CASSANALYTICS-60: CDC support for Cassandra 5.0 commit logs by lukasz-antoniak · Pull Request #175 · apache/cassandra-analytics

lukasz-antoniak · 2026-03-03T10:30:24Z

Fixes Analytics reader of Cassandra 5.0 commit logs (part of CASSANALYTICS-60).

Using environment variable CASSANDRA_VERSION=5.0.0 (or 4.0.0), developers can run CDC unit tests with different Cassandra versions.

Refactoring all tests to @ParameterizedTest requires more efforts, and will be performed in follow-up PR. Static variables like CdcTests.BRIDGE are references from multiple places in the code.

lukasz-antoniak · 2026-03-03T10:31:38Z

cassandra-analytics-cdc-codec/src/main/java/org/apache/cassandra/cdc/kafka/KafkaPublisher.java

    protected final KafkaStats kafkaStats;

-    public KafkaPublisher(TopicSupplier topicSupplier,
+    public KafkaPublisher(String version,


API change between Sidecar and Analytics.

Thanks for calling it out.
Thinking from API design perspective, how about passing CassandraVersion, instead of String version value? It, then, ensures the version is always valid. Sidecar anyways has access to CassandraVersion. It would be clear for Sidecar to create the version object.

lukasz-antoniak · 2026-03-03T10:33:39Z

...andra-analytics-common/src/main/java/org/apache/cassandra/cdc/msg/RangeTombstoneBuilder.java

 * @param <T>
 */
-public abstract class RangeTombstoneBuilder<T>
+public interface RangeTombstoneBuilder<T>


Classloader clash when FourZeroRangeTombstoneBuilder is loaded by bridge classloader and RangeTombstoneBuilder (abstract parent class) by application classloader.

maybe add the note in code so it stays as interface.

yifan-c · 2026-03-04T03:36:40Z

cassandra-analytics-cdc-codec/src/main/java/org/apache/cassandra/cdc/kafka/KafkaPublisher.java

    protected final KafkaStats kafkaStats;

-    public KafkaPublisher(TopicSupplier topicSupplier,
+    public KafkaPublisher(String version,


Thanks for calling it out.
Thinking from API design perspective, how about passing CassandraVersion, instead of String version value? It, then, ensures the version is always valid. Sidecar anyways has access to CassandraVersion. It would be clear for Sidecar to create the version object.

yifan-c · 2026-03-04T03:43:34Z

cassandra-analytics-cdc/src/main/java/org/apache/cassandra/bridge/CdcBridgeFactory.java

        }
    }
+
+    public static <T> T executeActionOnBridgeClassLoader(@NotNull CassandraVersion version, Throwing.Function<ClassLoader, T> action)


nit: add @VisibleForTesting

yifan-c · 2026-03-04T03:48:42Z

...andra-analytics-common/src/main/java/org/apache/cassandra/cdc/msg/RangeTombstoneBuilder.java

 * @param <T>
 */
-public abstract class RangeTombstoneBuilder<T>
+public interface RangeTombstoneBuilder<T>


maybe add the note in code so it stays as interface.

yifan-c

Noticed the PR is marked as Draft. I will re-review if there are new changes.
Please share the CI link too.

lukasz-antoniak · 2026-03-09T09:11:16Z

@yifan-c, refactored all CDC tests to @ParameterizedTest and replaced static bridge references. Opening the PR. I will not squash commits to make it easier to review. Once positive review is given, I will squash, rebase and add change log.

yifan-c

some minor comments. newer commits looks good overall.

yifan-c · 2026-03-10T04:05:46Z

cassandra-analytics-cdc/src/test/java/org/apache/cassandra/cdc/test/TestUtils.java

 */

-package org.apache.cassandra.cdc;
+package org.apache.cassandra.cdc.test;


What is the test package used for? Seems to group the test foundation classes, but CdcTester is not in this new package.

yifan-c · 2026-03-10T05:41:08Z

cassandra-analytics-cdc/src/test/java/org/apache/cassandra/cdc/test/CdcBridgeTestInjector.java

+ */
+public class CdcBridgeTestInjector implements InvocationInterceptor
+{
+    public void interceptTestTemplateMethod(Invocation<Void> invocation,


Please add @Override

yifan-c · 2026-03-10T05:42:22Z

cassandra-analytics-cdc/src/test/java/org/apache/cassandra/cdc/test/CdcBridgeTestInjector.java

+ * Setups all fields of {@code CdcTestBase} based on {@code CassandraVersion} method parameter
+ * before each test method execution.
+ */
+public class CdcBridgeTestInjector implements InvocationInterceptor


lukasz-antoniak · 2026-03-10T10:58:09Z

...ur-zero-bridge/src/main/java/org/apache/cassandra/db/commitlog/BufferingCommitLogReader.java

-                this.position = (int) log.maxOffset();
-            }
+            // TODO(lantoniak): Does not this change break markers in the stored state and prevents commit log from re-read?
+//            if (statusTracker.shouldContinue())


I think this change form commit 34c6c38 broke trunk as well. @jyothsnakonisa, can you also have a look?

The issue is being fixed in PR #179.

