fix: stop persisting CDC processor read marker across restarts

[divyagovindaiah]

Summary

GraphLogProcessor.init() registers the transaction-log processor with setProcessorIdentifier("janusgraph-cdc-processor"). An identified log processor persists its read marker in JanusGraph and resumes from it on every subsequent restart — silently ignoring setStartTime() after the very first run.

Root cause

Every JanusGraph stop/start replays old/backlogged transaction-log entries because of the persisted marker. Vertex property lookups (e.g. IL_FUNC_OBJECT_TYPE) performed on vertex handles bound to a replayed historical transaction come back empty — this appears to be a JanusGraph/storage-backend limitation when reading properties outside the specific transaction's own change-set. SunbirdLegacyMessageConverter falls back to vertex.label() (JanusGraph's generic "vertex" label) when that lookup is empty, which then hits the existing filter:

if ("vertex".equalsIgnoreCase(objectType)) {
    logger.debug("Skipping event for vertex {} with objectType='vertex' (no IL_FUNC_OBJECT_TYPE set)", vertex.id());
    return null;
}

Net effect: after any JanusGraph restart, every backlogged event is filtered out, and CDC silently stops producing anything until the backlog fully drains (which itself takes a long time and can appear to "never recover" on clusters with a large accumulated transaction-log history, e.g. after a disaster-recovery restore).

Reproduced and confirmed via debug logging on a live cluster: the marker loaded on restart showed Loaded identified ReadMarker start time <stale timestamp> instead of now, and every processed message hit Event filtered by converter for node <id> — 100% filter rate, including on messages only minutes old (i.e. not a formatting/legacy-data issue, purely a replay-vs-live distinction).

Fix

Remove .setProcessorIdentifier(...). Without an identifier, the log processor never persists a marker, so it always starts fresh from setStartTime() (now - 1 minute) on every restart. This trades replaying the exact outage window for CDC actually continuing to work after a restart — the correct tradeoff for a live content-indexing pipeline where a live/current pipeline matters more than a complete historical replay.

Test plan

• Deploy to a test JanusGraph instance with the CDC extension enabled
• Create/update content, confirm CDC event appears in the configured sink
• Stop and restart the JanusGraph pod
• Create/update new content immediately after restart, confirm the event still appears (this is the case that was broken before this fix)
• Confirm no persisted marker is created for janusgraph-cdc-processor in the backend after this change

Summary by CodeRabbit

• Bug Fixes
  • Improved change-data-capture recovery after restarts.
  • CDC now replays events from the configured start time instead of resuming from a persisted read position, helping prevent missed backlog events.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 67d19d32-f22f-4b51-b5f9-266a15e7d26a

📥 Commits

Reviewing files that changed from the base of the PR and between 6acd451 and ad918b9.

📒 Files selected for processing (1)

• janusgraph-cdc-extension/src/main/java/org/sunbird/janusgraph/cdc/GraphLogProcessor.java

---

📝 Walkthrough

Walkthrough

GraphLogProcessor now registers the CDC log processor without a persistent processor identifier, so initialization starts replay from the configured start time instead of resuming a persisted read marker.

Changes

CDC restart behavior

Layer / File(s)	Summary
Processor registration and replay behavior janusgraph-cdc-extension/src/main/java/org/sunbird/janusgraph/cdc/GraphLogProcessor.java	GraphLogProcessor registers the log processor using only LOG_IDENTIFIER and documents that restarts begin from the configured setStartTime().

Estimated code review effort: 2 (Simple) | ~10 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title is concise and accurately summarizes the main change: stopping CDC read-marker persistence across restarts.
Description check	✅ Passed	It covers the summary, root cause, fix, and testing plan; the issue reference and dependency list are the main omissions.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}