	`127.0.0.1:9300`	a comma-separated list of network addresses that should be used to connect to an external cluster

The following elasticsearch.type values are supported:

local: uses a full elasticsearch node inside the JVM (simple to set up, but only suitable for testing).
node: uses a storage-free elasticsearch node to participate in an external cluster. This may be more efficient that the transport connection type for some operations but may also require more JVM resources.
transport: communicates with an external cluster using the transport protocol.

5.6. Elasticsearch full index properties

The following properties can be used to configure the elasticsearch full index (for installations where this index configuration is used).

Table 9. Elasticsearch full index properties
Property	Default value	Description
`full.index.content.limit`	`-1`	blocks indexing of content larger than the given limit (can be used as a workaround for problems with indexing of very large content)
`full.index.name`	`edoras-objects`	the index name to be used by the full index (this must be the same for all edoras one nodes in a given cluster)

5.7. Encoding properties in property files

Sometime sensitive values need to be configured (such as user names or passwords) and storing these in plain text in the property files is a potential security risk. You can encode property values in a production property file so that the plain text value is not readable. Encoded values will be decoded when the property is loaded, so the original plaintext value is used for configuration.

As an example, we start with a plain text property:

sensitive.value=password:topSecret

We mark any sections of the property that should be encoded using the <ENCODE> delimiters. A property value may contain any combination of delimited encoded values and plain text:

sensitive.value=password:<ENCODE>topSecret<ENCODE>

This property definition (or the complete property file) is then passed through the command-line encoding tool (see below), resulting in an encoded equivalent marked with the <ENC> delimiters:

sensitive.value=password:<ENC>1TOKpX5hMQCd5mcSduoRxCofjybs6xgIEi4yo/zsW+E=<ENC>

This encoded value can then be used in the production property file. Encoded values in a property file will be decoded by edoras one when the property files are loaded, so the original property value can be used in the Spring configuration without exposing the plain text in the property file. A different encoding will also be used each time a value is encoded, so it is also not possible to identify encoded passwords using a known encoded value.

5.7.1. Running the command-line encoding tool

To run the command-line encoding tool you need to unpack the WAR file used to deploy the edoras one application. Next browse to the WEB-INF/lib folder of the extracted WAR file and run the command-line tool as follows:

java -cp edoras-one-server-core-1.6.0.jar:* com.edorasware.one.util.CommandLineParser -encode <value>

The value may either be a plain string to be encoded (e.g. the sensitive.value property value above) or the name of a complete property file. In both cases, the encoded text will be written to the standard output and can be either copied directly from the terminal window or redirected to a new file:

$ java -jar edoras-one-server-core-1.6.0.jar \
    -encode '<ENCODE>topSecret<ENCODE>'
<ENC>bnMeD/y2jbVzzw1OpLkTTo2qsO8obc4vwxmPzZJR3ao=<ENC>
$ java -jar edoras-one-server-core-1.6.0.jar \
    -encode /tmp/input.properties > /tmp/encoded.properties
$

Use " instead of ' on Windows

6. Tomcat configuration

6.1. Connector configuration

To allow to send UTF-8 characters in URIs (e.g. the search requests) we need to allow UTF-8 URI encoding in tomcat (server.xml).

    <!--
        Define a non-SSL HTTP/1.1 Connector on port 8080
        URIEncoding is set to UTF-8
    -->
    <Connector port="8080" protocol="HTTP/1.1"
               connectionTimeout="20000"
               redirectPort="8443"
      URIEncoding="UTF-8"
      />

    <!--
        Define an AJP 1.3 Connector on port 8009
        URIEncoding is set to UTF-8
    -->
    <Connector port="8009" protocol="AJP/1.3" redirectPort="8443" URIEncoding="UTF-8"/>

7. Clustering

To deploy edoras one on two or more nodes you will need the following prerequisites:

Apache HTTPD Server with mod_jk/mod_proxy or any other load balancing component (Apache HTTP Server, mod_jk, mod_proxy)
Two or more Apache Tomcat instances (or any other application server)
File content synchronization (like Gluster)
Redis 2.8 and greater (Redis)

NOTE

mod_jk is only needed if you use an Apache Tomcat application server. If you use another application server, you will need mod_proxy.

As a load balancer we use the Apache HTTPD Server which sends the request to the Apache Tomcat nodes as configured in the mod_jk configuration. The file content (like documents in edoras one) are stored on the file system which is synchronized beteween the nodes with the help of Gluster. The last part is the Redis server which acts as central point for the distributed caches and the user sessions.

