Automatic restart of SolR after Elastic Beanstalk deployment

Question

I have a RoR application on AWS. My app is using SolR for the search engine, but after each deployment, the app is unable to index again. So I have to reset permissions and restart Solr manually with:

chmod 777 -R /solr /tmp /log
RAILS_ENV=production rake sunspot:solr:stop # or I kill the processus if it doesn't work :D 
RAILS_ENV=production rake sunspot:solr:start
RAILS_ENV=production rake sunspot:reindex

Now i'm trying to setup it as an eb extension to automate the deployment. Here is what I tried in my .ebextensions/deploy.config:

container_commands:
  1_change_permissions:
    command: chmod 700 .ebextensions/setup.sh 
  2_restart_solr:
    command: bash .ebextensions/setup.sh 

And here is the setup.sh script:

#!/bin/bash
chmod 777 -R solr/ log/ tmp/
RAILS_ENV=production rake sunspot:solr:restart

The result is the deployment doesn't fail but, only the permissions are changed correctly, and the solr service is running but when I try to index something, it fails (the querying works fine).

I also tried to stop the server before the app is deployed by adding a commands block in my .ebextensions/deploy.config (and I changed my sh script to start the service instead of restart) :

commands:
  1_stop_solr:
    command: cd /var/app/current & RAILS_ENV=production rake sunspot:solr:stop

I got this error (I don't know from where it is executed):

[2015-06-25T09:51:35.510Z] INFO [13207] - [CMD-AppDeploy/AppDeployStage0/EbExtensionPreBuild/Infra-EmbeddedPreBuild/prebuild_0_My_First_Elastic_Beanstalk_Application/Command 1_stop_solr] : Activity execution failed, because: rake aborted! couldn't find HOME environment -- expanding `~'

EDIT 1 (following jay's comment): The indexation process is done when I save the objects.

Here is an example for an entity (and where it fails) :

class Document < ActiveRecord::Base

  # .....

  # SolR entity 
  searchable do
    text :title, :description, :tags 
    integer :user_id
  end 

  # .....

end

**EDIT 2: **

James' answer doesn't fix the problem but, I realize that manually on my EC2 instance, i can just run the 2 following lines :

chmod 777 -R solr/ tmp/ log/
RAILS_ENV=production rake sunspot:reindex"

I tried to use James' link to create a post-deployment script, and the chmod works well but when I add the reindex command into the file, the deployment fails with this error :

[2015-07-07T16:26:25.509Z] INFO  [20402] - [CMD-AppDeploy/AppDeployStage1/AppDeployPostHook/99_restart_delayed_job.sh] : Activity execution failed, because: rake aborted!
Could not find rake-10.4.2 in any of the sources 
/var/app/current/config/boot.rb:3:in `<top (required)>'
/var/app/current/config/application.rb:1:in `<top (required)>'
/var/app/current/Rakefile:4:in `<top (required)>'

Also, if i try to run the command manually (after the post-deployment script chmod), it fails with 500 errors on each item to reindex. So i need to kill the solr server, start and then reindex.

It's really painful :)

Answer 1

I know nothing about Beanstalk, but it looks like the shell that's executing your commands doesn't have the environment variables you're expecting. Assuming this blog post on post-deployment scripts is correct, you should be able to change your setup.sh script to something like this:

 #!/usr/bin/env bash
 . /opt/elasticbeanstalk/support/envvars
 cd $EB_CONFIG_APP_CURRENT
 su -c "RAILS_ENV=production /usr/local/bin/rvm 2.2 do rake sunspot:solr:restart" $EB_CONFIG_APP_USER