def get_hadoop_conf_dir():
"""
Gets the shared hadoop conf directory using:
1. Start with /etc/hadoop/conf
2. When the stack is greater than HDP-2.2, use /usr/hdp/current/hadoop-client/conf
3. Only when doing a RU and HDP-2.3 or higher, use the value as computed
by conf-select. This is in the form /usr/hdp/VERSION/hadoop/conf to make sure
the configs are written in the correct place
"""
config = Script.get_config()
hadoop_conf_dir = "/etc/hadoop/conf"
if Script.is_hdp_stack_greater_or_equal("2.2"):
from resource_management.libraries.functions.default import default
hadoop_conf_dir = "/usr/hdp/current/hadoop-client/conf"
direction = default("/commandParams/upgrade_direction", None)
ver = default("/commandParams/version", None)
stack_name = default("/hostLevelParams/stack_name", None)
if direction and ver and stack_name and Script.is_hdp_stack_greater_or_equal(
"2.3"):
select(stack_name, "hadoop", ver)
hadoop_conf_dir = "/usr/hdp/{0}/hadoop/conf".format(ver)
return hadoop_conf_dir
def get_hadoop_dir(target):
"""
Return the hadoop shared directory in the following override order
1. Use default for 2.1 and lower
2. If 2.2 and higher, use /usr/hdp/current/hadoop-client/{target}
3. If 2.2 and higher AND for an upgrade, use /usr/hdp/<version>/hadoop/{target}.
However, if the upgrade has not yet invoked hdp-select, return the current
version of the component.
:target: the target directory
"""
if not target in HADOOP_DIR_DEFAULTS:
raise Fail("Target {0} not defined".format(target))
hadoop_dir = HADOOP_DIR_DEFAULTS[target]
if Script.is_hdp_stack_greater_or_equal("2.2"):
hadoop_dir = HADOOP_DIR_TEMPLATE.format("current", "hadoop-client", target)
stack_info = _get_upgrade_stack()
if stack_info is not None:
stack_version = stack_info[1]
# determine if hdp-select has been run and if not, then use the current
# hdp version until this component is upgraded
current_hdp_version = get_role_component_current_hdp_version()
if current_hdp_version is not None and stack_version != current_hdp_version:
stack_version = current_hdp_version
hadoop_dir = HADOOP_DIR_TEMPLATE.format(stack_version, "hadoop", target)
return hadoop_dir
def get_hadoop_conf_dir(force_latest_on_upgrade=False):
"""
Gets the shared hadoop conf directory using:
1. Start with /etc/hadoop/conf
2. When the stack is greater than HDP-2.2, use /usr/hdp/current/hadoop-client/conf
3. Only when doing a RU and HDP-2.3 or higher, use the value as computed
by conf-select. This is in the form /usr/hdp/VERSION/hadoop/conf to make sure
the configs are written in the correct place. However, if the component itself has
not yet been upgraded, it should use the hadoop configs from the prior version.
This will perform an hdp-select status to determine which version to use.
:param force_latest_on_upgrade: if True, then force the returned path to always
be that of the upgrade target version, even if hdp-select has not been called. This
is primarily used by hooks like before-ANY to ensure that hadoop environment
configurations are written to the correct location since they are written out
before the hdp-select/conf-select would have been called.
"""
hadoop_conf_dir = "/etc/hadoop/conf"
if Script.is_hdp_stack_greater_or_equal("2.2"):
hadoop_conf_dir = "/usr/hdp/current/hadoop-client/conf"
stack_info = hdp_select._get_upgrade_stack()
# if upgrading to >= HDP 2.3
if stack_info is not None and Script.is_hdp_stack_greater_or_equal(
"2.3"):
stack_name = stack_info[0]
stack_version = stack_info[1]
# determine if hdp-select has been run and if not, then use the current
# hdp version until this component is upgraded
if not force_latest_on_upgrade:
current_hdp_version = hdp_select.get_role_component_current_hdp_version(
)
if current_hdp_version is not None and stack_version != current_hdp_version:
stack_version = current_hdp_version
# only change the hadoop_conf_dir path, don't conf-select this older version
hadoop_conf_dir = "/usr/hdp/{0}/hadoop/conf".format(stack_version)
# ensure the new HDP stack is conf-selected, but only if it exists
# there are cases where hadoop might not be installed, such as on a host with only ZK
if os.path.exists(hadoop_conf_dir):
select(stack_name, "hadoop", stack_version)
return hadoop_conf_dir
def get_hadoop_dir(target, force_latest_on_upgrade=False):
"""
Return the hadoop shared directory in the following override order
1. Use default for 2.1 and lower
2. If 2.2 and higher, use /usr/hdp/current/hadoop-client/{target}
3. If 2.2 and higher AND for an upgrade, use /usr/hdp/<version>/hadoop/{target}.
However, if the upgrade has not yet invoked hdp-select, return the current
version of the component.