Patched by Lukasz Antoniak; Reviewed by Yifan Cai for CASSANALYTICS-60

jyothsnakonisa

Looks good overall, some minor comments. I have local cdc setup, lets merge it after I test it with that setup on a 5.0 cassandra version

jyothsnakonisa · 2026-03-12T09:15:06Z

...andra-analytics-common/src/main/java/org/apache/cassandra/cdc/msg/RangeTombstoneBuilder.java

+    RangeTombstone buildTombstone(List<Value> start, boolean isStartInclusive, List<Value> end, boolean isEndInclusive);
+    boolean canBuild();
+    RangeTombstone build();
+    void add(T marker);


Can you add documentation to these interface methods?

They are little self-explanatory and not sure if JavaDoc will add much more value. Do you have something on your mind? This PR has changed abstract class to interface and those methods did not have explicit JavaDoc before.

jyothsnakonisa · 2026-03-12T09:22:43Z

...ur-zero-bridge/src/main/java/org/apache/cassandra/cdc/msg/AbstractRangeTombstoneBuilder.java

+    RangeTombstone rangeTombstone;
+    boolean expectOpen = true;


May be make them protected?

jyothsnakonisa · 2026-03-12T09:56:26Z

cassandra-bridge/src/main/java/org/apache/cassandra/bridge/CdcBridge.java

+    public abstract CommitLogInstance createCommitLogInstance(Path path);
+
+    public abstract TableIdLookup internalTableIdLookup();
+


Can you please add java docs?

jyothsnakonisa · 2026-03-12T09:58:43Z

...r-zero-bridge/src/main/java/org/apache/cassandra/bridge/AbstractCdcBridgeImplementation.java


+    public CommitLogInstance createCommitLogInstance(Path path)
+    {
+        return new FourZeroCommitLog(path);


As per my understanding commit log implementation has not changed between 4.0 and 5.0 but just double checking that returning FourZeroCommitLog is correct or not

jyothsnakonisa · 2026-03-12T10:14:26Z

...r-zero-bridge/src/main/java/org/apache/cassandra/bridge/AbstractCdcBridgeImplementation.java

+        return new TableIdLookup()
+        {
+            @Nullable
+            public UUID lookup(String keyspace, String table) throws NoSuchElementException
+            {
+                TableMetadata tm = Schema.instance.getTableMetadata(keyspace, table);
+                if (tm == null)
+                {
+                    throw new NoSuchElementException();
+                }
+                return tm.id.asUUID();
+            }
+        };


You are creating a new instance of TableIdLookup for every invocation of this method, instead can you have a single instance and return it on every call?

Good catch.

jyothsnakonisa · 2026-03-12T10:21:19Z

...ndra-five-zero-bridge/src/main/java/org/apache/cassandra/bridge/CdcBridgeImplementation.java

        DatabaseDescriptor.setCommitLogSyncGroupWindow(30);
        DatabaseDescriptor.setCommitLogSegmentSize(commitLogSegmentSize);
        DatabaseDescriptor.getRawConfig().commitlog_total_space = new DataStorageSpec.IntMebibytesBound(1024);
+        DatabaseDescriptor.setCommitLogWriteDiskAccessMode(Config.DiskAccessMode.direct);


Can u check if it's okay to hardcode it, curious if this changes based on hardware

We do not seem to have an option to skip this setting. Only mmap and direct are supported based on 5.0.5 code. direct seemed like a safer option. AbstractCommitLogSegmentManager#createSegmentBuilder uses other modes for encrypted or compressed commit logs automatically.

lukasz-antoniak commented Mar 3, 2026

View reviewed changes

yifan-c reviewed Mar 4, 2026

View reviewed changes

yifan-c approved these changes Mar 4, 2026

View reviewed changes

lukasz-antoniak marked this pull request as ready for review March 9, 2026 09:11

yifan-c reviewed Mar 10, 2026

View reviewed changes

lukasz-antoniak force-pushed the CASSANALYTICS-60 branch from db23c2f to b5b31a4 Compare March 10, 2026 10:51

lukasz-antoniak commented Mar 10, 2026

View reviewed changes

CASSANALYTICS-60: CDC support for Cassandra 5.0 commit logs

fffdf31

Patched by Lukasz Antoniak; Reviewed by Yifan Cai for CASSANALYTICS-60

lukasz-antoniak force-pushed the CASSANALYTICS-60 branch from b5b31a4 to fffdf31 Compare March 12, 2026 06:30

jyothsnakonisa approved these changes Mar 12, 2026

View reviewed changes

Apply review comments

e35909f

		public abstract CommitLogInstance createCommitLogInstance(Path path);

		public abstract TableIdLookup internalTableIdLookup();

Conversation

lukasz-antoniak commented Mar 3, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lukasz-antoniak Mar 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

yifan-c left a comment

Choose a reason for hiding this comment

Uh oh!

lukasz-antoniak commented Mar 9, 2026

Uh oh!

yifan-c left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jyothsnakonisa left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lukasz-antoniak Mar 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

lukasz-antoniak Mar 3, 2026 •

edited

Loading

lukasz-antoniak Mar 12, 2026 •

edited

Loading