AMBARI-22346. Beeline connection hangs for longer time when connection to HS2 with metastore DB down (aonishuk)
Project: http://git-wip-us.apache.org/repos/asf/ambari/repo Commit: http://git-wip-us.apache.org/repos/asf/ambari/commit/98915a11 Tree: http://git-wip-us.apache.org/repos/asf/ambari/tree/98915a11 Diff: http://git-wip-us.apache.org/repos/asf/ambari/diff/98915a11 Branch: refs/heads/branch-feature-AMBARI-21674 Commit: 98915a114769c7ddd590854ccf0ab33beabddcf3 Parents: 46ad6c6 Author: Andrew Onishuk <[email protected]> Authored: Fri Nov 10 13:28:27 2017 +0200 Committer: Andrew Onishuk <[email protected]> Committed: Fri Nov 10 13:28:27 2017 +0200 ---------------------------------------------------------------------- .../HIVE/0.12.0.2.0/package/alerts/alert_hive_thrift_port.py | 3 ++- .../1.2.1/package/scripts/alerts/alert_spark_thrift_port.py | 8 +++++++- .../2.0.0/package/scripts/alerts/alert_spark2_thrift_port.py | 8 +++++++- 3 files changed, 16 insertions(+), 3 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/ambari/blob/98915a11/ambari-server/src/main/resources/common-services/HIVE/0.12.0.2.0/package/alerts/alert_hive_thrift_port.py ---------------------------------------------------------------------- diff --git a/ambari-server/src/main/resources/common-services/HIVE/0.12.0.2.0/package/alerts/alert_hive_thrift_port.py b/ambari-server/src/main/resources/common-services/HIVE/0.12.0.2.0/package/alerts/alert_hive_thrift_port.py index 3560bf8..6db92b0 100644 --- a/ambari-server/src/main/resources/common-services/HIVE/0.12.0.2.0/package/alerts/alert_hive_thrift_port.py +++ b/ambari-server/src/main/resources/common-services/HIVE/0.12.0.2.0/package/alerts/alert_hive_thrift_port.py @@ -28,6 +28,7 @@ from resource_management.libraries.functions import format from resource_management.libraries.functions import get_kinit_path from ambari_commons.os_check import OSConst from ambari_commons.os_family_impl import OsFamilyFuncImpl, OsFamilyImpl +from resource_management.core.signal_utils import TerminateStrategy OK_MESSAGE = "TCP OK - {0:.3f}s response on port {1}" CRITICAL_MESSAGE = "Connection failed on host {0}:{1} ({2})" @@ -271,7 +272,7 @@ def execute(configurations={}, parameters={}, host_name=None): start_time = time.time() try: - Execute(cmd, user=hiveuser, timeout=30) + Execute(cmd, user=hiveuser, timeout=30, timeout_kill_strategy=TerminateStrategy.KILL_PROCESS_TREE) total_time = time.time() - start_time result_code = 'OK' label = OK_MESSAGE.format(total_time, port) http://git-wip-us.apache.org/repos/asf/ambari/blob/98915a11/ambari-server/src/main/resources/common-services/SPARK/1.2.1/package/scripts/alerts/alert_spark_thrift_port.py ---------------------------------------------------------------------- diff --git a/ambari-server/src/main/resources/common-services/SPARK/1.2.1/package/scripts/alerts/alert_spark_thrift_port.py b/ambari-server/src/main/resources/common-services/SPARK/1.2.1/package/scripts/alerts/alert_spark_thrift_port.py index 3f80fd9..6874cb4 100644 --- a/ambari-server/src/main/resources/common-services/SPARK/1.2.1/package/scripts/alerts/alert_spark_thrift_port.py +++ b/ambari-server/src/main/resources/common-services/SPARK/1.2.1/package/scripts/alerts/alert_spark_thrift_port.py @@ -27,6 +27,7 @@ from resource_management.libraries.script.script import Script from resource_management.libraries.functions import get_kinit_path from resource_management.core.resources import Execute from resource_management.core import global_lock +from resource_management.core.signal_utils import TerminateStrategy stack_root = Script.get_stack_root() @@ -140,7 +141,12 @@ def execute(configurations={}, parameters={}, host_name=None): start_time = time.time() try: - Execute(cmd, user=hiveruser, path=[beeline_cmd], timeout=CHECK_COMMAND_TIMEOUT_DEFAULT) + Execute(cmd, + user=hiveruser, + path=[beeline_cmd], + timeout=CHECK_COMMAND_TIMEOUT_DEFAULT, + timeout_kill_strategy=TerminateStrategy.KILL_PROCESS_TREE, + ) total_time = time.time() - start_time result_code = 'OK' label = OK_MESSAGE.format(total_time, port) http://git-wip-us.apache.org/repos/asf/ambari/blob/98915a11/ambari-server/src/main/resources/common-services/SPARK2/2.0.0/package/scripts/alerts/alert_spark2_thrift_port.py ---------------------------------------------------------------------- diff --git a/ambari-server/src/main/resources/common-services/SPARK2/2.0.0/package/scripts/alerts/alert_spark2_thrift_port.py b/ambari-server/src/main/resources/common-services/SPARK2/2.0.0/package/scripts/alerts/alert_spark2_thrift_port.py index 9e1afea..d3660de 100644 --- a/ambari-server/src/main/resources/common-services/SPARK2/2.0.0/package/scripts/alerts/alert_spark2_thrift_port.py +++ b/ambari-server/src/main/resources/common-services/SPARK2/2.0.0/package/scripts/alerts/alert_spark2_thrift_port.py @@ -27,6 +27,7 @@ from resource_management.libraries.script.script import Script from resource_management.libraries.functions import get_kinit_path from resource_management.core.resources import Execute from resource_management.core import global_lock +from resource_management.core.signal_utils import TerminateStrategy stack_root = Script.get_stack_root() @@ -141,7 +142,12 @@ def execute(configurations={}, parameters={}, host_name=None): start_time = time.time() try: - Execute(cmd, user=hiveruser, path=[beeline_cmd], timeout=CHECK_COMMAND_TIMEOUT_DEFAULT) + Execute(cmd, + user=hiveruser, + path=[beeline_cmd], + timeout=CHECK_COMMAND_TIMEOUT_DEFAULT, + timeout_kill_strategy=TerminateStrategy.KILL_PROCESS_TREE + ) total_time = time.time() - start_time result_code = 'OK' label = OK_MESSAGE.format(total_time, port)
