Add a section on how to use helm stable/postgresql for the hub

Reference #567

Signed-off-by: Gaetan Semet <gaetan@xeberon.net>

doc/source/advanced.md

@@ -75,6 +75,59 @@ and **google cloud's ingress controller**.
 This should provision a certificate, and keep renewing it whenever it gets close
 to expiry!
 
+
+## Using PostgreSQL as database for you Hub.
+
+Four types of settings are allowed for storing hub state, as described in
+the key ``hub.db.type``:
+
+- ``sqlite-pvc``
+- ``sqlite-memory``
+- ``mysql``
+- ``postgres``
+
+Using PostgreSql or MySQL provides the main advantage to allow seemless update
+of your hub without discontinuity of service (logged in users can always
+continue to work on running server, even with SQLite, but the difference between
+SQLite and MySQL/PostgreSQL are for new user, in order to allow them to log in
+even when a new hub is starting).
+
+The easiest way of running PostgreSQL for the Hub is to use Helm
+'stable/postgresal' chart:
+
+```bash
+$ cat > pg-values.yaml <<EOF
+postgresUser: jupyterhub
+postgresPassword: jupyterhub
+postgresDatabase: jupyterhub
+persistence:
+  enabled: False  # see helm 'stable/postgresql` documentation if you want pesistance
+EOF
+$ helm upgrade --install --name hub-db -f pg-values.yaml stable/postgresql
+```
+
+And then set `hub.db.type` to:
+
+```yaml
+hub:
+  db:
+    type: postgresql
+    url: postpostgres+psycopg2://jupyerhub:jupyterhub@hub-db-postgresql:5432/jupyterhub
+```
+
+Note the following points when using PostgreSQL as database for the Hub:
+
+- If you do not use a persistant volume on PostgreSQL, your database will be
+  lost when the PostgreSQL server will restart. Do not use `--recreate-pods`
+  helm option. It will not cause any data loss for user, just the proxy will
+  not be able to route logged users. 
+- even without persistance, your hub can still be upgraded smoothly for the
+  users, because a Hub restart will not cause the PostgreSQL pod to restart.
+- The Hub will use Rolling Update, so a new hub will be started while the 
+  other is still running and the transition will be transparent for the user.
+- if case of DB loss or reset, simply ensure all your server are shutdown
+  (all Kubernetes pods starting with `jupyter-*`) before restarting the Hub.
+
 ## Arbitrary extra code and configuration in `jupyterhub_config.py`
 
 Sometimes the various options exposed via the helm-chart's `values.yaml` is not