:target: the target directory
:force_latest_on_upgrade: if True, then this will return the "current" directory
without the HDP version built into the path, such as /usr/hdp/current/hadoop-client
"""
if not target in HADOOP_DIR_DEFAULTS:
raise Fail("Target {0} not defined".format(target))
hadoop_dir = HADOOP_DIR_DEFAULTS[target]
if Script.is_hdp_stack_greater_or_equal("2.2"):
# home uses a different template
if target == "home":
hadoop_dir = HADOOP_HOME_DIR_TEMPLATE.format(
"current", "hadoop-client")
else:
hadoop_dir = HADOOP_DIR_TEMPLATE.format("current", "hadoop-client",
target)
# if we are not forcing "current" for HDP 2.2, then attempt to determine
# if the exact version needs to be returned in the directory
if not force_latest_on_upgrade:
stack_info = _get_upgrade_stack()
if stack_info is not None:
stack_version = stack_info[1]
# determine if hdp-select has been run and if not, then use the current
# hdp version until this component is upgraded
current_hdp_version = get_role_component_current_hdp_version()
if current_hdp_version is not None and stack_version != current_hdp_version:
stack_version = current_hdp_version
if target == "home":
# home uses a different template
hadoop_dir = HADOOP_HOME_DIR_TEMPLATE.format(
stack_version, "hadoop")
else:
hadoop_dir = HADOOP_DIR_TEMPLATE.format(
stack_version, "hadoop", target)
return hadoop_dir
def get_hdfs_binary(distro_component_name):
"""
Get the hdfs binary to use depending on the stack and version.
:param distro_component_name: e.g., hadoop-hdfs-namenode, hadoop-hdfs-datanode
:return: The hdfs binary to use
"""
import params
hdfs_binary = "hdfs"
if params.stack_name == "HDP":
# This was used in HDP 2.1 and earlier
hdfs_binary = "hdfs"
if Script.is_hdp_stack_greater_or_equal("2.2"):
hdfs_binary = "/usr/hdp/current/{0}/bin/hdfs".format(distro_component_name)
return hdfs_binary
def start(self, env, rolling_restart=False):
import params
env.set_params(params)
self.configure(env) # FOR SECURITY
if params.has_ranger_admin and params.is_supported_yarn_ranger:
setup_ranger_yarn() #Ranger Yarn Plugin related calls
if not Script.is_hdp_stack_greater_or_equal("2.2"):
install_tez_jars()
else:
resource_created = copy_to_hdfs("tez", params.user_group,
params.hdfs_user)
if resource_created:
params.HdfsResource(None, action="execute")
service('resourcemanager', action='start')
def get_hdfs_binary(distro_component_name):
"""
Get the hdfs binary to use depending on the stack and version.
:param distro_component_name: e.g., hadoop-hdfs-namenode, hadoop-hdfs-datanode
:return: The hdfs binary to use
"""
import params
hdfs_binary = "hdfs"
if params.stack_name == "HDP":
# This was used in HDP 2.1 and earlier
hdfs_binary = "hdfs"
if Script.is_hdp_stack_greater_or_equal("2.2"):
hdfs_binary = "/usr/hdp/current/{0}/bin/hdfs".format(
distro_component_name)
return hdfs_binary
def get_hadoop_dir(target, force_latest_on_upgrade=False):
"""
Return the hadoop shared directory in the following override order
1. Use default for 2.1 and lower
2. If 2.2 and higher, use /usr/hdp/current/hadoop-client/{target}
3. If 2.2 and higher AND for an upgrade, use /usr/hdp/<version>/hadoop/{target}.
However, if the upgrade has not yet invoked hdp-select, return the current
version of the component.
