Exemple #1
0
def _get_hadoop_configs(node_group):
    cluster = node_group.cluster
    nn_hostname = vu.get_instance_hostname(vu.get_namenode(cluster))
    dirs = _get_hadoop_dirs(node_group)
    confs = {
        'Hadoop': {
            'fs.defaultFS': 'hdfs://%s:9000' % nn_hostname
        },
        'HDFS': {
            'dfs.namenode.name.dir': ','.join(dirs['hadoop_name_dirs']),
            'dfs.namenode.data.dir': ','.join(dirs['hadoop_data_dirs']),
            'dfs.hosts': '%s/dn-include' % HADOOP_CONF_DIR,
            'dfs.hosts.exclude': '%s/dn-exclude' % HADOOP_CONF_DIR
        }
    }

    res_hostname = vu.get_instance_hostname(vu.get_resourcemanager(cluster))
    if res_hostname:
        confs['YARN'] = {
            'yarn.nodemanager.aux-services':
            'mapreduce_shuffle',
            'yarn.resourcemanager.hostname':
            '%s' % res_hostname,
            'yarn.resourcemanager.nodes.include-path':
            '%s/nm-include' % (HADOOP_CONF_DIR),
            'yarn.resourcemanager.nodes.exclude-path':
            '%s/nm-exclude' % (HADOOP_CONF_DIR)
        }
        confs['MapReduce'] = {'mapreduce.framework.name': 'yarn'}

    oozie = vu.get_oozie(cluster)
    if oozie:
        hadoop_cfg = {
            'hadoop.proxyuser.hadoop.hosts': '*',
            'hadoop.proxyuser.hadoop.groups': 'hadoop'
        }
        confs['Hadoop'].update(hadoop_cfg)

        oozie_cfg = o_helper.get_oozie_required_xml_configs(HADOOP_CONF_DIR)
        if c_helper.is_mysql_enabled(cluster):
            oozie_cfg.update(o_helper.get_oozie_mysql_configs())

        confs['JobFlow'] = oozie_cfg

    if c_helper.get_config_value(c_helper.ENABLE_SWIFT.applicable_target,
                                 c_helper.ENABLE_SWIFT.name, cluster):
        swift_configs = {}
        for config in swift.get_swift_configs():
            swift_configs[config['name']] = config['value']

        confs['Hadoop'].update(swift_configs)

    if c_helper.is_data_locality_enabled(cluster):
        confs['Hadoop'].update(th.TOPOLOGY_CONFIG)
        confs['Hadoop'].update(
            {"topology.script.file.name": HADOOP_CONF_DIR + "/topology.sh"})

    return confs, c_helper.get_env_configs()
Exemple #2
0
def validate_existing_ng_scaling(cluster, existing):
    scalable_processes = _get_scalable_processes()
    dn_to_delete = 0
    for ng in cluster.node_groups:
        if ng.id in existing:
            if ng.count > existing[ng.id] and "datanode" in ng.node_processes:
                dn_to_delete += ng.count - existing[ng.id]

            if not set(ng.node_processes).issubset(scalable_processes):
                msg = ("Vanilla plugin cannot scale nodegroup "
                       "with processes: %s")
                raise ex.NodeGroupCannotBeScaled(
                    ng.name, msg % ' '.join(ng.node_processes))

    dn_amount = len(vu.get_datanodes(cluster))
    rep_factor = c_helper.get_config_value('HDFS', 'dfs.replication', cluster)

    if dn_to_delete > 0 and dn_amount - dn_to_delete < rep_factor:
        msg = ("Vanilla plugin cannot shrink cluster because it would be not "
               "enough nodes for replicas (replication factor is %s)")
        raise ex.ClusterCannotBeScaled(cluster.name, msg % rep_factor)