Skip to content

ppy/osu-elastic-indexer

Repository files navigation

ElasticIndexer

Component for loading osu! scores into Elasticsearch.

Requirements

  • .NET 6
  • Elasticsearch 7
  • Redis 6

Elasticsearch 8 compatiblity

If using Elasticsearch 8, a minimum version of Elasticsearch 8.2 is required.

The following env needs to be set on the indexer:

ELASTIC_CLIENT_APIVERSIONING=true

and the following must be set in elasticsearch server configuration elasticsearch.yml

xpack.security.enabled: false

or docker environment, e.g. in docker compose:

environment:
  xpack.security.enabled: false

This will enable http connections to elasticsearch and disable the https and authentication requirement, as well as, returning a compatible response to the client.

Usage

Schema

A string value is used to indicate the current schema version to be used.

When the queue processor is running, it will store the version it is processing in a set in Redis at osu-queue:score-index:${prefix}active-schemas.

If a queue processor is stops automatically due to a schema version change, it will remove the version it is processing from the set of versions; it will not be removed if the processor if stopped manually or from processor failures; this is to allow other services to continue pushing to those queues.

Adding items to be indexed

Scores with preserve=true belonging to a user with user_warnings=0 will be added to the index, scores where any of the previous conditions are false will be removed from the index.

Push items to osu-queue:score-index-${schema}

Switching to a new schema

Run dotnet run schema set ${schema} or set osu-queue:score-index:${prefix}schema directly in Redis

Automatic index switching

If there is an already running indexer watching the queue for the new schema version, it will automatically update the alias to point to the new index. When the alias is updated, any index previously used by the alias will be closed.

The alias will not be updated if:

  • the schema value does not change
  • the indexer processing the queue for that version was not running before the change.

When the schema version changes, all indexers processing the queues for any other version will automatically stop.

Configuration

Configuration is loaded from environment variables. No environment files are automatically loaded.

To read environment variables from an env file, you can prefix the command to run with env $(cat {envfile}) replacing {envfile} with your env file, e.g.

Note that this method of passing envvars does not support values with spaces.

env $(cat .env) dotnet run

Additional envs can be set:

env $(cat .env) SCHEMA=1 dotnet run

Environment Variables

BATCH_SIZE

Maximum number of items to handle/dequeue per batch. This affects the size of the _bulk request sent to Elasticsearch.

Defaults to 10000.

BUFFER_SIZE

Maximum number of BATCH_SIZE * BUFFER_SIZE items allowed inflight during queue processing.

Defaults to 5 (default of 50000 items).

DB_HOST

Host for MySQL.

Defaults to localhost.

DB_NAME

Database name.

Defaults to osu.

DB_USER

Database username.

Defaults to root.

DB_PASS

Database password.

DD_AGENT_HOST

Host to submit DataDog/StatsD metrics to.

Defaults to localhost.

DD_ENTITY_ID

Enables DataDog origin detection when running in a container. See DataDog documentation.

ES_INDEX_PREFIX

Optional prefix for the index names in elasticsearch.

ES_HOST

Url to the Elasticsearch host.

Defaults to http://localhost:9200

REDIS_HOST

Redis connection string; see here for configuration options.

Defaults to localhost

SCHEMA

Schema version for the queue; see Schema.

Commands

This documentation assumes dotnet run can be used; in cases where dotnet run is not available, the assembly should be used, e.g. dotnet osu.ElasticIndexer.dll

Watching a queue for new scores

Running queue will automatically create an index if an open index matching the requested schema does not exist. If a matching open index exists, it will be reused.

SCHEMA=${schema} dotnet run queue watch

e.g.

SCHEMA=1 dotnet run queue watch

Getting the current schema version

dotnet run schema get

Setting the schema version

dotnet run schema set ${schema}

Unsetting the schema version

This is used to unset the schema version for testing purposes.

dotnet run schema clear

Changing the alias to a new index

The index the alias points to can be changed manually:

dotnet run schema alias ${schema}

will update the index alias to the latest index with schema ${schema} tag.

List indices

To list all indices and their corresponding states (schema, aliased, open or closed)

dotnet run index list

Closing unused indices

This will close all score indices except the active one, unloading them from Elasticsearch's memory pool.

dotnet run index close

A specific index can be closed by passing in index's name as an argument; e.g. the following will close index_1:

dotnet run index close index_1

Cleaning up closed indices

This will delete all closed indices and free up the storage space used by those indices. The command will only delete an index if it is in the closed state.

dotnet run index delete

Passing arguments to the command will delete the matching index:

dotnet run index delete index_1

Adding fake items to the queue

For testing purposes, we can add fake items to the queue:

SCHEMA=1 dotnet run queue pump-fake

It should be noted that these items will not exist or match the ones in the database.

Queuing a specific score for indexing

SCHEMA=${schema} dotnet run queue pump-score ${id}

will queue the score with ${id} for indexing; the score will be added or deleted as necessary, according to the value of SoloScore.ShouldIndex.

See Queuing items for processing from another client

Adding existing database records to the queue

SCHEMA=1 dotnet run queue pump-all

will read existing solo_scores in chunks and add them to the queue for indexing. Only scores with a corresponding phpbb_users entry will be queued.

Extra options:

--from {id}: solo_scores.id to start reading from

--switch: Sets the schema version after the last item is queued; it does not wait for the item to be indexed; this option is provided as a conveninence for testing.

Listing known versions currently being processed

dotnet run active-schemas list

will list the versions known to have queue processors listening on the queue.

Manually add or remove known versions

For debugging purposes or to perform and manual maintenance or cleanups, the list of versions can be updated manually:

dotnet run active-schemas add ${schema}
dotnet run active-schemas remove ${schema}

(Re)Populating an index

Populating an index is done by pushing score items to a queue.

Docker

docker build -t ${tagname} -f osu.ElasticIndexer/Dockerfile osu.ElasticIndexer

docker run -e SCHEMA=1 -e "ES_HOST=http://host.docker.internal:9200" -e "ES_INDEX_PREFIX=docker." -e "REDIS_HOST=host.docker.internal" -e "DB_CONNECTION_STRING=Server=host.docker.internal;Database=osu;Uid=osuweb;SslMode=None;" ${tagname} ${cmd}

where ${cmd} is the command to run, e.g. dotnet osu.ElasticIndexer.dll queue

Typical use-cases

Queuing items for processing from another client

Push items into the Redis queue "osu-queue:score-index-${schema}" e.g.

ListLeftPush("osu-queue:score-index-1", "{ \"ScoreId\": 1 }");

or from redis-cli:

LPUSH "osu-queue:score-index-1" "{\"ScoreId\":1}"

Indexing a score by id

{ "ScoreId": 1 }

Queuing a whole score

{
    "Score": {Solo.Score}
}