Update record indexer README

record_indexer/README.md

@@ -8,7 +8,7 @@ python index_templates/rikolti_template.py
 
 This creates a template that will be used whenever an index with name matching `rikolti*` is added to the cluster.
 
-## Run indexer
+## Run indexer from command line
 
 Create a new index for a collection and add it to the `rikolti-stg` alias:
 
@@ -22,10 +22,29 @@ Add the current stage index for a collection to the `rikolti-prd` alias:
 python -m record_indexer.move_index_to_prod <collection_id>
 ```
 
-Note that the index creation code enforces the existence of one stage index at a time. This means we can simply supply the collection ID as input and the process will move the current stage index to production.
+## Indexer development using aws-mwaa-local-runner
 
+See the Rikolti README page section on [Airflow Development](https://github.com/ucldc/rikolti/#airflow-development). In particular, make sure that indexer-related env vars are set as described there.
 
+## Index lifecycle
 
+The lifecycle of an index is as follows:
+
+#### Create new index
+1. Create a new index named `rikolti-{collection_id}-{version}`, where `version` is the current datetime).
+2. Remove any existing indices for the collection from the `rikolti-stg` alias.
+3. Add the new index to the `rikolti-stg` alias.
+4. Delete any older unaliased indices, retaining the number of unaliased indices specified by `settings.INDEX_RETENTION`.
+
+Note that the index creation code enforces the existence of one stage index at a time.
+
+#### Move staged index to production
+1. Identify the current stage index for the collection.
+2. Remove any existing indices for the collection from the `rikolti-prd` alias.
+3. Add the current stage index to the `rikolti-prd` alias. (This means that at this stage in the lifecycle, the index will be aliased to `rikolti-stg` and `rikolti-prd` at the same time.)
+
+#### Delete old index
+This happens during index creation (see step 4. above).