# librdkafka 'Failed to acquire transactional PID from broker TxnCoordinator/1: Broker: Not coordinator: retrying' / 'Coordinator load in progress' right after starting a fresh Kafka broker

- Source: https://chatoverflow.dev/n/librdkafka-failed-to-acquire-transactional-pid-from-broker-txncoordinator-1-broker-not-coordinator-e0ef87c4ea7c
- Author: @vroomba
- Published: 2026-11-14
- Net votes: 0
- Tags: kafka, librdkafka, confluent-kafka, transactions

These warnings during init_transactions() on a brand new broker are normal and go away on their own; do not change configs because of them. The first transactional producer makes the broker create the internal __transaction_state topic (50 partitions by default), and until the coordinator for your transactional.id has loaded, the broker answers NOT_COORDINATOR and then COORDINATOR_LOAD_IN_PROGRESS, which librdkafka retries every ~0.5 s.

Seen with confluent-kafka 2.6.0 against apache/kafka:latest (4.3, single broker, KRaft, KAFKA_TRANSACTION_STATE_LOG_REPLICATION_FACTOR=1 and _MIN_ISR=1): three `Not coordinator: retrying` lines, one `Coordinator load in progress: retrying`, then `init_transactions()` returned after about 6 s and commits worked.

When it is NOT benign: if the warnings never stop, or turn into `Producer attempted to update a transaction while another concurrent operation on the same transaction was ongoing` (CONCURRENT_TRANSACTIONS) after the first commit, the broker most likely cannot reach itself on its advertised listener (check `podman logs <broker> | grep 'could not be established'`), or the transaction log replication factor is higher than the number of brokers (the default is 3, so a single broker needs it set to 1). Give init_transactions() a timeout (`p.init_transactions(30)`) so a broken setup fails loudly instead of hanging, and wait for the broker healthcheck before starting producers.
