Alexander Behm created IMPALA-6476:

             Summary: TestKrpcMemUsage.test_krpc_deferred_memory_usage fails on 
release build
                 Key: IMPALA-6476
             Project: IMPALA
          Issue Type: Bug
          Components: Infrastructure
            Reporter: Alexander Behm
            Assignee: Lars Volker

Jenkins console:
08:04:59 ==================================== ERRORS 
08:04:59  ERROR at setup of 
TestKrpcMemUsage.test_krpc_deferred_memory_usage[exec_option: {'batch_size': 0, 
'num_nodes': 0, 'disable_codegen_rows_threshold': 0, 'disable_codegen': False, 
'abort_on_error': 1, 'exec_single_node_rows_threshold': 0} | table_format: 
08:04:59 common/ in setup_method
08:04:59     self._start_impala_cluster(cluster_args)
08:04:59 common/ in _start_impala_cluster
08:04:59     check_call(cmd + options, close_fds=True)
08:04:59 /usr/lib64/python2.6/ in check_call
08:04:59     raise CalledProcessError(retcode, cmd)
08:04:59 E   CalledProcessError: Command 
 '--cluster_size=3', '--num_coordinators=3', 
 '--log_level=1', '--impalad_args="--use_krpc 
--stress_datastream_recvr_delay_ms=1000" ']' returned non-zero exit status 1
08:04:59 ---------------------------- Captured stdout setup 

The test uses stress options that are only compiled in debug builds.

See global-flags.h:
// Stress options that are only enabled in debug builds for testing.
#ifndef NDEBUG
DEFINE_int32(stress_fn_ctx_alloc, 0, "A stress option which causes memory 
allocations "
    "in function contexts to fail once every n allocations where n is the value 
of this "
    "flag. Effective in debug builds only.");
DEFINE_int32(stress_datastream_recvr_delay_ms, 0, "A stress option that causes 
data "
    "stream receiver registration to be delayed. Effective in debug builds 
DEFINE_bool(skip_file_runtime_filtering, false, "Skips file-based runtime 
filtering for"
    "testing purposes. Effective in debug builds only.");
DEFINE_int32(fault_injection_rpc_delay_ms, 0, "A fault injection option that 
causes "
    "rpc server handling to be delayed to trigger an RPC timeout on the caller 
side. "
    "Effective in debug builds only.");
DEFINE_int32(fault_injection_rpc_type, 0, "A fault injection option that 
specifies "
    "which rpc call will be injected with the delay. Effective in debug builds 
DEFINE_int32(fault_injection_rpc_exception_type, 0, "A fault injection option 
that "
    "specifies the exception to be thrown in the caller side of an RPC call. 
Effective "
    "in debug builds only");
DEFINE_int32(stress_scratch_write_delay_ms, 0, "A stress option which causes 
writes to "
    "scratch files to be to be delayed to simulate slow writes.");
DEFINE_bool(thread_creation_fault_injection, false, "A fault injection option 
that "
    " causes calls to Thread::Create() to fail randomly 1% of the time on 
eligible "
    " codepaths. Effective in debug builds only.");
DEFINE_int32(stress_catalog_init_delay_ms, 0, "A stress option that injects 
extra delay"
    " in milliseconds when initializing an impalad's local catalog replica. 
Delay <= 0"
    " inject no delay.");

See for how a similar issue was fixed 

This message was sent by Atlassian JIRA

Reply via email to