Mailing-List: contact dev-help@hawq.incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hawq.incubator.apache.org
From: jiny2 <git@git.apache.org>
To: dev@hawq.incubator.apache.org
Reply-To: dev@hawq.incubator.apache.org
References: <git-pr-58-incubator-hawq@git.apache.org>
In-Reply-To: <git-pr-58-incubator-hawq@git.apache.org>
Subject: [GitHub] incubator-hawq pull request: add check after submit Hawq AM
 to Yar...
Content-Type: text/plain
Message-Id: <20151029113844.697CCE0509@git1-us-west.apache.org>
Date: Thu, 29 Oct 2015 11:38:44 +0000 (UTC)

Github user jiny2 commented on a diff in the pull request:

    https://github.com/apache/incubator-hawq/pull/58#discussion_r43374690
  
    --- Diff: src/backend/resourcemanager/resourcebroker/resourcebroker_LIBYARN_proc.c ---
    @@ -1267,29 +1267,76 @@ int RB2YARN_connectToYARN(void)
     
     int RB2YARN_registerYARNApplication(void)
     {
    -	int yarnres = FUNCTION_SUCCEEDED;
    +	int retry = 5;
    +	int yarnres = FUNCTION_SUCCEEDED, result = FUNCTION_SUCCEEDED;
    +
     	yarnres = createJob(LIBYARNClient,
     					    YARNAppName.Str,
     						YARNQueueName.Str,
     						&YARNJobID);
    -	if ( yarnres != FUNCTION_SUCCEEDED ) {
    +	if ( yarnres != FUNCTION_SUCCEEDED )
    +	{
     		elog(WARNING, "YARN mode resource broker failed to create application "
     					  "in YARN resource manager. %s",
     					  getErrorMessage());
    +		return yarnres;
     	}
    -	else {
    -		elog(LOG, "YARN mode resource broker created job in YARN resource "
    -				  "manager %s as new application %s assigned to queue %s.",
    -				  YARNJobID,
    -				  YARNAppName.Str,
    -                  YARNQueueName.Str);
     
    -		ResBrokerStartTime = gettime_microsec();
    +	elog(LOG, "YARN mode resource broker created job in YARN resource "
    +			  "manager %s as new application %s assigned to queue %s.",
    +			  YARNJobID,
    +			  YARNAppName.Str,
    +			  YARNQueueName.Str);
    +
    +	/* check if hawq is registered successfully in Hadoop Yarn.
    +	 * if not, kill application from Hadoop Yarn.
    +	 */
    +	LibYarnApplicationReport_t *applicationReport = NULL;
    +	while (retry > 0)
    +	{
    +		result = getApplicationReport(LIBYARNClient, YARNJobID, &applicationReport);
    +		if (result != FUNCTION_SUCCEEDED || applicationReport == NULL)
    +		{
    +			if (retry > 0) {
    +				retry--;
    +				usleep(5*1000*1000L);
    +				continue;
    +			} else {
    +				elog(WARNING, "YARN mode resource broker failed to get application report, "
    +							  "so kill it from Hadoop Yarn.");
    +				result = forceKillJob(LIBYARNClient, YARNJobID);
    +				if (result != FUNCTION_SUCCEEDED)
    +					elog(WARNING, "YARN mode resource broker kill job failed.");
    +				return FUNCTION_FAILED;
    +			}
    +		}
     
    -		elog(LOG, "YARN mode resource broker registered new "
    -				  "YARN application. Start time stamp "UINT64_FORMAT,
    -				  ResBrokerStartTime);
    +		if (applicationReport->progress < 0.5)
    +		{
    +			if (retry > 0) {
    +				retry--;
    +				usleep(5*1000*1000L);
    --- End diff --
    
    this interval is a magic number, use a macro or guc variable ?


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---