add auto-detect

Author: Oreoxmt

online-unsafe-recovery.md

@@ -7,9 +7,11 @@ summary: Learn how to use Online Unsafe Recovery.
 
 > **Warning:**
 >
-> Online Unsafe Recovery is a type of lossy recovery. If you use this feature, the integrity of data and data indexes cannot be guaranteed.
+> - Online Unsafe Recovery is a type of lossy recovery. If you use this feature, the integrity of data and data indexes cannot be guaranteed. If you encounter any issues, other tools or steps are required to recover your data.
+>
+> - Online Unsafe Recovery is introduced since TiDB v6.1.0. For versions earlier than v6.1.0, this is an experimental feature and the behavior is not the same as this document describes. It is **NOT** recommended to use this in versions earlier than v6.1.0. When you use this feature in a version other than v6.1, refer to the corresponding version of this document.
 
-When permanently damaged replicas cause part of data on TiKV to be unreadable and unwritable, you can use the Online Unsafe Recovery feature to perform a lossy recovery operation.
+When permanently damaged replicas cause part of data on TiKV to be unreadable and unwritable, you can use the Online Unsafe Recovery feature to perform a lossy recovery operation and make TiKV continue providing services.
 
 ## Feature description
 
@@ -56,6 +58,8 @@ If the command returns `Failed`, PD Control has failed to register the task to P
 
 To specify the longest allowable duration of a recovery task, use the `--timeout <seconds>` option. If this option is not specified, the longest duration is 5 minutes by default. When the timeout occurs, the recovery is interrupted and returns an error.
 
+If you cannot confirm the store IDs of failed stores, in the scenario where [`pd-recover`](/pd-recover.md) is executed to recover PD cluster, and the store information of unrecoverable TiKV nodes is lost, you can use the `--auto-detect` parameter to allow an empty store ID list. In this mode, all store IDs that are not in the PD store list is considered as unrecoverable and will be removed.
+
 > **Note:**
 >
 > - Because this command needs to collect information from all peers, it might cause an increase in memory usage (100,000 peers are estimated to use 500 MiB of memory).
@@ -87,8 +91,11 @@ Each of the above stages is output in the JSON format, including information, ti
 ```json
 [
     {
-        "info": "Unsafe recovery enters collect report stage: failed stores 4, 5, 6",
-        "time": "......"
+        "info": "Unsafe recovery enters collect report stage",
+        "time": "......",
+        "details" : [
+            "failed stores 4, 5, 6",
+        ]
     },
     {
         "info": "Unsafe recovery enters force leader stage",