def attach(self, cc, out):
out.info("Started installing %s in %s (%s) ---------------------------------------" % (self.role, self.hostname, self.ip_address))
# Regenerate host files of all the managed nodes (before acquiring the node lock)
if not cc.options.dns:
try:
self.config_description = 'regenerating /etc/hosts files of the cluster'
bigdata.acquire('cluster')
cc = bigdata.create_config_context(cc.options)
self.regenerate_etc_hosts_files(cc, out)
finally:
bigdata.release('cluster')
# Connect by SSH
bigdata.acquire("node-%s-files" % self.hostname)
ssh = SSHConnection(self.hostname, out)
try:
out.info("Connecting to %s" % self.hostname)
ssh.connect()
# Check operating system type and version and remote hostname
self.check_operating_system_version(cc, out, ssh)
self.check_remote_hostname(cc, out, ssh)
# Set state
self.put_config_state("attaching")
# Regenerate host files of all the managed nodes
try:
bigdata.acquire('cluster')
self.regenerate_etc_hosts_files(cc, out)
finally:
bigdata.release('cluster')
# Make template params
template_params = make_template_params(cc)
# Check if there is an installation in the node already
self.check_possible_big_data_installation(cc, out, ssh)
# List of rpm lists for uninstallation
rpms_list = RpmList(self)
rpms_list.push([])
# Install RPMs
if cc.options.rpms:
self.config_description = 'installing rpms'
## Running yum update
#out.info("Running YUM update") # TODO: this shouldn't be done in the final version
#ssh.execute("yum clean all ; yum --assumeyes --quiet update")
# Install open java
out.info("Installing OpenJDK")
ssh.install(['java-1.6.0-openjdk'])
if self.role == "hmaster":
out.info("Installing HMaster RPM files")
if self.type == 'hbase':
rpms = [
'zookeeper',
'hadoop-0.20-mapreduce-jobtracker',
'hadoop-hdfs-secondarynamenode',
'hadoop-hdfs-namenode',
'hbase',
'hbase-master',
'hadoop-hdfs',
]
else:
rpms = [
'hadoop-0.20-mapreduce-jobtracker',
'hadoop-hdfs-secondarynamenode',
'hadoop-hdfs-namenode',
'hadoop-hdfs',
]
ssh.install(rpms)
rpms_list.push(rpms)
elif self.role == "zookeeper":
# Install RPM packages
out.info("Installing ZooKeeper RPM files")
rpms = [
'zookeeper',
'zookeeper-server',
]
ssh.install(rpms)
rpms_list.push(rpms)
elif self.role == "slave":
# Install RPM packages
if self.type == 'hbase':
rpms = [
'zookeeper',
'hadoop-hdfs',
'hadoop-hdfs-datanode',
'hadoop-0.20-mapreduce-tasktracker',
'hbase',
'hbase-regionserver',
]
else:
rpms = [
'hadoop-hdfs',
'hadoop-hdfs-datanode',
'hadoop-0.20-mapreduce-tasktracker',
]
ssh.install(rpms)
rpms_list.push(rpms)
elif self.role == "hive":
# Install RPM packages
rpms = [
'hive',
'hive-metastore',
'hive-server2',
'MariaDB-server',
'MariaDB-client',
'mysql-connector-java',
]
if self.type == 'hbase':
rpms += ['hive-hbase']
if cc.pig_support:
rpms += ['pig']
rpms += ['pig-udf-datafu']
ssh.install(rpms)
rpms_list.push(rpms)
else:
raise "Unknown role: %s" % (self.role)
