17376 FIX Empty postgres_bloat section under windows

This werk could be relevant for you in case you monitor multiple postgres databases on a windows host.
If the corresponding `postgres_bloat` section is empty:
```
<<<postgres_bloat:sep(59)>>>
[[[data]]]
[databases_start]
postgres
another_db
yet_another_db
[databases_end]

<<<win_netstat:cached(12345676,300)>>>
```

Under windows, the bloat query will now be executed _per_ database in order to avoid that issue.

Change-Id: Icedec27b3e5c40135adc233d3e9606541aa3796d
JIRA-Ref: SUP-23843

Author: TimotheusBachinger

.werks/17376.md

@@ -0,0 +1,28 @@
+[//]: # (werk v2)
+# Empty postgres_bloat section under windows
+
+key        | value
+---------- | ---
+date       | 2025-07-28T13:32:30+00:00
+version    | 2.3.0p36
+class      | fix
+edition    | cre
+component  | checks
+level      | 1
+compatible | yes
+
+This werk could be relevant for you in case you monitor multiple postgres databases on a windows host.
+If the corresponding `postgres_bloat` section is empty:
+```
+<<<postgres_bloat:sep(59)>>>
+[[[data]]]
+[databases_start]
+postgres
+another_db
+yet_another_db
+[databases_end]
+
+<<<win_netstat:cached(12345676,300)>>>
+```
+
+Under windows, the bloat query will now be executed _per_ database in order to avoid that issue.

agents/plugins/mk_postgres.py

@@ -786,13 +786,22 @@ def get_bloat(self, databases, numeric_version):
 
         query = "\\pset footer off \\\\"
 
+        response = ""
         cur_rows_only = False
+        # On windows, we saw issues when more than one database was queried at once.
+        # See SUP-23843
+        # We therefore execute the query per database and accumulate the results.
         for idx, database in enumerate(databases):
-            query = "%s \\c %s \\\\ %s" % (query, database, bloat_query)
-            if idx == 0:
-                query = "%s \\pset tuples_only on" % query
-
-        return self.run_sql_as_db_user(query, mixed_cmd=True, rows_only=cur_rows_only)
+            current_query = "%s \\c %s \\\\ %s" % (query, database, bloat_query)
+            if idx != 0:
+                # Switch off the header for all but the first database: this is now needed as we
+                # execute psql _per_ database but want to get the same format as before.
+                current_query = "\\pset tuples_only on %s" % current_query
+                response += "\n"
+            response += self.run_sql_as_db_user(
+                current_query, mixed_cmd=True, rows_only=cur_rows_only
+            )
+        return response
 
 
 class PostgresLinux(PostgresBase):