docs: Add some docs on persistence implementaions

README.md

@@ -6,6 +6,10 @@ A highly available load balancer with support for queueing, routing and auto-sca
 * [Example setups](./docs/example-setups.md)
 * [Routing](./docs/routing/index.md)
   * [PythonScriptRouter](./docs/routing/PythonScriptRouter.md)
+* [Persistence](./docs/persistence/index.md)
+  * [In-memory](./docs/persistence/in-memory.md)
+  * [Redis](./docs/persistence/redis.md)
+  * [Postgres](./docs/persistence/postgres.md)
 * [Scaling](./docs/scaling/index.md)
   * [Stackable](./docs/scaling/stackable.md)
 

docs/design.md

@@ -13,9 +13,9 @@ This imposes the following requirements on the underlying persistence:
 
 The trait `trino_lb_persistence::Persistence` (think of an interface) allows for multiple persistence implementations. Currently the following implementations are available:
 
-1. **In-memory**: Volatile persistence mainly intended for development or testing purposes.
-2. **Redis**: Uses a [Redis Cluster](https://redis.io/docs/management/scaling/) as a distributed key-value store.
-3. **Postgres**: **Experimental** We have seen a few queries too much being send to the Trino clusters, probably related to some transactional problems. Also performance measurements are missing.
+1. [In-memory](./persistence/in-memory.md): Volatile persistence mainly intended for development or testing purposes.
+2. [Redis](./persistence/redis.md): Uses a [Redis Cluster](https://redis.io/docs/management/scaling/) as a distributed key-value store.
+3. [Postgres](./persistence/postgres.md): **Experimental**, as performance measurements are missing.
 
 In case you are interested in a implementation for a different persistence please feel free to open an issue or pull request!
 

docs/persistence/in-memory.md

@@ -0,0 +1,17 @@
+# In-memory persistence
+
+This persistence stores the data in the RAM of the trino-lb server.
+This implies all queued and running queries will be lost when trino-lb is restarted.
+Additionally, as multiple trino-lb instances do not share their memory, you can not run multiple trino-lb instances for scalability or availability reasons.
+
+Because of these restrictions the in-memory persistence is only recommended for testing purpose.
+
+## Configuration
+
+The configuration of the in-memory persistence is the most simple, as it does not require any configurations and can be configured as follows:
+
+```yaml
+trinoLb:
+  persistence:
+    inMemory: {}
+```

docs/persistence/index.md

@@ -0,0 +1,7 @@
+# Persistence
+
+Currently the following persistence implementations are available:
+
+1. [In-memory](./in-memory.md): Volatile persistence mainly intended for development or testing purposes.
+2. [Redis](./redis.md): Uses a [Redis Cluster](https://redis.io/docs/management/scaling/) as a distributed key-value store.
+3. [Postgres](./postgres.md): **Experimental**, as performance measurements are missing.

docs/persistence/postgres.md

@@ -0,0 +1,20 @@
+# Postgres persistence
+
+## Configuration
+
+You only need to provide the url the Postgres is available at.
+Authentication is supported by providing username and password in the url.
+
+```yaml
+trinoLb:
+  persistence:
+    postgres:
+      url: postgres://trino-lb:[email protected]/trino_lb
+      maxConnections: 10 # optional, defaults to 10
+```
+
+## Example installation
+
+The above configuration works with a Postgres installed with the following command:
+
+`helm install postgres bitnami/postgresql --version 13.2.18 --set auth.username=trino-lb,auth.password=trino-lb,auth.database=trino_lb`

docs/persistence/redis.md

@@ -0,0 +1,20 @@
+# Redis persistence
+
+The Redis persistence currently only supports a [Redis Cluster](https://redis.io/docs/management/scaling/) as a distributed key-value store.
+A single-instance Redis server is currently not supported but would be an easy addition.
+
+The Redis persistence is the most tested implementation and is currently the recommended choice for production systems.
+
+## Configuration
+
+You only need to provide the endpoint the Redis cluster is available at.
+Authentication is supported by providing the password in the endpoint (the username is set to an empty string).
+
+The following configuration connects to the Redis cluster running at `trino-lb-redis-cluster.trino-lb.svc.cluster.local` on Port `6379` secured with the password `redis`.
+
+```yaml
+trinoLb:
+  persistence:
+    redis:
+      endpoint: redis://:[email protected]:6379/
+```