Test cluster Service Unavailable

[khoan]

Other than sleep, is there a way to wait until index is ready?

when testing Elasticsearch in rspec, cannot query index after it's created.

# spec/bam.rb
require 'rake'
require 'elasticsearch/extensions/test/cluster/tasks'
require 'elasticsearch/persistence/model'

class Bam
  include Elasticsearch::Persistence::Model
end

RSpec.describe 'Bam' do

  it 'cannot use elasticsearch endpoint after created' do
    Elasticsearch::Persistence.client = Elasticsearch::Client.new(host: 'localhost:9250', log: true)
    Elasticsearch::Extensions::Test::Cluster.start(nodes: 1)

    Bam.create_index!(force: true)

    # neither of the following work
    # Bam.refresh_index!
    # Elasticsearch::Persistence.client.indices.refresh

    # workaround 1 - wait for index to come online
    # sleep(0.1)

    # workaround 2 - index a document and wait for it to refresh
    # bam = Bam.create(id: 'bam')
    # Bam.refresh_index!
    # bam.delete

    Bam.count

    Elasticsearch::Extensions::Test::Cluster.stop
  end

end

$ bundle exec rspec spec/bam.rb
...
2015-06-17 10:34:46 +1000: PUT http://localhost:9250/bams [status:200, request:0.678s, query:n/a]
2015-06-17 10:34:46 +1000: > {"settings":{},"mappings":{"bam":{"properties":{"created_at":{"type":"date"},"updated_at":{"type":"date"}}}}}
2015-06-17 10:34:46 +1000: < {"acknowledged":true}
2015-06-17 10:34:46 +1000: GET http://localhost:9250/bams/bam/_search?search_type=count [status:503, request:0.016s, query:N/A]
2015-06-17 10:34:46 +1000: > {"query":{"match_all":{}}}
2015-06-17 10:34:46 +1000: < {"error":"SearchPhaseExecutionException[Failed to execute phase [query], all shards failed]","status":503}
2015-06-17 10:34:46 +1000: [503] {"error":"SearchPhaseExecutionException[Failed to execute phase [query], all shards failed]","status":503}
...

[karmi]

Sure, just use refresh_index! (or client.indices.refresh of course)

[khoan]

@karmi thanks for the suggestion. Neither of Bam.refresh_index! nor Elasticsearch::Persistence.client.indices.refresh works. But sleeping for a bit seems to work.

[karmi]

On a really slow system, you might want to wait for cluster health, see eg. until client.cluster.health(level: 'indices')['indices']['bicycles.stackexchange.com']['status'] == 'green' in https://github.com/elastic/elasticsearch-ruby/tree/master/elasticsearch-dsl#features-overview

[khoan]

that's a more reliable way of checking.

I guess other way is to index a document, call refresh_index!, then delete document.