Mailing-List: contact dev-help@hawq.incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hawq.incubator.apache.org
From: linwen <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: <20151029091709.F2AAFE1091@git1-us-west.apache.org>
Date: Thu, 29 Oct 2015 09:17:09 +0000 (UTC)

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

    https://github.com/apache/incubator-hawq/pull/58#discussion_r43363199
  
    --- Diff: depends/libyarn/src/libyarnclient/LibYarnClient.cpp ---
    @@ -214,6 +214,60 @@ int LibYarnClient::createJob(string &jobName, string &queue,string &jobId) {
         }
     }
     
    +int LibYarnClient::forceKillJob(string &jobId) {
    +
    +#ifndef MOCKTEST
    +    if ( keepRun ) {
    +        keepRun=false;
    +        void *thrc = NULL;
    +        int rc = pthread_join(heartbeatThread, &thrc);
    +        if ( rc != 0 ) {
    +            LOG(INFO, "LibYarnClient::foreceKillJob, fail to join heart-beat thread. "
    +                      "error code %d", rc);
    +            return FR_FAILED;
    +        }
    +    }
    +#endif
    +
    +    try{
    +        if (jobId != clientJobId) {
    +            throw std::invalid_argument("The jobId is wrong, please check the jobId argument");
    +        }
    +
    +        for (map<int,Container*>::iterator it = jobIdContainers.begin(); it != jobIdContainers. end(); it++) {
    +            ostringstream key;
    +            Container *container = it->second;
    +            key << container->getNodeId().getHost() << ":" << container->getNodeId().getPort();
    +            Token nmToken = nmTokenCache[key.str()];
    +            ((ContainerManagement*)nmClient)->stopContainer((*container), nmToken);
    +            LOG(INFO,"LibYarnClient::foreceKillJob, container:%d are stopped",container->getId().getId());
    +        }
    +
    +        ((ApplicationClient*) appClient)->forceKillApplication(clientAppId);
    +        LOG(INFO, "LibYarnClient::foreceKillJob, forceKillApplication");
    +
    +        for (map<int,Container*>::iterator it = jobIdContainers.begin(); it != jobIdContainers.end(); it++) {
    +            LOG(INFO,"LibYarnClient::foreceKillJob, container:%d in jobIdContainers are delete",it->second->getId().getId());
    +            delete it->second;
    +            it->second = NULL;
    +        }
    +        jobIdContainers.clear();
    +        activeFailContainerIds.clear();
    +        return FR_SUCCEEDED;
    +    } catch(std::exception& e){
    +        stringstream errorMsg;
    +        errorMsg << "LibYarnClient::forceKillJob, Catch the Exception:" << e.what();
    +        setErrorMessage(errorMsg.str());
    +        return FR_FAILED;
    +    } catch (...) {
    +        stringstream errorMsg;
    +        errorMsg << "LibYarnClient::forceKillJob, catch unexpected exception.";
    +        setErrorMessage(errorMsg.str());
    +        return FR_FAILED;
    +    }
    +}
    +
    --- End diff --
    
    yes. in this case no container are returned to hawq. what I am thinking is, in some other cases in future, forceKillApplication() is called when hawq works normally, then containers should be returned. 


---
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.
---