I have been trying to start Hive2 Interactive on HDP 2.6.5 that ships with Hive 2.1.0. I have followed the guidelines mentioned in the following articles about how to size and tune LLAP with Yarn https://community.hortonworks.com/articles/149486/llap-sizing-and-setup.html
But LLAP daemon fails to start and Hive interactive gives up after configured retries. Also followed the following steps to troubleshoot and the error doesn't seem to be sizing related. https://community.hortonworks.com/articles/149899/investigating-when-llap-doesnt-start.html Following is my cluster setup in brief: hive.llap.daemon.yarn.container.mb= 10 GB llap_heap_size= 7 GB In-Memory Cache per Daemon= 2 GB Head room= 1 GB (hive.llap.daemon.yarn.container.mb should be = llap_heap_size + In-Memory Cache per Daemon + Headroom) Yarn container max= 14.6 GB, min 1 GB Number of executors per LLAP Daemon= 1 Yarn pre-emption enabled Retries for checking LLAP status= 20 Attached the startup script logs and Yarn container relevant logs. https://pastebin.com/raw/4ywJHp4M https://pastebin.com/raw/AiJkfwUR Following is a snapshot of Slide AM page Application: llap0 Status all containers allocated Total number of containers 5 Create time: 17 Jan 2019 05:40:15 GMT Running since: 17 Jan 2019 05:40:15 GMT Time last flexed: N/A Application storage path: hdfs://hsft/user/hive/.slider/cluster/llap0/database Application configuration path: hdfs://hsft/user/hive/.slider/cluster/llap0/snapshot Component Instances *Component* *Desired* *Actual* *Outstanding Requests* *Failed* *Failed to start* *Placement* LLAP 4 4 0 0 0 slider-appmaster 1 1 0 0 0 Application Container Diagnostics *Container ID* *Component* *State* *Exit Code* *Logs* *Diagnostics* container_e27_1547703090681_0001_01_000002 LLAP 3 -1000 Logs container_e27_1547703090681_0001_01_000003 LLAP 3 -1000 Logs container_e27_1547703090681_0001_01_000004 LLAP 3 -1000 Logs container_e27_1547703090681_0001_01_000005 LLAP 3 -1000 Logs
