YARN-4087. Followup fixes after YARN-2019 regarding RM behavior when

state-store error occurs. Contributed by Jian He

(cherry picked from commit 9b78e6e33d)
This commit is contained in:
Xuan 2015-09-07 17:45:47 -07:00
parent d983e55714
commit a0b7ef15d0
4 changed files with 17 additions and 8 deletions

View File

@ -837,6 +837,9 @@ Release 2.7.2 - UNRELEASED
YARN-4105. Capacity Scheduler headroom for DRF is wrong (Chang Li via
jlowe)
YARN-4087. Followup fixes after YARN-2019 regarding RM behavior when
state-store error occurs. (Jian He via xgong)
Release 2.7.1 - 2015-07-06
INCOMPATIBLE CHANGES

View File

@ -402,7 +402,7 @@ public class YarnConfiguration extends Configuration {
public static final boolean DEFAULT_RM_RECOVERY_ENABLED = false;
public static final String YARN_FAIL_FAST = YARN_PREFIX + "fail-fast";
public static final boolean DEFAULT_YARN_FAIL_FAST = true;
public static final boolean DEFAULT_YARN_FAIL_FAST = false;
public static final String RM_FAIL_FAST = RM_PREFIX + "fail-fast";

View File

@ -349,9 +349,12 @@
<property>
<description>Should YARN fail fast if it encounters any errors.
This is a global config for all other components including RM,NM etc.
If no value is set for component-specific config (e.g yarn.resourcemanager.fail-fast),
this value will be the default.
</description>
<name>yarn.fail-fast</name>
<value>true</value>
<value>false</value>
</property>
<property>

View File

@ -45,6 +45,7 @@ import org.apache.hadoop.yarn.api.records.ApplicationId;
import org.apache.hadoop.yarn.api.records.ApplicationSubmissionContext;
import org.apache.hadoop.yarn.api.records.ReservationId;
import org.apache.hadoop.yarn.api.records.impl.pb.ApplicationSubmissionContextPBImpl;
import org.apache.hadoop.yarn.conf.HAUtil;
import org.apache.hadoop.yarn.conf.YarnConfiguration;
import org.apache.hadoop.yarn.event.AsyncDispatcher;
import org.apache.hadoop.yarn.event.Dispatcher;
@ -1013,18 +1014,20 @@ public abstract class RMStateStore extends AbstractService {
*/
protected void notifyStoreOperationFailed(Exception failureCause) {
LOG.error("State store operation failed ", failureCause);
if (failureCause instanceof StoreFencedException) {
if (HAUtil.isHAEnabled(getConfig())) {
LOG.warn("State-store fenced ! Transitioning RM to standby");
updateFencedState();
Thread standByTransitionThread =
new Thread(new StandByTransitionThread());
standByTransitionThread.setName("StandByTransitionThread Handler");
standByTransitionThread.start();
} else if (YarnConfiguration.shouldRMFailFast(getConfig())) {
LOG.fatal("Fail RM now due to state-store error!");
rmDispatcher.getEventHandler().handle(
new RMFatalEvent(RMFatalEventType.STATE_STORE_OP_FAILED,
failureCause));
} else {
if (YarnConfiguration.shouldRMFailFast(getConfig())) {
rmDispatcher.getEventHandler().handle(
new RMFatalEvent(RMFatalEventType.STATE_STORE_OP_FAILED,
failureCause));
}
LOG.warn("Skip the state-store error.");
}
}