Installing EPrints on RHEL/Fedora/CentOS
Contents
- 1 Requirements
- 2 Installing EPrints 3.4.x from Source
- 3 Non-critical Dependencies
- 4 Next steps
- 5 Installing using RPM package
- 6 Operating System Version Specific Requirements
- 6.1 RHEL / CentOS 7
- 6.2 RHEL / CentOS / Rocky 8
- 6.3 RHEL / CentOS / Rocky 9
- 6.3.1 elinks package is no longer available
- 6.3.2 perl-IO-String package is no longer available
- 6.3.3 Sub-dependency packages perl-MIME-Types and perl-Digest-SHA1 are no longer available
- 6.3.4 perl-Apache-DBI and Pod::LaTeX packages is no longer available
- 6.3.5 Compatibility issues with XML::LibXSLT
- 6.3.6 MySQL root user cannot be used
- 7 Troubleshooting
Requirements
- RedHat Enterprise (RHEL) / CentOS 7, 8 or 9 (or Fedora 25+).
- Pre-install MySQL server and client and make sure the former is enabled and running.
- RHEL / Fedora:
yum install mysql-server mysql mysql-devel systemctl enable --now mysqld
- CentOS / Rocky Linux:
yum install mariadb-server mariadb mariadb-devel systemctl enable --now mariadb
Installing EPrints 3.4.x from Source
Ideally you will install from source from EPrints Core Git repository. Installing from source is now the preferred way to install production repositories, having previously proved to be a useful installation method for those wishing to develop EPrints code or a plugin or Bazaar package too.
First, install the EPEL YUM package repository and run yum repolist to ensure the GPG key is installed:
yum install epel-release yum repolist
Now, install all the dependencies that would be installed through the eprints RPM.
yum install libxml2 libxslt httpd mod_perl perl-Apache-DBI perl-DBI perl-DBD-MySQL perl-IO-Socket-SSL \ perl-Time-HiRes perl-CGI perl-Digest-MD5 perl-Digest-SHA perl-Digest-SHA1 perl-JSON perl-XML-LibXML \ perl-XML-LibXSLT perl-XML-SAX perl-MIME-Lite perl-Text-Unidecode perl-JSON perl-Unicode-Collate \ perl-Pod-LaTeX perl-LWP-Protocol-https perl-IO-String tetex-latex wget gzip tar ImageMagick elinks \ poppler-utils chkconfig unzip cpan python3-html2text expat-devel libxslt-devel
Next, create the eprints user
adduser eprints
Next, clone the Git repository from GitHub to /opt/ and set the eprints user as the owner. It is probably best to checkout the latest release rather than using HEAD, which may have known issues:
yum install git mkdir /opt/eprints3 chown eprints:eprints /opt/eprints3 chmod 2775 /opt/eprints3 su eprints git clone https://github.com/eprints/eprints3.4.git /opt/eprints3 cd /opt/eprints3/ git checkout tags/v3.4.5
Next, as the eprints user copy /opt/eprints3/perl_lib/EPrints/SystemSettings.pm.tmpl to /opt/eprints3/perl_lib/EPrints/SystemSettings.pm
EPrints is now fully installed at /opt/eprints3. However, you will not yet have a running archive. Follow the Getting Started with EPrints 3 instructions to set this up.
Now update Apache configuration to set it up to use the eprints user and group, by ensuring the following lines are set thus in /etc/httpd/conf/httpd.conf:
User eprints Group eprints
Now add the file /etc/httpd/conf.d/eprints.conf with the following line:
Include /opt/eprints3/cfg/apache.conf
There is a good chance your operating system will have SELinux enabled. If so, follow the instructions under Troubleshooting for how to let EPrints do what it needs to do through SELinux.
Finally, restart Apache:
service httpd restart
You should now be able to access your vanilla repository at the hostname you specified when running epadmin create from the Getting Started with EPrints 3 instructions. You should then login as the admin user you created during this process and turn on the indexer. This can be done by clicking on the Admin link under the Logged in menu, then clicking on the System Tools tab and finally the Start Indexer button.
Non-critical Dependencies
Some of EPrints non-core functionality may not work out of the box as these dependencies are not included as part of the installation process. Here is a list of known additional packages (and the YUM repositories in which they can be found) or CPAN modules than can be installed. The latter should be installed using the cpan command as the root user.
- perl-Geo-IP (epel) - Allows IRstats2 to determine the location of views/downloads for publications.]
- perl-Spreadsheet-WriteExcel (epel) - To allow Multiline Excel export.
- CPAN module Text::Refer - Required for EndNote import.
