Do you know in how much time does this timeout occur? There might be some tuning needed to increase a timeout. Also, I think this (S3 specifically) has been seen before... So you might find a solution within the mailing list archives. Did you try looking there?
From: Alan Höng Sent: Wednesday, September 20, 8:46 AM Subject: User client timeout with results > 2M rows To: [email protected] Hello, I'm getting errors when trying to fetch results from drill with queries that evaluate to bigger tables. Surprisingly it works like a charm if the returned table has less than 2M rows. It also seems like the query is executed and finishes successfully.... I'm querying parquet files with GZIP compression on S3. I'm running drill in distributed mode with zookeeper. I use version 1.9 from the container available on dockerhub "harisekhon/apache-drill:1.9". I'm using the pydrill package which uses the rest api to submit queries and gather results. I get the following error message from the client: TransportError(500, '{\n "errorMessage" : "CONNECTION ERROR: Connection /172.19.0.3:52382<http://172.19.0.3:52382> <--> ef53daab0ef8/172.19.0.6:31010<http://172.19.0.6:31010> (user client) closed unexpectedly. Drillbit down?\\n\\n\\n[Error Id: 6a19835b-2325-431e-9bad-dde8f1d3c192 ]"\n}' I would appreciate any help with this. Best Alan Höng
