Document force recovery (#4446)

doc/code_snippets/snippets/config/instances.enabled/force_recovery/README.md

@@ -0,0 +1,18 @@
+# Force recovery
+
+A sample application demonstrating how to start an instance with a corrupted write-ahead log file.
+
+## Running
+
+- To start the instance using the `tt` utility, execute the following command in the [config](../../../config) directory:
+
+  ```console
+  $ TT_FORCE_RECOVERY=true tt start force_recovery
+  ```
+
+- To start the instance using the `tarantool` command, you can use both the `TT_FORCE_RECOVERY` environment variable and add the `--force-recovery` option.
+  In the example below, the `tarantool` command is executed in the [config/instances.enabled/force_recovery](../../../config/instances.enabled/force_recovery) directory:
+
+  ```console
+  $ tarantool --name instance001 --config config.yaml --force-recovery -i
+  ```

doc/code_snippets/snippets/config/instances.enabled/force_recovery/config.yaml

@@ -0,0 +1,9 @@
+groups:
+  group001:
+    replicasets:
+      replicaset001:
+        instances:
+          instance001:
+            iproto:
+              listen:
+              - uri: '127.0.0.1:3301'

doc/code_snippets/snippets/config/instances.enabled/force_recovery/instances.yml

@@ -0,0 +1 @@
+instance001:

doc/reference/tarantool_cli_options.rst

@@ -65,6 +65,25 @@ Options
 
     Start an external coordinator used for a :ref:`supervised failover <repl_supervised_failover>`.
 
+.. _tarantool_cli_force_recovery:
+
+..  option:: --force-recovery
+
+    **Since:** :doc:`3.0.0 </release/3.0.0>`.
+
+    Try to start an instance if there is an error while reading a corrupted snapshot or write-ahead log file during the recovery process:
+
+    -   For a corrupted :ref:`snapshot file <index-box_persistence>` -- at the instance start.
+    -   For a corrupted :ref:`write-ahead log file <internals-wal>` -- at the instance start or when applying an update at a replica.
+
+    With this option enabled, Tarantool skips invalid records, reads as much data as possible, and lets the process finish with a warning.
+    When the instance has started, call :ref:`box.snapshot() <box-snapshot>` to make a new snapshot so that the corrupted snapshots or write-ahead logs aren't used for recovery anymore.
+
+    You can also enable force recovery using the ``TT_FORCE_RECOVERY`` environment variable.
+    ``TT_FORCE_RECOVERY`` has a lower priority than the ``--force-recovery`` option.
+
+    Example on GitHub: `force_recovery <https://github.com/tarantool/doc/tree/latest/doc/code_snippets/snippets/config/instances.enabled/force_recovery>`_
+
 .. _index-tarantool_version:
 
 ..  option:: -v, -V, --version