- CPAN module TeX::Encode - Required for BibTex export needed for publications flavour.
- perl-Image-ExifTool (epel) - Sometimes needed to support generation of thumbnails for uploaded documents/images.
- perl-LWP-Protocol-https (base) - Needed by DataCite plugin to register repository DOIs.
Next steps
Configure eprints per Getting_Started_with_EPrints_3.
Installing using RPM package
As of EPrints 3.4.5 RPM (.rpm) packages will no longer be produced. Due to the complexity of the installation process there are a number of inherent problems with installing using a RPM package. As new operating systems (e.g. RHEL, Fedora, CentOS, Rocky Linux, etc.) are released the available packages have changed. This often leads to packages upon which that EPrints depends ceasing to be available. This makes it very difficult or impossible to install the RPM package, as well as requiring retrospective effort to install packages using other sources, (e.g. additional RPM repositories or using CPAN). However, the main concern of installation the RPM package, particularly through EPrints' RPM repository, is that it can lead to EPrints being upgraded automatically to a new version. Upgrading EPrints using a RPM package is likely to lead to EPrints producing errors, as configuration changes are often required as part of the upgrade process. Automating these changes within the RPM package would be all but impossible due to EPrints' high level of customisability.
Operating System Version Specific Requirements
RHEL / CentOS 7
No significant issues.
RHEL / CentOS / Rocky 8
elinks package is no longer available
The RPM package elinks is no longer available in RHEL / CentOS / Rocky 8 even with EPEL package repository. elinks is only needed if are intending to upload web pages or XML files and these need to be indexed by EPrints, so if installing from source use the following DNF command line:
dnf install libxml2 libxslt httpd mod_perl perl-DBI perl-DBD-MySQL perl-IO-Socket-SSL perl-Time-HiRes \ perl-CGI perl-Digest-MD5 perl-Digest-SHA perl-JSON perl-XML-LibXML perl-XML-LibXSLT perl-XML-SAX \ perl-MIME-Lite perl-Text-Unidecode perl-JSON perl-Unicode-Collate perl-Pod-LaTeX tetex-latex wget \ gzip tar ImageMagick poppler-utils chkconfig unzip cpan
Alternatively, for RHEL 8 you can enable the codeready-builder repository, which still has elinks available:
subscription-manager repos --enable "codeready-builder-for-rhel-8-$(arch)-rpms"
If you are running CentOS 8 (or Rocky 8) you can similarly enable the powertools repository:
dnf config-manager --set-enabled powertools
perl-IO-String package is no longer available
The RPM package perl-IO-String is no longer available in RHEL / CentOS / Rocky 8 even with EPEL package repository. You can enable the codeready-builder repository, which still has perl-IO-String available:
subscription-manager repos --enable "codeready-builder-for-rhel-8-$(arch)-rpms"
If you are running CentOS 8 (or Rocky 8) you can similarly enable the powertools repository:
dnf config-manager --set-enabled powertools
Sub-dependency packages perl-MIME-Types and perl-Digest-SHA1 are no longer available
perl-MIME-Lite package requires perl-MIME-Types package and perl-Apache-DBI requires perl-Digest-SHA1 package neither of which available on RHEL / CentOS / Rocky 8, even with EPEL repository enabled. You can enable the codeready-builder repository, which still has these packahes available:
subscription-manager repos --enable "codeready-builder-for-rhel-8-$(arch)-rpms"
If you are running CentOS 8 (or Rocky 8) you can similarly enable the powertools repository:
dnf config-manager --set-enabled powertools
Compatibility issues with XML::LibXSLT
Installing XML::LibXSLT either through the perl-XML-LibXSLT RPM or CPAN causes httpd to generate a segmentation fault when you try to start it. The most straightfarward solution to this is to run httpd in mpm-prefork rather than mpm-event mode. This can be acheived by editing /etc/httpd/conf.modules.d/00-mpm.conf and commenting out the following line:
LoadModule mpm_event_module modules/mod_mpm_event.so
and uncommenting the mpm-prefork line:
LoadModule mpm_prefork_module modules/mod_mpm_prefork.so
The alternative solution is to make sure that XML::LibXSLT is not installed through the perl-XML-LibXSLT RPM or CPAN. EPrints will still run without this but you will not be able to use XSLT-based import plugins. Unless you have installed any additional import plugins, the only ones affected are Atom XML and OpenXML Bibliography.