:target: the target directory
:force_latest_on_upgrade: if True, then this will return the "current" directory
without the HDP version built into the path, such as /usr/hdp/current/hadoop-client
"""
if not target in HADOOP_DIR_DEFAULTS:
raise Fail("Target {0} not defined".format(target))
hadoop_dir = HADOOP_DIR_DEFAULTS[target]
if Script.is_hdp_stack_greater_or_equal("2.2"):
# home uses a different template
if target == "home":
hadoop_dir = HADOOP_HOME_DIR_TEMPLATE.format("current", "hadoop-client")
else:
hadoop_dir = HADOOP_DIR_TEMPLATE.format("current", "hadoop-client", target)
# if we are not forcing "current" for HDP 2.2, then attempt to determine
# if the exact version needs to be returned in the directory
if not force_latest_on_upgrade:
stack_info = _get_upgrade_stack()
if stack_info is not None:
stack_version = stack_info[1]
# determine if hdp-select has been run and if not, then use the current
# hdp version until this component is upgraded
current_hdp_version = get_role_component_current_hdp_version()
if current_hdp_version is not None and stack_version != current_hdp_version:
stack_version = current_hdp_version
if target == "home":
# home uses a different template
hadoop_dir = HADOOP_HOME_DIR_TEMPLATE.format(stack_version, "hadoop")
else:
hadoop_dir = HADOOP_DIR_TEMPLATE.format(stack_version, "hadoop", target)
return hadoop_dir
dfs_dn_ipc_address = config['configurations']['hdfs-site']['dfs.datanode.ipc.address']
secure_dn_ports_are_in_use = False
# hadoop default parameters
mapreduce_libs_path = "/usr/lib/hadoop-mapreduce/*"
hadoop_libexec_dir = hdp_select.get_hadoop_dir("libexec")
hadoop_bin = hdp_select.get_hadoop_dir("sbin")
hadoop_bin_dir = hdp_select.get_hadoop_dir("bin")
hadoop_home = hdp_select.get_hadoop_dir("home")
hadoop_secure_dn_user = hdfs_user
hadoop_conf_dir = conf_select.get_hadoop_conf_dir()
hadoop_conf_secure_dir = os.path.join(hadoop_conf_dir, "secure")
hadoop_lib_home = hdp_select.get_hadoop_dir("lib")
# hadoop parameters for 2.2+
if Script.is_hdp_stack_greater_or_equal("2.2"):
mapreduce_libs_path = "/usr/hdp/current/hadoop-mapreduce-client/*"
if not security_enabled:
hadoop_secure_dn_user = '******'
else:
dfs_dn_port = utils.get_port(dfs_dn_addr)
dfs_dn_http_port = utils.get_port(dfs_dn_http_addr)
dfs_dn_https_port = utils.get_port(dfs_dn_https_addr)
# We try to avoid inability to start datanode as a plain user due to usage of root-owned ports
if dfs_http_policy == "HTTPS_ONLY":
secure_dn_ports_are_in_use = utils.is_secure_port(dfs_dn_port) or utils.is_secure_port(dfs_dn_https_port)
elif dfs_http_policy == "HTTP_AND_HTTPS":
secure_dn_ports_are_in_use = utils.is_secure_port(dfs_dn_port) or utils.is_secure_port(dfs_dn_http_port) or utils.is_secure_port(dfs_dn_https_port)
else: # params.dfs_http_policy == "HTTP_ONLY" or not defined:
secure_dn_ports_are_in_use = utils.is_secure_port(dfs_dn_port) or utils.is_secure_port(dfs_dn_http_port)
dfs_dn_https_addr = default('/configurations/hdfs-site/dfs.datanode.https.address', None)
dfs_http_policy = default('/configurations/hdfs-site/dfs.http.policy', None)
dfs_dn_ipc_address = config['configurations']['hdfs-site']['dfs.datanode.ipc.address']
secure_dn_ports_are_in_use = False
# hadoop default parameters
mapreduce_libs_path = "/usr/lib/hadoop-mapreduce/*"
hadoop_libexec_dir = "/usr/lib/hadoop/libexec"
hadoop_bin = "/usr/lib/hadoop/sbin"
hadoop_bin_dir = "/usr/bin"
hadoop_home = "/usr/lib/hadoop"
hadoop_secure_dn_user = hdfs_user
hadoop_conf_dir = conf_select.get_hadoop_conf_dir()
# hadoop parameters for 2.2+
if Script.is_hdp_stack_greater_or_equal("2.2"):
mapreduce_libs_path = "/usr/hdp/current/hadoop-mapreduce-client/*"
hadoop_libexec_dir = "/usr/hdp/current/hadoop-client/libexec"
hadoop_bin = "/usr/hdp/current/hadoop-client/sbin"
hadoop_bin_dir = "/usr/hdp/current/hadoop-client/bin"
hadoop_home = "/usr/hdp/current/hadoop-client"
if not security_enabled:
hadoop_secure_dn_user = '******'
else:
dfs_dn_port = utils.get_port(dfs_dn_addr)
dfs_dn_http_port = utils.get_port(dfs_dn_http_addr)
dfs_dn_https_port = utils.get_port(dfs_dn_https_addr)
# We try to avoid inability to start datanode as a plain user due to usage of root-owned ports
if dfs_http_policy == "HTTPS_ONLY":
secure_dn_ports_are_in_use = utils.is_secure_port(dfs_dn_port) or utils.is_secure_port(dfs_dn_https_port)
def get_hadoop_conf_dir(force_latest_on_upgrade=False):
"""
Gets the shared hadoop conf directory using:
1. Start with /etc/hadoop/conf
2. When the stack is greater than HDP-2.2, use /usr/hdp/current/hadoop-client/conf
3. Only when doing a RU and HDP-2.3 or higher, use the value as computed
by conf-select. This is in the form /usr/hdp/VERSION/hadoop/conf to make sure
the configs are written in the correct place. However, if the component itself has
not yet been upgraded, it should use the hadoop configs from the prior version.
This will perform an hdp-select status to determine which version to use.
