wiki:apachesolr

Version 50 (modified by Bart, 10 years ago) (diff)

--

Apache Solr on mirabal.mayfirst.org

Apache Solr is an indexing and searching program that can replace the Drupal core search engine with a much improved system.

mirabal.mayfirst.org is a dedicated Apache Solr Server that is available for use by any May First/People Link member. Feel free to contact the support list, open a ticket or contact the support team if you want to use Apache Solr on your website.

Security

Solr does not come with a generic authentification mechanism. The Solr installation on mirabal uses a two-level security system. The connection between the web and the Solr server is secured by ssh against attacks from the web. The instances (denoted by $SITENAME) on the Solr server are protected by random admin paths (denoted by $SITEPATH) from one another. The random admin paths are generated by core_ followed by pwgen -1 --secure 30

Reliability

The ssh connections between the web and the Solr server is supervised by runit. If there is a problem ssh will end the connection and runit will restart ssh. Sometimes Solr crashes Tomcat if there are too many requests at once. So we use monit to monitor Solr and restart Tomcat automatically if it crashed.

Web site Administrators

If you run a Drupal web site, you can use Solr. The first step is to determine if you primary host is connected to our Solr server. The following hosts are currently connected:

  • lucy.mayfirst.org
  • lumumba.mayfirst.org
  • june.mayfirst.org
  • sojourner.mayfirst.org
  • at.indymedia.org
  • bethemedia.org.uk

If your primary server is not listed, please open a ticket or contact the support team to request having your primary host added.

Enable Apache Solr on Drupal

To enable Solr you have to install Apache Solr Drupal module which is done best through drush:

drush dl apachesolr
drush solr-phpclient

Configure Apache Solr on Drupal

Once your host is solr-enabled then you can then connect to the Apache Solr server. Enter on your website at admin/settings/apachesolr:

Solr host name: localhost
Solr port: 9080
Solr path: /solr/$SITEPATH

Add TcpForward exception

MFPL servers by default do not allow TCP Forwarding. You can create an exception by adding:

  $sshd_tcp_forwarding = "yes"

To the server's puppet configuration before the include ssh line.

Administrators

Below are the steps administrators must take to provide solr access.

On the primary server

Create a new user on the server (called $SERVERNAME with host $SERVERHOST from now on) with the user name solr-ssh and add:

ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQDK9Rnog/DuOeFN0gusPadGHtAFs/vrfuNyxsNpakycA/+hkKbgEid/xR4Tbbs25ak03bBJePdCf/PpORcB2rgiFwte4fOAfoXX/VHBnqZLFizhZKPCtG0gstQXTwdvOkx+8p5yPODVQvkqeOJCxF7EAI5B9VgcoJWdh2tQX+e07v/DgRtTs+01re0ZmVxmpGpgRWICdaPms9Hh/DwJT7gs19TpPv1qBgzDTN/z12b/6BRsShk/eEGVwslGF3meFA5+saWPmCSJKKK3Pg0btp3LxLkZfWUtdnVr7ASeJlevnthOPlSMp0ITs6oznhQEjycWK+nBYNUrK+cguiMepe/t solr-ssh@mirabal

to the new user's ~/.ssh/authorized_keys file. mirabal will connect to $SERVERHOST with these credentials and provide an ssh tunnel to the Apache Solr server.

On Mirabal

There are two things to do on mirabal.mayfirst.org: add a ssh tunnel from mirabal to the web server and create a new Apache Solr instance in Apache Tomcat.

Create SSH tunnel

Log as solr-ssh into $SERVER from solr-ssh on mirabal.mayfirst.org

sudo -u solr-ssh ssh solr-ssh@$SERVERHOST

End the connection with exit

On mirabal, add a SSH tunnel by using solr_addssh (see solr_addssh --man for help and more options)

solr_addssh $SERVERNAME $SERVERHOST

You can delete (stop/disable/remove) port forwarding with solr_delssh $SERVERNAME (see solr_delssh --man for help and more options).

Test SSH tunnel

Log as solr-ssh into $SERVERHOST and try

curl http://localhost:9080

to see the Apache Tomcat welcome page from mirabal.mayfirst.org.

Create new Solr configuration

Use solr_addsite to create a new Solr instance

solr_addsite $SITENAME

The script will output the Solr admin path ($SITEPATH) for the new Solr instance.

You can delete an instance with solr_delsite $SITENAME (see solr_delsite --man for help and more options).

Restart Apache Tomcat

Restart Apache Tomcat with

service tomcat6 restart

Test new Solr site

Check that the new core is accessible with:

curl http://localhost:8080/solr/$SITEPATH/admin/.

Troubleshooting

Make sure that the ssh tunnel is up

sv status solr-ssh-$SERVERNAME

If the tunnel is down you should have a look at the logs: tail -f /var/log/supervise/solr-ssh-$SERVERNAME/current

For more information add -v to the ssh command in /etc/sv/$SERVERNAME/run and check the logs again.

Check the running Solr instances with

solr_check

(see solr_check --man for help and more options)