dataflow/ERR/2023_013
Dataflow job reading from spanner did not fail due to deadline exceeded error.
Product: Dataflow
Rule class: ERR - Something that is very likely to be wrong
Description
The Dataflow job reading from spanner, failed with deadline exceeded error
a. If the job is not already using shuffle service, try enabling shuffle service. b. Timeouts are caused by too large work items. To make sure there are no timeouts, it is recommended trying to tweak some configurations in Spanner Read such as “maxPartittions” and “partitionSizeBytes”. c. If the CPU utilization is high, might have to scale up the metadata database.