Class: ElasticGraph::Admin::ClusterConfigurator::ClusterSettingsManager

Inherits:
Object
  • Object
show all
Defined in:
lib/elastic_graph/admin/cluster_configurator/cluster_settings_manager.rb

Overview

Responsible for updating datastore cluster settings based on the mode EG is in, maintenance mode or indexing mode

Instance Method Summary collapse

Constructor Details

#initialize(datastore_clients_by_name:, datastore_config:, logger:) ⇒ ClusterSettingsManager 



16
17
18
19
20
 
# File 'lib/elastic_graph/admin/cluster_configurator/cluster_settings_manager.rb', line 16

def initialize(datastore_clients_by_name:, datastore_config:, logger:)
  @datastore_clients_by_name = datastore_clients_by_name
  @datastore_config = datastore_config
  @logger = logger
end

Instance Method Details

#end_index_maintenance_mode!(cluster_spec) ⇒ Object

Ends index maintenance mode, if it has not already ended. This method is idempotent.

Outside of this mode, you cannot safely delete or update the index configuration. However, new rollover indices will correctly be auto-created as documents that fall in new months or years are indexed.

‘cluster_spec` can be the name of a specific cluster (as a string) or `:all_clusters`.



43
44
45
46
47
48
49
 
# File 'lib/elastic_graph/admin/cluster_configurator/cluster_settings_manager.rb', line 43

def end_index_maintenance_mode!(cluster_spec)
  cluster_names_for(cluster_spec).each do |cluster_name|
    datastore_client_named(cluster_name).put_persistent_cluster_settings(
      desired_cluster_settings(cluster_name, auto_create_index_patterns: ["*#{ROLLOVER_INDEX_INFIX_MARKER}*"])
    )
  end
end

#in_index_maintenance_mode(cluster_spec) ⇒ Object

Runs a block in index maintenance mode. Should be used to wrap any code that updates your index configuration.

‘cluster_spec` can be the name of a specific cluster (as a string) or `:all_clusters`.



54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
 
# File 'lib/elastic_graph/admin/cluster_configurator/cluster_settings_manager.rb', line 54

def in_index_maintenance_mode(cluster_spec)
  start_index_maintenance_mode!(cluster_spec)

  begin
    yield
  rescue => e
    @logger.warn "WARNING: ClusterSettingsManager#in_index_maintenance_mode is not able to exit index maintenance mode due to exception #{e}.\n A bit of manual cleanup may be required (although a re-try should be idempotent)."
    raise # re-raise the same error
  else
    # Note: we intentionally do not end maintenance mode in an `ensure` block, because if an exception
    # happens while we `yield`, we do _not_ want to exit maintenance mode. Exiting maintenance mode
    # could put us in a state where indices are dynamically created when we do not want them to be.
    end_index_maintenance_mode!(cluster_spec)
  end
end

#start_index_maintenance_mode!(cluster_spec) ⇒ Object

Starts index maintenance mode, if it has not already been started. This method is idempotent.

In index maintenance mode, you can safely delete or update the index configuration without worrying about indices being auto-created with dynamic mappings (e.g. due to an indexing race condition). While in this mode, indexing operations on documents that fall into new rollover indices may fail since the auto-creation of those indices is disabled.

‘cluster_spec` can be the name of a specific cluster (as a string) or `:all_clusters`.



30
31
32
33
34
 
# File 'lib/elastic_graph/admin/cluster_configurator/cluster_settings_manager.rb', line 30

def start_index_maintenance_mode!(cluster_spec)
  cluster_names_for(cluster_spec).each do |cluster_name|
    datastore_client_named(cluster_name).put_persistent_cluster_settings(desired_cluster_settings(cluster_name))
  end
end