:param force_latest_on_upgrade: if True, then force the returned path to always
be that of the upgrade target version, even if hdp-select has not been called. This
is primarily used by hooks like before-ANY to ensure that hadoop environment
configurations are written to the correct location since they are written out
before the hdp-select/conf-select would have been called.
"""
hadoop_conf_dir = "/etc/hadoop/conf"
stack_name = None
version = None
allow_setting_conf_select_symlink = False
if not Script.in_stack_upgrade():
# During normal operation, the HDP stack must be 2.3 or higher
if Script.is_hdp_stack_greater_or_equal("2.2"):
hadoop_conf_dir = "/usr/hdp/current/hadoop-client/conf"
if Script.is_hdp_stack_greater_or_equal("2.3"):
hadoop_conf_dir = "/usr/hdp/current/hadoop-client/conf"
stack_name = default("/hostLevelParams/stack_name", None)
version = default("/commandParams/version", None)
if stack_name and version:
version = str(version)
allow_setting_conf_select_symlink = True
else:
# During an upgrade/downgrade, which can be a Rolling or Express Upgrade, need to calculate it based on the version
'''
Whenever upgrading to HDP 2.2, or downgrading back to 2.2, need to use /etc/hadoop/conf
Whenever upgrading to HDP 2.3, or downgrading back to 2.3, need to use a versioned hadoop conf dir
Type__|_Source_|_Target_|_Direction_____________|_Comment_____________________________________________________________
Normal| | 2.2 | | Use /etc/hadoop/conf
Normal| | 2.3 | | Use /etc/hadoop/conf, which should be a symlink to /usr/hdp/current/hadoop-client/conf
EU | 2.1 | 2.3 | Upgrade | Use versioned /usr/hdp/current/hadoop-client/conf
| | | No Downgrade Allowed | Invalid
EU/RU | 2.2 | 2.2.* | Any | Use /usr/hdp/current/hadoop-client/conf
EU/RU | 2.2 | 2.3 | Upgrade | Use /usr/hdp/$version/hadoop/conf, which should be a symlink destination
| | | Downgrade | Use /usr/hdp/current/hadoop-client/conf
EU/RU | 2.3 | 2.3.* | Any | Use /usr/hdp/$version/hadoop/conf, which should be a symlink destination
'''
# The method "is_hdp_stack_greater_or_equal" uses "stack_version" which is the desired stack, e.g., 2.2 or 2.3
# In an RU, it is always the desired stack, and doesn't change even during the Downgrade!
# In an RU Downgrade from HDP 2.3 to 2.2, the first thing we do is
# rm /etc/[component]/conf and then mv /etc/[component]/conf.backup /etc/[component]/conf
if Script.is_hdp_stack_greater_or_equal("2.2"):
hadoop_conf_dir = "/usr/hdp/current/hadoop-client/conf"
# This contains the "version", including the build number, that is actually used during a stack upgrade and
# is the version upgrading/downgrading to.
stack_info = hdp_select._get_upgrade_stack()
if stack_info is not None:
stack_name = stack_info[0]
version = stack_info[1]
else:
raise Fail("Unable to get parameter 'version'")
Logger.info("In the middle of a stack upgrade/downgrade for Stack {0} and destination version {1}, determining which hadoop conf dir to use.".format(stack_name, version))
# This is the version either upgrading or downgrading to.
if compare_versions(format_hdp_stack_version(version), "2.3.0.0") >= 0:
# Determine if hdp-select has been run and if not, then use the current
# hdp version until this component is upgraded.
if not force_latest_on_upgrade:
current_hdp_version = hdp_select.get_role_component_current_hdp_version()
if current_hdp_version is not None and version != current_hdp_version:
version = current_hdp_version
Logger.info("hdp-select has not yet been called to update the symlink for this component, keep using version {0}".format(current_hdp_version))
# Only change the hadoop_conf_dir path, don't conf-select this older version
hadoop_conf_dir = "/usr/hdp/{0}/hadoop/conf".format(version)
Logger.info("Hadoop conf dir: {0}".format(hadoop_conf_dir))
allow_setting_conf_select_symlink = True
if allow_setting_conf_select_symlink:
