Impala returns error: "Bad status for request 5241: TGetOperationStatusResp"

[Георгий Безруких <gb...@gmail.com>]

Hello,

I'm running *CDH 5.2.1.*

When I tried to execute impala query on huge tables via HUE UI I got such
error:
*Bad status for request 5241:
TGetOperationStatusResp*(status=TStatus(errorCode=None,
errorMessage=None, sqlState=None, infoMessages=None, statusCode=0),
operationState=5, errorMessage=None, sqlState=None, errorCode=None)

First, I've tried to execute the same query via *impala-shell* on one of my
worker-nodes. I was confused that error message was:
Query did not have enough memory to get the minimum required buffers.
Backend 3:*Memory Limit Exceeded*

I've checked impalad startup options. Command that gives me `ps aux | grep
impalad` is:
/opt/cloudera/parcels/CDH-5.2.1-1.cdh5.2.1.p0.12/lib/impala/sbin-retail/impalad
--flagfile=/var/run/cloudera-scm-agent/process/7492-impala-IMPALAD/impala-conf/impalad_flags

And there is the content of the above flag_file:
-beeswax_port=21000
-fe_port=21000
-be_port=22000
-llama_callback_port=28000
-hs2_port=21050
-enable_webserver=true
*-mem_limit=128849018880*
-webserver_port=25000
-max_result_cache_size=100000
-state_store_subscriber_port=23000
-statestore_subscriber_timeout_seconds=30
-scratch_dirs=/disk1/impala/impalad,/disk10/impala/impalad,/disk2/impala/impalad,/disk3/impala/impalad,/disk4/impala/impalad,/disk5/impala/impalad,/disk6/impala/impalad,/disk7/impala/impalad,/disk8/impala/impalad,/disk9/impala/impalad,/opt/impala/impalad,/disk11/impala/impalad,/disk12/impala/impalad
*-default_query_options*
-log_filename=impalad
-hostname=my_hostname
-state_store_host=my_host
-local_nodemanager_url=my_host:8042
-llama_host=my_host
-llama_port=15000
-enable_rm=true
-pool_conf_file=pool-acls.txt
-cgroup_hierarchy_path=/var/run/cloudera-scm-agent/cgroups/cpu/hadoop-yarn
-state_store_port=24000
-catalog_service_host=my_host
-catalog_service_port=26000
-local_library_dir=/var/lib/impala/udfs
-llama_max_request_attempts=5
-llama_registration_timeout_secs=30
-llama_registration_wait_secs=3
-fair_scheduler_allocation_path=/var/run/cloudera-scm-agent/process/7494-impala-IMPALAD/impala-conf/fair-scheduler.xml
-llama_site_path=/var/run/cloudera-scm-agent/process/7494-impala-IMPALAD/impala-conf/llama-site.xml
-disk_spill_encryption=false

So, I've tried to change *-default_query_options *to:
*-default_query_options=mem_limit=**128849018880*

Now, if I do 'set;' in HUE or impala-shell I see:
* MEM_LIMIT: [128849018880]*
Before there was value "0". My queries were succesfully executed.

Can you explain, why I have to set mem_limit under parameter "
*-default_query_options*"? I thought that default memory limit for impalad
is set by "mem_limit" option.

-- 
Regards,
Georgy