# FIXME: earlier these all refered to 'hbase' user, but changed to the correct ones. Does this work?
hbase_homedir = ssh.get_homedir_of('hbase')
hdfs_homedir = ssh.get_homedir_of('hdfs')
zookeeper_homedir = ssh.get_homedir_of('zookeeper')
hive_homedir = ssh.get_homedir_of('hive')
# Ensure that update-alternatives configuration is correct
if self.role in ['hmaster', 'slave']:
out.info("Configure update alternatives")
ssh.execute("mkdir -p /etc/hadoop/conf.cluster")
ssh.execute("cp -pfR /etc/hadoop/conf.empty/* /etc/hadoop/conf.cluster/")
ssh.execute("update-alternatives --install /etc/hadoop/conf hadoop-conf /etc/hadoop/conf.cluster 20")
ssh.execute("update-alternatives --set hadoop-conf /etc/hadoop/conf.cluster")
# Wait until expected number of dependencies are ready
try:
self.config_description = 'waiting for other nodes'
bigdata.release('node-%s' % self.hostname)
if self.role == 'hmaster':
if len(cc.zookeepers) > 0:
self.config_description = 'waiting for zookeepers'
while count_config_states(cc.zookeepers, 'attached') < 3:
sleep(2.0)
cc = self.recreate_config_context(cc.options)
self.config_description = 'waiting for initial slaves'
while count_config_states(cc.slaves, 'attached') < 3:
sleep(2.0)
cc = self.recreate_config_context(cc.options)
elif self.role == 'zookeeper':
pass
elif self.role == 'slave':
if len(cc.zookeepers) > 0:
self.config_description = 'waiting for zookeepers'
while count_config_states(cc.zookeepers, 'attached') < 3:
sleep(2.0)
cc = self.recreate_config_context(cc.options)
elif self.role == 'hive':
# Wait for ZooKeepers
if len(cc.zookeepers) > 0:
self.config_description = 'waiting for zookeepers'
while count_config_states(cc.zookeepers, 'attached') < 3:
sleep(2.0)
cc = self.recreate_config_context(cc.options)
# Wait for HMaster
self.config_description = 'waiting for hmaster'
while count_config_states(cc.hmasters, 'attached') < 1:
sleep(2.0)
cc = self.recreate_config_context(cc.options)
else:
raise MgmtException("Uknown role %s" % self.role)
finally:
bigdata.acquire('node-%s' % self.hostname)
self.config_description = ''
# Re-read nodes and generate template parameters
try:
bigdata.acquire('cluster')
cc = bigdata.create_config_context(cc.options)
template_params = make_template_params(cc)
finally:
bigdata.release('cluster')
# Populate templates and send them to the remote server
self.config_description = 'populating templates'
out.info("Populating and applying configuration templates")
for content, remote_filename, mode in populate_templates('hbase', self.role, template_params):
ssh.send_file_to(content, remote_filename, mode=mode)
# Repopulate and start the first slaves
if self.role == 'hmaster':
self.config_description = "updating and restarting slaves"
out.info("Updating and restarting first 3 slaves")
for snode in cc.slaves[0:3]:
self.config_description = "updating and restarting slaves: %s" % snode.hostname
rssh = SSHConnection(snode.hostname, out)
try:
bigdata.acquire('node-%s' % snode.hostname)
rssh.connect()
# Repopulate and send the templates
for content, remote_filename, mode in populate_templates('hbase', 'slave', template_params):
rssh.send_file_to(content, remote_filename, mode=mode)
# (Re)start the services
rssh.execute("service hadoop-hdfs-datanode restart");
rssh.execute("service hadoop-0.20-mapreduce-tasktracker restart");
if self.type == 'hbase':
rssh.execute("service hbase-regionserver restart");
finally:
bigdata.release('node-%s' % snode.hostname)
rssh.disconnect()
# Run post install script
self.config_description = 'executing post install script'
out.info("Executing post-install script")
ssh.execute("cd /tmp && ./post-install.sh", raise_on_non_zero=True,
raise_on_keywords=["java.net.ConnectException"])
# Process startups
if self.role == "hmaster":
# Copy SSH keys of hdfs and hbase to the bigdata storage
self.config_description = 'sending ssh keys'
if self.type == 'hbase':
out.info("Save public SSH keys of hbase and hdfs")
self.hdfs_public_ssh_key = ssh.receive_file_from("%s/.ssh/id_rsa.pub" % hdfs_homedir)
self.hbase_public_ssh_key = ssh.receive_file_from("%s/.ssh/id_rsa.pub" % hbase_homedir)
else:
out.info("Save public SSH keys of hdfs")
self.hdfs_public_ssh_key = ssh.receive_file_from("%s/.ssh/id_rsa.pub" % hdfs_homedir)
# Start the services
self.config_description = 'starting services'
if self.type == 'hbase':
out.info("Starting services in HMaster")
ssh.execute("service hbase-master stop", raise_on_non_zero=False)
ssh.execute("/etc/init.d/hadoop-0.20-mapreduce-jobtracker stop", raise_on_non_zero=False)
ssh.execute("service hadoop-hdfs-secondarynamenode stop", raise_on_non_zero=False)
ssh.execute("service hadoop-hdfs-namenode stop", raise_on_non_zero=False)
ssh.execute("service hadoop-hdfs-namenode start")
ssh.execute("service hadoop-hdfs-secondarynamenode start")
ssh.execute("/etc/init.d/hadoop-0.20-mapreduce-jobtracker restart")
ssh.execute("service hbase-master start")
else:
out.info("Starting services in Hadoop Master")
ssh.execute("/etc/init.d/hadoop-0.20-mapreduce-jobtracker stop", raise_on_non_zero=False)