# If not in the middle of an upgrade and on HDP 2.3 or higher, or if
# upgrading stack to version 2.3.0.0 or higher (which may be upgrade or downgrade), then consider setting the
# symlink for /etc/hadoop/conf.
# If a host does not have any HDFS or YARN components (e.g., only ZK), then it will not contain /etc/hadoop/conf
# Therefore, any calls to conf-select will fail.
# For that reason, if the hadoop conf directory exists, then make sure it is set.
if os.path.exists(hadoop_conf_dir):
Logger.info("The hadoop conf dir {0} exists, will call conf-select on it for version {1}".format(hadoop_conf_dir, version))
select(stack_name, "hadoop", version)
Logger.info("Using hadoop conf dir: {0}".format(hadoop_conf_dir))
return hadoop_conf_dir
'dfs.datanode.ipc.address']
secure_dn_ports_are_in_use = False
# hadoop default parameters
mapreduce_libs_path = "/usr/lib/hadoop-mapreduce/*"
hadoop_libexec_dir = hdp_select.get_hadoop_dir("libexec")
hadoop_bin = hdp_select.get_hadoop_dir("sbin")
hadoop_bin_dir = hdp_select.get_hadoop_dir("bin")
hadoop_home = hdp_select.get_hadoop_dir("home")
hadoop_secure_dn_user = hdfs_user
hadoop_conf_dir = conf_select.get_hadoop_conf_dir()
hadoop_conf_secure_dir = os.path.join(hadoop_conf_dir, "secure")
hadoop_lib_home = hdp_select.get_hadoop_dir("lib")
# hadoop parameters for 2.2+
if Script.is_hdp_stack_greater_or_equal("2.2"):
mapreduce_libs_path = "/usr/hdp/current/hadoop-mapreduce-client/*"
if not security_enabled:
hadoop_secure_dn_user = '******'
else:
dfs_dn_port = utils.get_port(dfs_dn_addr)
dfs_dn_http_port = utils.get_port(dfs_dn_http_addr)
dfs_dn_https_port = utils.get_port(dfs_dn_https_addr)
# We try to avoid inability to start datanode as a plain user due to usage of root-owned ports
if dfs_http_policy == "HTTPS_ONLY":
secure_dn_ports_are_in_use = utils.is_secure_port(
dfs_dn_port) or utils.is_secure_port(dfs_dn_https_port)
elif dfs_http_policy == "HTTP_AND_HTTPS":
secure_dn_ports_are_in_use = utils.is_secure_port(
dfs_dn_port) or utils.is_secure_port(
dfs_dn_ipc_address = config['configurations']['hdfs-site']['dfs.datanode.ipc.address']
secure_dn_ports_are_in_use = False
# hadoop default parameters
mapreduce_libs_path = "/usr/lib/hadoop-mapreduce/*"
hadoop_libexec_dir = hdp_select.get_hadoop_dir("libexec")
hadoop_bin = hdp_select.get_hadoop_dir("sbin")
hadoop_bin_dir = hdp_select.get_hadoop_dir("bin")
hadoop_home = hdp_select.get_hadoop_dir("home")
hadoop_secure_dn_user = hdfs_user
hadoop_conf_dir = conf_select.get_hadoop_conf_dir()
hadoop_conf_secure_dir = os.path.join(hadoop_conf_dir, "secure")
hadoop_lib_home = hdp_select.get_hadoop_dir("lib")
# hadoop parameters for 2.2+
if Script.is_hdp_stack_greater_or_equal("2.2"):
mapreduce_libs_path = "/usr/hdp/current/hadoop-mapreduce-client/*"
if not security_enabled:
hadoop_secure_dn_user = '******'
else:
dfs_dn_port = utils.get_port(dfs_dn_addr)
dfs_dn_http_port = utils.get_port(dfs_dn_http_addr)
dfs_dn_https_port = utils.get_port(dfs_dn_https_addr)
# We try to avoid inability to start datanode as a plain user due to usage of root-owned ports
if dfs_http_policy == "HTTPS_ONLY":
secure_dn_ports_are_in_use = utils.is_secure_port(dfs_dn_port) or utils.is_secure_port(dfs_dn_https_port)
elif dfs_http_policy == "HTTP_AND_HTTPS":
secure_dn_ports_are_in_use = utils.is_secure_port(dfs_dn_port) or utils.is_secure_port(dfs_dn_http_port) or utils.is_secure_port(dfs_dn_https_port)
else: # params.dfs_http_policy == "HTTP_ONLY" or not defined:
secure_dn_ports_are_in_use = utils.is_secure_port(dfs_dn_port) or utils.is_secure_port(dfs_dn_http_port)
dfs_dn_ipc_address = config['configurations']['hdfs-site'][
'dfs.datanode.ipc.address']
secure_dn_ports_are_in_use = False
# hadoop default parameters
mapreduce_libs_path = "/usr/lib/hadoop-mapreduce/*"
hadoop_libexec_dir = conf_select.get_hadoop_dir("libexec")
hadoop_bin = conf_select.get_hadoop_dir("sbin")
hadoop_bin_dir = conf_select.get_hadoop_dir("bin")
hadoop_home = "/usr/lib/hadoop"
hadoop_secure_dn_user = hdfs_user
hadoop_conf_dir = conf_select.get_hadoop_conf_dir()
hadoop_lib_home = conf_select.get_hadoop_dir("lib")
# hadoop parameters for 2.2+
if Script.is_hdp_stack_greater_or_equal("2.2"):
mapreduce_libs_path = "/usr/hdp/current/hadoop-mapreduce-client/*"
hadoop_home = "/usr/hdp/current/hadoop-client"
if not security_enabled:
hadoop_secure_dn_user = '******'
else:
dfs_dn_port = utils.get_port(dfs_dn_addr)
dfs_dn_http_port = utils.get_port(dfs_dn_http_addr)
dfs_dn_https_port = utils.get_port(dfs_dn_https_addr)
# We try to avoid inability to start datanode as a plain user due to usage of root-owned ports
if dfs_http_policy == "HTTPS_ONLY":
secure_dn_ports_are_in_use = utils.is_secure_port(
dfs_dn_port) or utils.is_secure_port(dfs_dn_https_port)
elif dfs_http_policy == "HTTP_AND_HTTPS":
secure_dn_ports_are_in_use = utils.is_secure_port(
stack_name = default("/hostLevelParams/stack_name", None)
current_version = default("/hostLevelParams/current_version", None)
component_directory = status_params.component_directory
# New Cluster Stack Version that is defined during the RESTART of a Rolling Upgrade
version = default("/commandParams/version", None)
# default parameters
zk_home = "/usr/hdp/2.3.2.0-2950/zookeeper"
zk_bin = "/usr/hdp/2.3.2.0-2950/zookeeper/bin"
zk_cli_shell = "/usr/hdp/2.3.2.0-2950/zookeeper/bin/zkCli.sh"
config_dir = "/etc/zookeeper/conf"
# hadoop parameters for 2.2+
if Script.is_hdp_stack_greater_or_equal("2.2"):
zk_home = format("/usr/hdp/current/{component_directory}")
zk_bin = format("/usr/hdp/current/{component_directory}/bin")
zk_cli_shell = format("/usr/hdp/current/{component_directory}/bin/zkCli.sh")
config_dir = status_params.config_dir
zk_user = config['configurations']['zookeeper-env']['zk_user']
hostname = config['hostname']
user_group = config['configurations']['cluster-env']['user_group']
zk_env_sh_template = config['configurations']['zookeeper-env']['content']
zk_log_dir = config['configurations']['zookeeper-env']['zk_log_dir']
zk_data_dir = config['configurations']['zoo.cfg']['dataDir']
zk_pid_dir = status_params.zk_pid_dir
zk_pid_file = status_params.zk_pid_file
def get_hadoop_conf_dir(force_latest_on_upgrade=False):
"""
Gets the shared hadoop conf directory using:
1. Start with /etc/hadoop/conf
2. When the stack is greater than HDP-2.2, use /usr/hdp/current/hadoop-client/conf
3. Only when doing a RU and HDP-2.3 or higher, use the value as computed
by conf-select. This is in the form /usr/hdp/VERSION/hadoop/conf to make sure
the configs are written in the correct place. However, if the component itself has
not yet been upgraded, it should use the hadoop configs from the prior version.
This will perform an hdp-select status to determine which version to use.
:param force_latest_on_upgrade: if True, then force the returned path to always
be that of the upgrade target version, even if hdp-select has not been called. This
is primarily used by hooks like before-ANY to ensure that hadoop environment
configurations are written to the correct location since they are written out
before the hdp-select/conf-select would have been called.
"""
hadoop_conf_dir = "/etc/hadoop/conf"
stack_name = None
version = None
allow_setting_conf_select_symlink = False
if not Script.in_stack_upgrade():
# During normal operation, the HDP stack must be 2.3 or higher
if Script.is_hdp_stack_greater_or_equal("2.2"):
hadoop_conf_dir = "/usr/hdp/current/hadoop-client/conf"
if Script.is_hdp_stack_greater_or_equal("2.3"):
hadoop_conf_dir = "/usr/hdp/current/hadoop-client/conf"
stack_name = default("/hostLevelParams/stack_name", None)
version = default("/commandParams/version", None)
if stack_name and version:
version = str(version)
allow_setting_conf_select_symlink = True
else:
# During an upgrade/downgrade, which can be a Rolling or Express Upgrade, need to calculate it based on the version
'''
Whenever upgrading to HDP 2.2, or downgrading back to 2.2, need to use /etc/hadoop/conf
Whenever upgrading to HDP 2.3, or downgrading back to 2.3, need to use a versioned hadoop conf dir
Type__|_Source_|_Target_|_Direction_____________|_Comment_____________________________________________________________
Normal| | 2.2 | | Use /etc/hadoop/conf
Normal| | 2.3 | | Use /etc/hadoop/conf, which should be a symlink to /usr/hdp/current/hadoop-client/conf
EU | 2.1 | 2.3 | Upgrade | Use versioned /usr/hdp/current/hadoop-client/conf
| | | No Downgrade Allowed | Invalid
EU/RU | 2.2 | 2.2.* | Any | Use /usr/hdp/current/hadoop-client/conf
EU/RU | 2.2 | 2.3 | Upgrade | Use /usr/hdp/$version/hadoop/conf, which should be a symlink destination
| | | Downgrade | Use /usr/hdp/current/hadoop-client/conf
EU/RU | 2.3 | 2.3.* | Any | Use /usr/hdp/$version/hadoop/conf, which should be a symlink destination
'''