RHEL / CentOS / Rocky 9
elinks package is no longer available
The RPM package elinks is no longer available in RHEL / CentOS / Rocky 9 even with EPEL package repository. elinks is only needed if are intending to upload web pages or XML files and these need to be indexed by EPrints, so if installing from source use the following DNF command line:
dnf install libxml2 libxslt httpd mod_perl perl-DBI perl-DBD-MySQL perl-IO-Socket-SSL perl-Time-HiRes \ perl-CGI perl-Digest-MD5 perl-Digest-SHA perl-JSON perl-XML-LibXML perl-XML-LibXSLT perl-XML-SAX \ perl-MIME-Lite perl-Text-Unidecode perl-JSON perl-Unicode-Collate tetex-latex wget gzip tar \ ImageMagick poppler-utils chkconfig unzip cpan python3-html2text
In future versions of EPrints (3.4.5+) EPrints will be switching to use html2text rather than elinks. changeset shows how the places where elinks is used can be changed to switch over to using html2text. To make use of this you will need to add python3-html2text to the DNF command above.
perl-IO-String package is no longer available
The RPM package perl-IO-String is no longer available in RHEL / CentOS / Rocky 9 even with EPEL package repository. You can enable the Code Ready Builder repository, which still has perl-IO-String available:
subscription-manager repos --enable "codeready-builder-for-rhel-9-$(arch)-rpms"
If you are running CentOS 9 (or Rocky 9) you can similarly enable the crb repository:
dnf config-manager --set-enabled crb
Then install perl-IO-String:
dnf install perl-IO-String
Sub-dependency packages perl-MIME-Types and perl-Digest-SHA1 are no longer available
perl-MIME-Lite package requires perl-MIME-Types package and perl-Apache-DBI requires perl-Digest-SHA1 package neither of which available on RHEL / CentOS / Rocky 9, even with EPEL repository enabled. You can enable the Code Ready Builder repository, which still has these packages available:
subscription-manager repos --enable "codeready-builder-for-rhel-9-$(arch)-rpms"
If you are running CentOS 9 (or Rocky 9) you can similarly enable the crb repository:
dnf config-manager --set-enabled crb
Then install perl-MIME-Types and perl-Digest-SHA1:
dnf install perl-MIME-Types perl-Digest-SHA1
perl-Apache-DBI and Pod::LaTeX packages is no longer available
The RPM packages perl-Apache-DBI are no longer available in RHEL / CentOS / Rocky 9 even with EPEL or Code Ready Builder package repository. You can install this using CPAN instead:
cpan Apache::DBI cpan Pod::LaTeX
Compatibility issues with XML::LibXSLT
Installing XML::LibXSLT either through the perl-XML-LibXSLT RPM or CPAN causes httpd to generate a segmentation fault and core dump when you try to start it. The most straightfarward solution to this is to run httpd in mpm-prefork rather than mpm-event mode. This can be acheived by editing /etc/httpd/conf.modules.d/00-mpm.conf and commenting out the following line:
LoadModule mpm_event_module modules/mod_mpm_event.so
and uncommenting the mpm-prefork line:
LoadModule mpm_prefork_module modules/mod_mpm_prefork.so
The alternative solution is to make sure that XML::LibXSLT is not installed through the perl-XML-LibXSLT RPM or CPAN. EPrints will still run without this but you will not be able to use XSLT-based import plugins. Unless you have installed any additional import plugins, the only ones affected are Atom XML and OpenXML Bibliography.
MySQL root user cannot be used
On RHEL/Centos/Rocky Linux 9 rather than prompting you to set a password for MySQL server when installing this as a dependency, it just installs it without. However, this means on the root user can login to MySQL to add a database. As "epadmin create" runs at the eprints user, this will not be able to create a database for EPrints. To get round this as the root user connect to MySQL (just type "mysql" at the command line) and type the following two commands. You will want to set your own password (i.e. not 'changeme') , you may also want to restrict which databases the eprints user has control over. If you know the "Archive ID" you are going to use.
CREATE USER 'eprints'@'localhost' IDENTIFIED by 'changeme'; GRANT ALL PRIVILEGES ON *.* TO 'eprints'@'localhost' WITH GRANT OPTION;
Troubleshooting
- If setup is successful but then you cannot access EPrints through your web browser this is often due to firewalld being enabled by default on recent versions of RHEL, Fedora and CentOS. If firewalld is running either disable it (systemctl disable --now firewalld) or configure it to allow HTTP/HTTPS traffic.
- If you are having problems uploading files or similar functional issues, this may be due to SELinux being enabled. See advice for enabling SELinux with EPrints.
- If you are having trouble getting additional subdomains beyond username.eprints-hosting.org working, then it could be you need wildcards enabled on the DNS records to facilitate the likes of somethingcoolhere.username.eprints-hosting.org urls.