From a552f7ce4ccd9cb2d62c94f754f4831c212c6ff5 Mon Sep 17 00:00:00 2001
From: Stefan Miklosovic <smiklosovic@apache.org>
Date: Thu, 22 Jan 2026 15:33:20 +0100
Subject: [PATCH] Do not serialize dropped columns to SSTable header after
 compaction

patch by Stefan Miklosovic; reviewed by TBD for CASSANDRA-21000
---
 .../cassandra/db/SerializationHeader.java     |  19 ++-
 .../SerializationHeaderCompactionCQLTest.java | 149 ++++++++++++++++++
 2 files changed, 167 insertions(+), 1 deletion(-)
 create mode 100644 test/unit/org/apache/cassandra/db/SerializationHeaderCompactionCQLTest.java

diff --git a/src/java/org/apache/cassandra/db/SerializationHeader.java b/src/java/org/apache/cassandra/db/SerializationHeader.java
index 841f7b305198..1a89eaabd0c7 100644
--- a/src/java/org/apache/cassandra/db/SerializationHeader.java
+++ b/src/java/org/apache/cassandra/db/SerializationHeader.java
@@ -88,13 +88,30 @@ public static SerializationHeader make(TableMetadata metadata, Collection<SSTabl
         // but rather on their stats stored in StatsMetadata that are fully accurate.
         EncodingStats.Collector stats = new EncodingStats.Collector();
         RegularAndStaticColumns.Builder columns = RegularAndStaticColumns.builder();
+        RegularAndStaticColumns currentSchemaColumns = metadata.regularAndStaticColumns();
         // We need to order the SSTables by descending generation to be sure that we use latest column metadata.
         for (SSTableReader sstable : orderByDescendingGeneration(sstables))
         {
             stats.updateTimestamp(sstable.getMinTimestamp());
             stats.updateLocalDeletionTime(sstable.getMinLocalDeletionTime());
             stats.updateTTL(sstable.getMinTTL());
-            columns.addAll(sstable.header.columns());
+            for (ColumnMetadata column : sstable.header.columns())
+            {
+                // Include column in a header only if it is in the current schema.
+                // It is not enough to check if a column is not among dropped ones,
+                // because that would not cover the scenario when a column is dropped
+                // and re-added back before flush. Consider:
+                //
+                // insert into ks.tb (id, val) values (1, 2);
+                // alter table ks.tb drop val;
+                // alter table ks.tb add val int;
+                //
+                // If we backed our logic only by a check against a dropped column,
+                // then we would see that it was dropped (in metadata.droppedColumn)
+                // but we would omit the fact that it was actually re-added.
+                if (currentSchemaColumns.contains(column))
+                    columns.add(column);
+            }
         }
         return new SerializationHeader(true, metadata, columns.build(), stats.get());
     }
diff --git a/test/unit/org/apache/cassandra/db/SerializationHeaderCompactionCQLTest.java b/test/unit/org/apache/cassandra/db/SerializationHeaderCompactionCQLTest.java
new file mode 100644
index 000000000000..8ea0ae640c3e
--- /dev/null
+++ b/test/unit/org/apache/cassandra/db/SerializationHeaderCompactionCQLTest.java
@@ -0,0 +1,149 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.cassandra.db;
+
+import org.junit.Test;
+
+import org.apache.cassandra.cql3.CQLTester;
+import org.apache.cassandra.io.sstable.format.SSTableReader;
+import org.apache.cassandra.schema.ColumnMetadata;
+
+import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.assertFalse;
+import static org.junit.Assert.assertTrue;
+
+public class SerializationHeaderCompactionCQLTest extends CQLTester
+{
+    @Test
+    public void testDroppedColumnPresenceInSerialisationHeader() throws Throwable
+    {
+        createTable("CREATE TABLE %s (id int PRIMARY KEY, a text, b text, c text)");
+        execute("INSERT INTO %s (id, a, b, c) VALUES (1, 'a1', 'b1', 'c1')");
+        flush(keyspace());
+        alterTable("ALTER TABLE %s DROP b");
+
+        ColumnFamilyStore cfs = getCurrentColumnFamilyStore();
+        assertEquals(1, cfs.getLiveSSTables().size());
+
+        SSTableReader sstable = cfs.getLiveSSTables().iterator().next();
+        assertTrue("Column 'b' should be in the flushed SSTable header",
+                   sstableHeaderContainsColumn(sstable, "b"));
+
+        // same but alter before flush
+
+        createTable("CREATE TABLE %s (id int PRIMARY KEY, a text, b text, c text)");
