def ru_downgrade_state(cls):
'''
Downgrades Namenode
A downgrade is done - may need to convert state to previous version or state is compatible - again upgrade is being abandoned
NOTE: this command will not return until namenode shuts down
'''
command = "sudo su - -c 'hadoop namenode -rollingUpgrade downgrade' hdfs"
if HDFS.isHAEnabled():
nodes = []
nodes.append(HDFS.getNamenodeByState('standby'))
nodes.append(HDFS.getNamenodeByState('active'))
for node in nodes:
HDFS.resetNamenode('stop', host=node)
(exitcode, stdout) = Machine.runas(Machine.getAdminUser(),
command, node, None, None,
"True",
Machine.getAdminPasswd())
ruAssert(
"HDFS", exitcode == 0,
"[NNDowngrade] hadoop namenode -rollingUpgrade downgrade command failed"
)
return
HDFS.stopNamenode()
node = HDFS.getNamenode()
(exitcode, stdout) = Machine.runas(Machine.getAdminUser(), command,
node, None, None, "True",
Machine.getAdminPasswd())
ruAssert(
"HDFS", exitcode == 0,
"[NNDowngrade] hadoop namenode -rollingUpgrade downgrade command failed"
)
def ru_rollback_state(cls):
'''
Saved state is rolled back - upgrade is abandonded
NOTE: this command will not return until namenode shuts down
:return:
'''
logger.info("[INFO][HDFS][Upgrade] HA Journal Node Upgrade Started ")
hdfs_user = Config.get('hadoop', 'HDFS_USER')
nodes = []
nodes.append(HDFS.getNamenodeByState('standby'))
nodes.append(HDFS.getNamenodeByState('active'))
logger.info("[INFO][HDFS][Upgrade] HA Namenode Upgrade Started")
for node in nodes:
HDFS.resetZkfc('stop', hosts=node.split())
HDFS.resetNamenode('stop', host=node)
HDFS.resetNamenode('start',
config=None,
host=nodes[0],
option=" -rollingUpgrade rollback")
HDFS.resetZkfc('start', hosts=nodes[0].split())
# lets make sure the NN is out of safemode before we proceed to the next namenode
HDFS.waitForNNOutOfSafemode(options='-fs hdfs://%s:8020' % nodes[0])
command = "sudo su - -c 'hadoop namenode -bootstrapStandby -force' hdfs"
(exitcode, stdout) = Machine.runas(Machine.getAdminUser(), command,
nodes[1], None, None, "True",
Machine.getAdminPasswd())
ruAssert("HDFS", exitcode == 0,
"hadoop namenode -bootstrapStandby -force")
HDFS.resetNamenode('start', config=None, host=nodes[1], option="")
HDFS.resetZkfc('start', hosts=nodes[1].split())
# lets make sure the NN is out of safemode before we proceed to the next namenode
HDFS.waitForNNOutOfSafemode(options='-fs hdfs://%s:8020' % nodes[1])
def updateJobProperties(cls,
propFile,
properties=None,
haEnabled=False,
debug=False):
fileSystemName = Hadoop.getFSDefaultValue()
jobTrackerIP = MAPRED.getJobtrackerAddress()
jobTracker = jobTrackerIP[0] + ":" + jobTrackerIP[1]
if not properties:
properties = {}
if not properties.has_key('nameNode'):
properties['nameNode'] = fileSystemName
if not properties.has_key('jobTracker'):
properties['jobTracker'] = jobTracker
if "hcatalog" in propFile:
if Hadoop.isSecure():
kerberosPrincipal = Hive.getConfigValue(
"hive.metastore.kerberos.principal")
properties[
'hive.metastore.kerberos.principal'] = kerberosPrincipal
logger.info("Updating for hcatalog workflow")
hcatNode = Hive.getConfigValue("hive.metastore.uris").replace(
'thrift', 'hcat')
logger.info("Hcat node is " + hcatNode)
properties['hcatNode'] = hcatNode
if Hadoop.isSecure():
# determine the namenode and the jobtracker principal
nnPrincipal = None
if haEnabled:
nnPrincipal = HDFS.getNameNodePrincipal().replace(
'_HOST', HDFS.getNamenodeByState('active'))
else:
nnPrincipal = HDFS.getNameNodePrincipal().replace(
'_HOST',
HDFS.getNamenodeHttpAddress()[0])
jtPrincipal = MAPRED.getMasterPrincipal().replace(
'_HOST', jobTrackerIP[0])
properties['dfs.namenode.kerberos.principal'] = nnPrincipal
properties['mapreduce.jobtracker.kerberos.principal'] = jtPrincipal
wfPath = util.getPropertyValueFromFile(propFile,
"oozie.wf.application.path")