# The method "is_hdp_stack_greater_or_equal" uses "stack_version" which is the desired stack, e.g., 2.2 or 2.3
# In an RU, it is always the desired stack, and doesn't change even during the Downgrade!
# In an RU Downgrade from HDP 2.3 to 2.2, the first thing we do is
# rm /etc/[component]/conf and then mv /etc/[component]/conf.backup /etc/[component]/conf
if Script.is_hdp_stack_greater_or_equal("2.2"):
hadoop_conf_dir = "/usr/hdp/current/hadoop-client/conf"
# This contains the "version", including the build number, that is actually used during a stack upgrade and
# is the version upgrading/downgrading to.
stack_info = hdp_select._get_upgrade_stack()
if stack_info is not None:
stack_name = stack_info[0]
version = stack_info[1]
else:
raise Fail("Unable to get parameter 'version'")
Logger.info(
"In the middle of a stack upgrade/downgrade for Stack {0} and destination version {1}, determining which hadoop conf dir to use."
.format(stack_name, version))
# This is the version either upgrading or downgrading to.
if compare_versions(format_hdp_stack_version(version),
"2.3.0.0") >= 0:
# Determine if hdp-select has been run and if not, then use the current
# hdp version until this component is upgraded.
if not force_latest_on_upgrade:
current_hdp_version = hdp_select.get_role_component_current_hdp_version(
)
if current_hdp_version is not None and version != current_hdp_version:
version = current_hdp_version
Logger.info(
"hdp-select has not yet been called to update the symlink for this component, keep using version {0}"
.format(current_hdp_version))
# Only change the hadoop_conf_dir path, don't conf-select this older version
hadoop_conf_dir = "/usr/hdp/{0}/hadoop/conf".format(version)
Logger.info("Hadoop conf dir: {0}".format(hadoop_conf_dir))
allow_setting_conf_select_symlink = True
if allow_setting_conf_select_symlink:
# If not in the middle of an upgrade and on HDP 2.3 or higher, or if
# upgrading stack to version 2.3.0.0 or higher (which may be upgrade or downgrade), then consider setting the
# symlink for /etc/hadoop/conf.
# If a host does not have any HDFS or YARN components (e.g., only ZK), then it will not contain /etc/hadoop/conf
# Therefore, any calls to conf-select will fail.
# For that reason, if the hadoop conf directory exists, then make sure it is set.
if os.path.exists(hadoop_conf_dir):
Logger.info(
"The hadoop conf dir {0} exists, will call conf-select on it for version {1}"
.format(hadoop_conf_dir, version))
select(stack_name, "hadoop", version)
Logger.info("Using hadoop conf dir: {0}".format(hadoop_conf_dir))
return hadoop_conf_dir
# zeppelin-env.sh
zeppelin_env_content = config["configurations"]["zeppelin-env"]["content"]
# detect HS2 details and java home
master_configs = config["clusterHostInfo"]
hive_server_host = str(master_configs["hive_server_host"][0])
hive_metastore_host = str(master_configs["hive_metastore_host"][0])
hive_metastore_port = str(get_port_from_url(config["configurations"]["hive-site"]["hive.metastore.uris"]))
java64_home = config["hostLevelParams"]["java_home"]
# e.g. 2.3
stack_version_unformatted = str(config["hostLevelParams"]["stack_version"])
# e.g. 2.3.0.0
hdp_stack_version = format_hdp_stack_version(stack_version_unformatted)
if hasattr(Script, "is_hdp_stack_greater_or_equal") and Script.is_hdp_stack_greater_or_equal("2.3"):
mvn_spark_tag = "spark-1.3"
else:
mvn_spark_tag = "spark-1.2"
# e.g. 2.3.0.0-2130
full_version = default("/commandParams/version", None)
hdp_version = full_version
if hasattr(functions, "get_hdp_version"):
spark_version = functions.get_hdp_version("spark-client")
dfs_http_policy = default('/configurations/hdfs-site/dfs.http.policy', None)
dfs_dn_ipc_address = config['configurations']['hdfs-site'][
'dfs.datanode.ipc.address']
secure_dn_ports_are_in_use = False
# hadoop default parameters
mapreduce_libs_path = "/usr/lib/hadoop-mapreduce/*"
hadoop_libexec_dir = "/usr/lib/hadoop/libexec"
hadoop_bin = "/usr/lib/hadoop/sbin"
hadoop_bin_dir = "/usr/bin"
hadoop_home = "/usr/lib/hadoop"
hadoop_secure_dn_user = hdfs_user
hadoop_conf_dir = conf_select.get_hadoop_conf_dir()
# hadoop parameters for 2.2+
if Script.is_hdp_stack_greater_or_equal("2.2"):