+        execute("INSERT INTO %s (id, a, b, c) VALUES (1, 'a1', 'b1', 'c1')");
+        alterTable("ALTER TABLE %s DROP b");
+        flush(keyspace());
+
+        ColumnFamilyStore cfs2 = getCurrentColumnFamilyStore();
+        assertEquals(1, cfs2.getLiveSSTables().size());
+
+        SSTableReader sstable2 = cfs2.getLiveSSTables().iterator().next();
+        assertTrue("Column 'b' should be in the flushed SSTable header",
+                   sstableHeaderContainsColumn(sstable2, "b"));
+
+        // flush but without dropped column
+
+        createTable("CREATE TABLE %s (id int PRIMARY KEY, a text, b text, c text)");
+        // we are not populating column we go to drop on purpose
+        execute("INSERT INTO %s (id, a, c) VALUES (1, 'a1', 'c1')");
+        alterTable("ALTER TABLE %s DROP b");
+        flush(keyspace());
+
+        ColumnFamilyStore cfs3 = getCurrentColumnFamilyStore();
+        assertEquals(1, cfs3.getLiveSSTables().size());
+
+        SSTableReader sstable3 = cfs3.getLiveSSTables().iterator().next();
+        assertFalse("Column 'b' should not be in the flushed SSTable header",
+                    sstableHeaderContainsColumn(sstable3, "b"));
+    }
+
+    @Test
+    public void testDroppedColumnNotInCompactedSerialisationHeader() throws Throwable
+    {
+        createTable("CREATE TABLE %s (id int PRIMARY KEY, a text, b text, c text)");
+        execute("INSERT INTO %s (id, a, b, c) VALUES (1, 'a1', 'b1', 'c1')");
+        alterTable("ALTER TABLE %s DROP b");
+        flush(keyspace());
+
+        ColumnFamilyStore cfs = getCurrentColumnFamilyStore();
+        assertEquals(1, cfs.getLiveSSTables().size());
+        assertTrue("Column 'b' should be in the flushed SSTable header",
+                   sstableHeaderContainsColumn(cfs.getLiveSSTables().iterator().next(), "b"));
+
+        compact();
+
+        assertEquals(1, cfs.getLiveSSTables().size());
+
+        // column dropped for a header after compaction
+        assertFalse("Dropped column 'b' should not be in the compacted SSTable header",
+                    sstableHeaderContainsColumn(cfs.getLiveSSTables().iterator().next(), "b"));
+
+        assertTrue("Column 'a' should be in the compacted SSTable header",
+                   sstableHeaderContainsColumn(cfs.getLiveSSTables().iterator().next(), "a"));
+        assertTrue("Column 'c' should be in the compacted SSTable header",
+                   sstableHeaderContainsColumn(cfs.getLiveSSTables().iterator().next(), "c"));
+
+        assertRows(execute("SELECT id, a, c FROM %s"), row(1, "a1", "c1"));
+    }
+
+    @Test
+    public void testReAddedColumnInCompactedSerializationHeader() throws Throwable
+    {
+        createTable("CREATE TABLE %s (id int PRIMARY KEY, a text, b text, c text)");
+        execute("INSERT INTO %s (id, a, b, c) VALUES (1, 'a1', 'b1', 'c1')");
+        alterTable("ALTER TABLE %s DROP b");
+        alterTable("ALTER TABLE %s ADD b text");
+        flush(keyspace());
+
+        ColumnFamilyStore cfs = getCurrentColumnFamilyStore();
+        assertEquals(1, cfs.getLiveSSTables().size());
+        assertTrue("Column 'b' should be in the flushed SSTable header",
+                   sstableHeaderContainsColumn(cfs.getLiveSSTables().iterator().next(), "b"));
+
+        compact();
+
+        assertEquals(1, cfs.getLiveSSTables().size());
+
+        // re-added column b
+        assertTrue("Dropped column 'b' should be in the compacted SSTable header",
+                   sstableHeaderContainsColumn(cfs.getLiveSSTables().iterator().next(), "b"));
+
+        assertTrue("Column 'a' should be in the compacted SSTable header",
+                   sstableHeaderContainsColumn(cfs.getLiveSSTables().iterator().next(), "a"));
+        assertTrue("Column 'c' should be in the compacted SSTable header",
+                   sstableHeaderContainsColumn(cfs.getLiveSSTables().iterator().next(), "c"));
+
+        assertRows(execute("SELECT id, a, b, c FROM %s"), row(1, "a1", null, "c1"));
+    }
+
+    /**
+     * Checks if a column with the given name exists in the SSTable header.
+     */
+    private boolean sstableHeaderContainsColumn(SSTableReader sstable, String columnName)
+    {
+        for (ColumnMetadata column : sstable.header.columns())
+        {
+            if (column.name.toString().equals(columnName))
+                return true;
+        }
+        return false;
+    }
+}