Elasticsearch service restart at each chef run

[kamaradclimber]

Elasticsearch restart at each run due to ark provider which create a symlink and execute a ruby_blockat each run. I'll open a ticket in ark cookbook too.

[karmi]

Cool. Also please see: #98 (comment)

[kamaradclimber]

yes indeed, if you don't use ark it will fix the issue
I can probably make a MR with rpm support if you want

[karmi]

"MR with rpm support"?

[kamaradclimber]

merge request adding a way to download and install a rpm package for rhel family :)

[karmi]

There is some rudimentary support in https://github.com/elasticsearch/cookbook-elasticsearch/blob/master/recipes/rpm.rb to kick the tires and try it out.

As said in other issues, I'd like to move to a more fine grained layout, which would allow to install either from tar.gr, or .deb, or .rpm -- but we need to figure out how to properly set up configs, etc etc, and test the whole refactoring.

[nukemberg]

it seems the cookbook (with elasticsearch::default recipe) will remove /usr/local/elasticsearch due to resource bash "remove the elasticsearch user home"
changing the not_if condition to check for symlink as well as directory solves the issue:

home_dir = "#{node.elasticsearch[:dir]}/elasticsearch"
bash "remove the elasticsearch user home" do
  user    'root'
  code    "rm -rf  #{node.elasticsearch[:dir]}/elasticsearch"
  only_if { ::File.directory?(home_dir) and not ::File.symlink?(home_dir) }
end

[Pneumatus]

This issue also goes away if you drop back to ark 0.1.0. Ark 0.3.0 seems to trigger the restart where the older version does not.

[jsirex]

http://tickets.opscode.com/browse/COOK-2444 - may be this helps somebody...
And this: fewbytes-cookbooks/ark@a5f30f1

[dosercz]

I have updated ES cookbook today and I have the same problem, but downgrade ark to 0.1.0 helps and it's the simplest way for now, as recommends @Pneumatus

[karmi]

@avishai-ish-shalom Thanks for identifying the culprit for removing the /usr/local/elasticsearch directory

[karmi]

Everybody, this issue should now be solved -- no re-downloads or service restarts on each Chef run.

Verified in Vagrant and EC2 (both Chef-Solo only).