if wfPath != None and wfPath.find("hdfs://localhost:9000") != -1:
wfPath = wfPath.replace("hdfs://localhost:9000", fileSystemName)
logger.info("Value of replaced oozie.wf.application.path is " +
wfPath)
properties['oozie.wf.application.path'] = wfPath
util.writePropertiesToFile(propFile, propFile, properties)
if debug:
logger.info('Content of properties file %s' % propFile)
f = open(propFile, 'r')
# print the file to the console
logger.info(f.read())
f.close()
def ensure_nn_is_active(cls, timeout=11 * 60):
from beaver.component.rollingupgrade.ruUpgrade import UpgradePerNode
active_nn = None
curr_time = int(time.time())
end_time = int(time.time()) + timeout
while not active_nn and curr_time <= end_time:
active_nn = HDFS.getNamenodeByState('active')
# we we found active NN return it
if active_nn:
UpgradePerNode.reportProgress(
"[INFO][HDFS] Active Namenode is %s" % active_nn)
return
# wait for 30s
time.sleep(30)
curr_time = int(time.time())
if not active_nn:
UpgradePerNode.reportProgress("[FAILED][HDFS] No Active Namenode")
assert active_nn
return
def downgrade_master(cls, version, config=None):
'''
Downgrade HDFS Master services
:param version: Version to be downgraded to
:param config: Configuration location
'''
from beaver.component.rollingupgrade.ruCommon import hdpSelect
from beaver.component.rollingupgrade.ruUpgrade import UpgradePerNode
if HDFS.isHAEnabled():
UpgradePerNode.reportProgress(
"[INFO][HDFS][Downgrade] HA Namenode Downgrade Started ")
nodes = []
nodes.append(HDFS.getNamenodeByState('standby'))
nodes.append(HDFS.getNamenodeByState('active'))
for node in nodes:
HDFS.resetNamenode('stop', host=node)
HDFS.resetZkfc('stop', hosts=node.split())
# BUG-25534: There is no package for zkfc. So just updating the NN is enough.
hdpSelect.changeVersion("hadoop-hdfs-namenode", version, node)
HDFS.resetNamenode('start', config=config, host=node)
HDFS.resetZkfc('start', hosts=node.split())
# lets make sure the NN is out of safemode before we proceed to the next namenode
HDFS.waitForNNOutOfSafemode(options='-fs hdfs://%s:8020' %
node)
UpgradePerNode.reportProgress(
"[INFO][HDFS][Downgrade] HA Namenode Downgrade Finished ")
jn_nodes = HDFS.getJournalNodes()
if len(jn_nodes) < 3:
UpgradePerNode.reportProgress(
"[INFO][HDFS][Downgrade] Less than three Journal Nodes. Not going to do Downgrade "
)
return
#Loop through all the JNs and stop flip start one at a time
hdfs_user = Config.get('hadoop', 'HDFS_USER')
UpgradePerNode.reportProgress(
"[INFO][HDFS][Downgrade] HA Journal Node Upgrade Started")
cls.ensure_all_jns_are_up(jn_nodes)
#Loop through all the JNs and stop flip start one at a time
for node in jn_nodes:
Hadoop.resetService(hdfs_user,
node,
"journalnode",
'stop',
binFolder="sbin")
hdpSelect.changeVersion("hadoop-hdfs-journalnode", version,
node)
Hadoop.resetService(hdfs_user,
node,
"journalnode",
'start',
binFolder="sbin")
time.sleep(5)
cls.ensure_all_jns_are_up(jn_nodes)
cls.ensure_nn_is_active()
UpgradePerNode.reportProgress(
"[INFO][HDFS][Downgrade] HA Journal Node Downgrade Finished ")
else:
## TODO add code to upgrade SNN
UpgradePerNode.reportProgress(
"[INFO][HDFS][Downgrade] Namenode Downgrade Started ")
node = HDFS.getNamenode()
HDFS.stopNamenode()
hdpSelect.changeVersion("hadoop-hdfs-namenode", version, node)
HDFS.startNamenode(config=config)
UpgradePerNode.reportProgress(
"[INFO][HDFS][Downgrade] Namenode Downgrade Finished ")
# upgrade SNN
UpgradePerNode.reportProgress(
"[INFO][HDFS][Downgrade]Secondary Namenode Downgrade Started ")
node = HDFS.getSecondaryNamenode()
HDFS.stopSecondaryNamenode()
hdpSelect.changeVersion("hadoop-hdfs-secondarynamenode", version,
node)
HDFS.startSecondaryNamenode(config=config)
UpgradePerNode.reportProgress(
"[INFO][HDFS][Downgrade] Secondary Namenode Downgrade Finished"
)
