Suhas Karanth
06/19/2025, 4:47 AMmetadataStoreBatchingEnabled?RunLLM
06/19/2025, 4:47 AMmetadataStoreBatchingEnabled in Pulsar improves performance by reducing RPC traffic and the number of write transactions, as multiple metadata operations are grouped into a single batch and sent as one ZooKeeper transaction. However, if the batched data size exceeds ZooKeeper's jute.maxbuffer limit, it can cause connection loss or errors, and in some cases, disabling batching may be necessary for stability, especially if large metadata entries are common or if you encounter issues like Packet len ... is out of range! in logs. According to Pulsar documentation and GitHub issue discussion, batching is critical for high-throughput environments but may require tuning or disabling in edge cases.
Would you like more detail on configuration or troubleshooting strategies?
Sources:
- GitHub Issue: ConnectionLoss in broker logs
- Pulsar documentation 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.