When at least one job times out while connecting to Kafka, it affects the import of other jobs, slowing down global Routine Load imports. | At least one job times out while connecting to Kafka. | Shared-nothing and shared-storage | Stop or manually pause the job to resolve the issue. | <2.1.9 <3.0.5 | 2.1.9 3.0.5 | #47530 |
User data may be lost after restarting the FE Master. | The job's offset is set to OFFSET_END, and the FE is restarted. | Shared-storage | Change the consumption mode to OFFSET_BEGINNING. | 3.0.2-3.0.4 | 3.0.5 | #46149 |
A large number of small transactions are generated during import, causing compaction to fail and resulting in continuous -235 errors. | Doris consumes data too quickly, or Kafka data flow is in small batches. | Shared-nothing and shared-storage | Pause the Routine Load job and execute the following command: ALTER ROUTINE LOAD FOR jobname FROM kafka ("property.enable.partition.eof" = "false"); | <2.1.8 <3.0.4 | 2.1.8 3.0.4 | #45528, #44949, #39975 |
Kafka third-party library destructor hangs, causing data consumption to fail. | Kafka topic deletion (possibly other conditions). | Shared-nothing and shared-storage | Restart all BE nodes. | <2.1.8 <3.0.4 | 2.1.8 3.0.4 | #44913 |
Routine Load scheduling hangs. | Timeout occurs when FE aborts a transaction in Meta Service. | Shared-storage | Restart the FE node. | <3.0.2 | 3.0.2 | #41267 |
Routine Load restart issue. | Restarting BE nodes. | Shared-nothing and shared-storage | Manually resume the job. | <2.1.7 <3.0.2 | 2.1.7 3.0.2 | #3727 |