mapreduce_libs_path = "/usr/hdp/current/hadoop-mapreduce-client/*"
hadoop_libexec_dir = "/usr/hdp/current/hadoop-client/libexec"
hadoop_bin = "/usr/hdp/current/hadoop-client/sbin"
hadoop_bin_dir = "/usr/hdp/current/hadoop-client/bin"
hadoop_home = "/usr/hdp/current/hadoop-client"
if not security_enabled:
hadoop_secure_dn_user = '******'
else:
dfs_dn_port = utils.get_port(dfs_dn_addr)
dfs_dn_http_port = utils.get_port(dfs_dn_http_addr)
dfs_dn_https_port = utils.get_port(dfs_dn_https_addr)
# We try to avoid inability to start datanode as a plain user due to usage of root-owned ports
if dfs_http_policy == "HTTPS_ONLY":
secure_dn_ports_are_in_use = utils.is_secure_port(
from resource_management.libraries import functions
# server configurations
config = Script.get_config()
tmp_dir = Script.get_tmp_dir()
sudo = AMBARI_SUDO_BINARY
stack_name = default("/hostLevelParams/stack_name", None)
# node hostname
hostname = config["hostname"]
# This is expected to be of the form #.#.#.#
stack_version_unformatted = str(config["hostLevelParams"]["stack_version"])
hdp_stack_version_major = format_hdp_stack_version(stack_version_unformatted)
stack_is_hdp21 = Script.is_hdp_stack_greater_or_equal("2.0") and Script.is_hdp_stack_less_than("2.2")
# this is not available on INSTALL action because hdp-select is not available
hdp_stack_version = functions.get_hdp_version("hive-server2")
# New Cluster Stack Version that is defined during the RESTART of a Rolling Upgrade.
# It cannot be used during the initial Cluser Install because the version is not yet known.
version = default("/commandParams/version", None)
# current host stack version
current_version = default("/hostLevelParams/current_version", None)
# When downgrading the 'version' and 'current_version' are both pointing to the downgrade-target version
# downgrade_from_version provides the source-version the downgrade is happening from
downgrade_from_version = default("/commandParams/downgrade_from_version", None)
knox_master_secret_path = format('/var/lib/knox/data/security/master')
knox_cert_store_path = format('/var/lib/knox/data/security/keystores/gateway.jks')
knox_user = default("/configurations/knox-env/knox_user", "knox")
# server configurations
knox_data_dir = '/var/lib/knox/data'
knox_logs_dir = '/var/log/knox'
# default parameters
knox_bin = '/usr/lib/knox/bin/gateway.sh'
knox_conf_dir = '/etc/knox/conf'
ldap_bin = '/usr/lib/knox/bin/ldap.sh'
knox_client_bin = '/usr/lib/knox/bin/knoxcli.sh'
# HDP 2.2+ parameters
if Script.is_hdp_stack_greater_or_equal("2.2"):
knox_bin = '/usr/hdp/current/knox-server/bin/gateway.sh'
knox_conf_dir = '/usr/hdp/current/knox-server/conf'
ldap_bin = '/usr/hdp/current/knox-server/bin/ldap.sh'
knox_client_bin = '/usr/hdp/current/knox-server/bin/knoxcli.sh'
knox_master_secret_path = '/usr/hdp/current/knox-server/data/security/master'
knox_cert_store_path = '/usr/hdp/current/knox-server/data/security/keystores/gateway.jks'
knox_data_dir = '/usr/hdp/current/knox-server/data/'
knox_group = default("/configurations/knox-env/knox_group", "knox")
mode = 0644
stack_version_unformatted = str(config['hostLevelParams']['stack_version'])
hdp_stack_version = format_hdp_stack_version(stack_version_unformatted)
zeppelin_env_content = config['configurations']['zeppelin-env']['content']
#detect HS2 details and java home
master_configs = config['clusterHostInfo']
hive_server_host = str(master_configs['hive_server_host'][0])
hive_metastore_host = str(master_configs['hive_metastore_host'][0])
hive_metastore_port = str(
get_port_from_url(
config['configurations']['hive-site']['hive.metastore.uris']))
java64_home = config['hostLevelParams']['java_home']
#e.g. 2.3
stack_version_unformatted = str(config['hostLevelParams']['stack_version'])
#e.g. 2.3.0.0
hdp_stack_version = format_hdp_stack_version(stack_version_unformatted)
if hasattr(Script, 'is_hdp_stack_greater_or_equal'
) and Script.is_hdp_stack_greater_or_equal("2.3"):
mvn_spark_tag = 'spark-1.3'
else:
mvn_spark_tag = 'spark-1.2'
#e.g. 2.3.0.0-2130
full_version = default("/commandParams/version", None)
hdp_version = full_version
if hasattr(functions, 'get_hdp_version'):
spark_version = functions.get_hdp_version('spark-client')
