savanna的启动步骤

savanna为openstack下面部署hadoop的组件，其原理如下：

savanna启动步骤：

1、    service/api.pycreate_cluster(values)

a)    plugin.validate(cluster):验证集群中各功能节点的数目

b)    _provison_cluster

2、    对应1.b service/api.py _provision_cluster

a)    plugin.update_infra(cluster):没有任何动作，直接pass

b)    instance.create_cluster(cluster)

c)     plugin.configure_cluster(cluster)

d)    plugin.start_cluster(cluster)

3、    对应2.b service/instances.pycreate_cluster

a)    “Spawning” _create_instances(cluster)

                   i.           _generate_user_data_script 密钥对

                ii.           _run_instance 启动虚拟机，讲密钥对注入虚拟机

b)    “waiting” _await_instances(cluster)

                   i.           _check_if_up(instance)

1.     remote.execute_command(“hostname”)

c)     volumes.attach(cluster)

d)    “preparing”__configure_instance(cluster)

                   i.           _generate_etc_hosts(cluster) 产生主机名和ip的对应关系

                ii.           remote.write_file_to(‘etc-hosts’,hosts)

              iii.           remote.execute_command(‘sudo mv etc-hosts/etc/hosts’)

              iv.           remote.execute_command(‘chmod 400 .ssh/id_rsa’)

4、    对应2.c  plugins/vanilla/plugin.py configure_cluster

a)    __extract_configs(cluster)

b)    __push_configs_to_nodes(cluster)

c)     __write_hadoop_user_keys(cluster.private_key,utils.get_instances(cluster))

5、    对应4.b plugins/vanilla/plugin.py

a)    files={core-site.xml,mapred-site.xml,hdf-site.xml,savanna-hadoop-init.sh}

b)    sudo chown –R$USER:$USER /etc/hadoop

c)     write_files_to(files)

d)    sudo chmod 0500/tmp/savanna-hadoop-init.sh

e)    sudo/tm/savanna-hadoop-init.sh >> /tmp/savanna-hadoop-init.log 2>&1

f)      针对namenode：write_file_to(‘/etc/hadoop/dn.incl’,utils.generate_fqdn_host_names(utils.get_datanodes(cluster)))

g)    针对jobtracker：write_file_to(‘/etc/hadoop/tt.incl’,utils.generate_fqdn_host_names(utils.get_tasktrackers(cluster)))

6、    对应4.c plugins/vanilla/plugin.py

a)    id_rsa:private_key,authorized_keys:public_key

b)    write_files_to(key)

c)     sudo mkdir –p/home/hadoop/.ssh/

d)    sudo mv id_rsaauthorized_keys /home/hadoop/.shh

e)    sudo chown –Rhadoop:hadoop /home/hadoop/.shh

f)      sudo chmod 600/home/hadoop/.shh/{id_rsa,authorized_keys}

7、    对应2.dplugins/vanilla/plugin.py  start_cluster 通过调用plugins／vanilla／run_scripts.py中的函数来远程执行：

a)    run.format_namenode(remote)

                   i.           sudo su –c ‘hadoop namenode –format’ hadoop

b)    run.start_process(remote,”namenode”)

                   i.           sudo su –c “/usr/sbin/hadoop-daemon.sh startnamenode” hadoop

c)     run.start_process(remote,”secondarynamenode”)

d)    run.start_process(dn.remote,”datanode”)

e)    run.start_process(jt_instance.temote,”jobtracker”)

f)      run.start_process(tt.remote,”tasktracker”)

g)    __set_cluster_info(cluster)启动成功，设置县官信息。