The following diagram shows this setup graphically:

This guide will just show you how to configure edoras one to use Redis as distributed cache and as a distributed session store. Please refer to the following guides on how to install, configure these tools:

7.1. Configuration of edoras one

The following list shows the configuration options needed to run edoras one in a clustered environment:

Use the shared file system to store cached converted document content. Because converted document content cache is shared between the nodes, each node has to point to the shared directory where converted content is stored. Each node’s content.conversionCacheDir property has to point to the same shared directory for the all nodes. When you start the node in the cluster, evict the cache to keep nodes synchronized.
Use Redis as distributed cache and as a distributed session store. To do this you need to enable the cache-redis and session-redis Spring profiles. You can either add these profiles to the spring.profiles.active system property, or add it in the web.xml as spring.profiles.default context parameter. Next you need to configure the Redis connection. To do this add the following system properties with the appropriate values. .Redis connection properties

Property Description

Property	Description
`redis.hostname`	the hostname where the Redis server is running (defaults to: `localhost`)
`redis.port`	the port where the Redis server is listening to incoming connections (defaults to: `6379`)
`redis.password`	the password of the Redis server (default is empty)

redis.hostname

the hostname where the Redis server is running (defaults to: localhost)

redis.port

the port where the Redis server is listening to incoming connections (defaults to: 6379)

redis.password

the password of the Redis server (default is empty)

If you use the FilesystemContentProvider in a clustered environment, you first need to check if the shared file system supports the needed locking mechanisms to ensure data consistency. We provide a command-line lock check tool which creates a test file in the same folder and then tries to lock the file, write a line to the file and unlock it again. This will be executed as many times as specified in the given iterations parameter. Later in this section we also provide a bash script which executes this process several times to simulate concurrent access to this lock file which then tests the real use case. To run the command-line encoding tool you need to unpack the WAR file used to deploy the edoras one application. Next browse to the WEB-INF/lib folder of the extracted WAR file and run the command-line tool as follows:

java -cp edoras-one-server-core-1.6.0.jar:* com.edorasware.one.util.CommandLineParser -lockcheck <iterations>

You need to run the lock check test tool, and extract the WAR file, on the shared file system you set up before (see first point in the clustered configuration) as there are network file systems which do not properly support the locking. If there is any error while executing the lock check please contact our edorasware support.

The <iterations> parameter is used to set the amount of iterations as described above. This check simulates accessing the file with a single instance. You need to use the following bash script to simulate access of multiple processes on the same file which then checks if the file is properly locked and that there are no data inconsistencies. Please create a new file called lock-check.sh inside the WEB-INF/lib folder of the extracted WAR file on the shared file system with the following content:

#!/bin/bash

echo Starting $1 lock check processes ...
pids=""
result=0

# start the lock check java processes
for i in $(seq 1 $1);
do
    java -cp edoras-one-server-core-1.6.0.jar:* com.edorasware.one.util.CommandLineParser -lockcheck $2 &
    pids="$pids $!"
done

# wait for all the processes to finish
for pid in $pids; do
    wait $pid || let "result=1"
done

if [ "$result" == "1" ];
    then
       echo A lock check process failed.
       exit 1
fi

echo All lock check processes completed successfully!

Next make the file executable and then execute it with the following parameters:

./lock-check.sh 10 1000

This will then spawn 10 Java processes with the lock check tool running with 1000 iterations. When everything runs through, a success message is shown and this ensures that the FilesystemContentProvider is able to run inside the cluster.

After these configurations edoras one is configured to run in a clustered environment.

8. Process debugger (experimental)

NOTE: the process debugger is an experimental feature and should not be used in a production environment. Use the debugger at your own risk.

8.1. Enabling the debugger

To allow process debugging in a particular system you should perform the following steps as the tenant administrator:

import and deploy the ProcessDebugger.zip application (this App can be found in the release bundle)
create a Process Debugging Dashboard case based on the definition with the same name. There must be only one case with this name
share the Process Debugging Dashboard case with a group of users that should be allowed to use the debugger (e.g. edoras one Modeler)
enable the Spring profile debugger (e.g. using the spring.profiles.active JVM property in the intellij idea Run→Run…→Edit configuration…→{project run configuration}→VM options)
restart the server

8.2. Updating the debugger

To update the debugger when migrating an edoras one installation to a newer version, you need to perform the following steps as the tenant administrator:

update the process debugger App by import the ProcessDebugger.zip file from the new release
deploy the updated App
in the process debugger dashboard case model, use the Migrate action to update the dashboard case
without this step the dashboard will continue to use the old debugger