Implement partition tracking and rebalance handling in Consumer and BatchConsumer

[sonnes]

Fix "Local: Erroneous state" panic during Kafka rebalancing

Problem

When a Kafka rebalance occurs during batch processing (especially with concurrency > 1), goroutines attempt to commit offsets for partitions they no longer own, causing librdkafka to return "Erroneous state" errors which bubble up as panics.

┌─────────────────────────────────────────────────────────────────────────┐
│                        REBALANCE RACE CONDITION                         │
├─────────────────────────────────────────────────────────────────────────┤
│                                                                         │
│   Pod A              Pod B              Pod C (new)                     │
│   ┌─────┐            ┌─────┐            ┌─────┐                         │
│   │ P0  │            │ P1  │            │     │  ← Spawns               │
│   └──┬──┘            └──┬──┘            └──┬──┘                         │
│      │                  │                  │                            │
│      ▼                  ▼                  │                            │
│  ┌────────┐        ┌────────┐              │                            │
│  │ Long   │        │ Long   │              │                            │
│  │ Batch  │        │ Batch  │              ▼                            │
│  │Process │        │Process │     ┌─────────────────┐                   │
│  └───┬────┘        └───┬────┘     │   REBALANCE     │                   │
│      │                 │          │   TRIGGERED     │                   │
│      │ ◄───────────────┼──────────┤                 │                   │
│      │    Partitions   │          │ P0, P1 REVOKED! │                   │
│      │    Revoked      │          └─────────────────┘                   │
│      │                 │                                                │
│      ▼                 ▼                                                │
│  ┌────────────┐   ┌────────────┐                                        │
│  │  Commit    │   │  Commit    │   ← Partition no longer owned!         │
│  │  Offset    │   │  Offset    │                                        │
│  └─────┬──────┘   └─────┬──────┘                                        │
│        │                │                                               │
│        ▼                ▼                                               │
│   ╔════════════════════════════════╗                                    │
│   ║  💥 Local: Erroneous state 💥  ║                                    │
│   ║     (librdkafka error)         ║                                    │
│   ╚════════════════════════════════╝                                    │
│        │                                                                │
│        ▼                                                                │
│   ┌─────────────┐                                                       │
│   │   PANIC()   │  → Container exits                                    │
│   └─────────────┘                                                       │
│                                                                         │
└─────────────────────────────────────────────────────────────────────────┘

Root Cause

Local: Erroneous state is a librdkafka internal error that occurs when the client attempts a Kafka operation while in an invalid/transition state (e.g., after partitions were revoked during a rebalance).

With batch concurrency > 1: Multiple goroutines are processing simultaneously. When rebalance starts, all goroutines try to store offsets/commit at nearly the same time while the rebalance is in progress. Kafka's transactional state machine cannot handle overlapping operations during a rebalance — one succeeds, the others panic.

Solution

• Track active partitions via a rebalance callback (onPartitionsAssigned / onPartitionsRevoked)
• Skip offset commits for partitions that have been revoked
• Filter batch offsets to only include partitions still assigned before calling StoreOffsets

Changes

File	Change
consumer.go	Added activePartitions map, rebalance callback, isPartitionActive() check in storeMessage()
batch_consumer.go	Same partition tracking + filters storeBatch() to only commit offsets for active partitions
*_test.go	Added tests for rebalance scenarios

[codecov-commenter]

Codecov Report

❌ Patch coverage is 96.29630% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 84.24%. Comparing base (6ce2a14) to head (0500965).
⚠️ Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
xkafka/batch_consumer.go	96.15%	1 Missing and 1 partial ⚠️
xkafka/consumer.go	95.65%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #67      +/-   ##
==========================================
+ Coverage   83.30%   84.24%   +0.93%     
==========================================
  Files          60       60              
  Lines        2810     2316     -494     
==========================================
- Hits         2341     1951     -390     
+ Misses        456      351     -105     
- Partials       13       14       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[sonnes]

Code review

No issues found. Checked for bugs and CLAUDE.md compliance.

🤖 Generated with Claude Code