Open
Description
The following error can be seen when the fluentd client attempts to write a message that is too big: write: message too long or short write and can be attributed to this line of code.
Specifically, this has been observed when eventpb.QueryExecute events are logged with large placeholder values, resulting in conn.Write in fluent_client.go to error. To fix, I see two potential options:
- Add some truncation logic to truncate the data in these events if they are too big
- Break the fluent sink writes up into small chunks to avoid message too long errors.
Jira issue: CRDB-44655