Update failing-backup-restore.mdx (#3907)

Author: fpagny

serverless/sql-databases/troubleshooting/failing-backup-restore.mdx

@@ -28,4 +28,30 @@ These issues are caused by using `pg_dump` and `pg_restore` versions that are no
 To solve these issues, upgrade your `pg_dump` and/or `pg_restore` modules:
 
 - You can upgrade them by installing PostgreSQL 16 which includes these tools.
-- If you are using a third-party tool that includes these libraries, upgrade your tool. For instance, pgAdmin supports PostgreSQL 16 ecosystem from version 7.8.
+- If you are using a third-party tool that includes these libraries, upgrade your tool. For instance, pgAdmin supports PostgreSQL 16 ecosystem from version 7.8.
+
+## Automated backup fails due to long-running queries
+
+### Problem
+
+When performing long-running queries (lasting several hours or days), backup operations might not be performed successfully. Backup status will then appear as `error` or `unknown_status`. 
+
+### Cause
+
+These issues are caused by queries locking database rows (usually long running transactions), preventing logical backup operations from reading database rows.
+
+### Solution
+
+To solve these issues, stop these queries:
+
+- List PostgreSQL processes and identify the ones that have been running transactions since several hours ('xact_start' colmun) with the following command:
+    ```
+    SELECT pid, state, usename, query, xact_start, query_start FROM pg_stat_activity ORDER BY xact_start;
+    ```
+- Stop the corresponding queries with:
+    ```
+    SELECT pg_cancel_backend({pid});
+    ```
+    Where `{pid}` is the process id from the long-running query causing the issue (`pid` column of the previous step).
+
+Alternatively, you can also stop long-running queries using a graphical PostgreSQL client such as [pgAdmin](https://www.pgadmin.org/) or [DBeaver](https://dbeaver.io/).