ssh.execute("service hadoop-hdfs-secondarynamenode stop", raise_on_non_zero=False)
ssh.execute("service hadoop-hdfs-namenode stop", raise_on_non_zero=False)
ssh.execute("service hadoop-hdfs-namenode start")
ssh.execute("service hadoop-hdfs-secondarynamenode start")
ssh.execute("/etc/init.d/hadoop-0.20-mapreduce-jobtracker restart")
elif self.role == "zookeeper":
# Initialize the service
self.config_description = 'starting services'
out.info("Initiating ZooKeeper")
ssh.execute("service zookeeper-server init");
# update zookeeper id before start
ssh.execute("echo %s > /var/lib/zookeeper/myid" % (cc.zookeepers.index(self)));
# Start the service
out.info("Starting services in ZooKeeper")
ssh.execute("service zookeeper-server restart");
elif self.role == "slave":
# Copy SSH public keys
if self.type == 'hbase':
out.info("Copying the public SSH keys of hbase and hdfs from master to the node")
ssh.send_file_to(self.hbase_public_ssh_key, "/tmp/id_rsa.pub")
ssh.execute("cat /tmp/id_rsa.pub >> %s/.ssh/authorized_keys && rm /tmp/id_rsa.pub" % hbase_homedir)
ssh.execute("mkdir %s/.ssh && chmod 0700 %s/.ssh" % (hdfs_homedir, hdfs_homedir), raise_on_non_zero=False)
ssh.send_file_to(self.hdfs_public_ssh_key, "/tmp/id_rsa.pub")
ssh.execute("cat /tmp/id_rsa.pub >> %s/.ssh/authorized_keys && rm /tmp/id_rsa.pub" % hdfs_homedir)
else:
out.info("Copying the public SSH keys of hdfs from master to the node")
ssh.execute("mkdir %s/.ssh && chmod 0700 %s/.ssh" % (hdfs_homedir, hdfs_homedir), raise_on_non_zero=False)
ssh.send_file_to(self.hdfs_public_ssh_key, "/tmp/id_rsa.pub")
ssh.execute("cat /tmp/id_rsa.pub >> %s/.ssh/authorized_keys && rm /tmp/id_rsa.pub" % hdfs_homedir)
# Start the services
if len(cc.slaves) > 3:
self.config_description = 'starting services'
if self.type == 'hbase':
out.info("Starting services in HBase slave")
ssh.execute("service hadoop-hdfs-datanode restart");
ssh.execute("/etc/init.d/hadoop-0.20-mapreduce-tasktracker restart");
ssh.execute("service hbase-regionserver restart");
else:
out.info("Starting services in Hadoop slave")
ssh.execute("service hadoop-hdfs-datanode restart");
ssh.execute("/etc/init.d/hadoop-0.20-mapreduce-tasktracker restart");
else:
out.info("Not starting slave services before HMaster")
elif self.role == "hive":
# Initialize the service
self.config_description = 'starting services'
# Start the service
out.info("Starting services in Hive")
ssh.execute("service mysql restart");
ssh.execute("service hive-metastore restart");
ssh.execute("service hive-server2 restart");
else:
raise "Unknown role: %s" % (self.role)
# Run status check script
self.config_description = 'checking status'
out.info("Run status check script")
ssh.execute("cd /tmp && ./status-check.sh")
## Remove post install and status check scripts
#self.config_description = 'cleaning /tmp'
#out.info("Remove installation scripts from /tmp")
#ssh.remove_file("/tmp/post-install.sh")
#ssh.remove_file("/tmp/status-check.sh")
#ssh.remove_file("/tmp/hadoop-config.sh.diff")
# Release host lock because we won't touch in this host anymore
bigdata.release('node-%s' % self.hostname)
# Configure related nodes
self.config_description = 'configuring related nodes...'
# Re-read nodes and generate template parameters
try:
bigdata.acquire('cluster')
cc = bigdata.create_config_context(cc.options)
template_params = make_template_params(cc)
finally:
bigdata.release('cluster')
# Reconfigure zoo.cfg on all zookeeper nodes and restart services
if self.role == "hmaster":
# Notice: if we need to add something here, be careful with
# node locking...
pass
elif self.role == "zookeeper":
# Update ZooKeepers (if this is the last zookeeper being configured in parallel)
if cc.zookeepers.index(self) == len(cc.zookeepers) - 1:
for node in cc.zookeepers:
if node != self:
rssh = SSHConnection(node.hostname, out)
try:
self.config_description = "reconfiguring zookeeper of %s" % node.hostname
bigdata.acquire('node-%s' % node.hostname)
rssh.connect()
# copy zoo.cfg including new zookeeper to all zookeeper nodes
# Populate templates and send them to the remote server
out.info("Populating and applying configuration templates")
for content, remote_filename, mode in populate_templates('hbase', 'zookeeper', template_params):
# send configurations to new node
rssh.send_file_to(content, remote_filename, mode=mode)
# restart zookeepers
out.info("Restarting ZooKeeper services in node: %s" % node.hostname)
rssh.execute("service zookeeper-server restart");
finally:
bigdata.release('node-%s' % node.hostname)
rssh.disconnect()
# Update and restart HMaster
if cc.hmaster != None and len(cc.zookeepers) > 3:
rssh = SSHConnection(cc.hmaster.hostname, out)
try:
self.config_description = "updating and restarting hmaster"
bigdata.acquire('node-%s' % cc.hmaster.hostname)
rssh.connect()
# update hbase-site.xml for hmaster
for content, remote_filename, mode in populate_templates('hbase', 'hmaster', template_params):
rssh.send_file_to(content, remote_filename, mode=mode)
# refresh hmaster (hbase)?? - restart required?
rssh.execute("service hbase-master restart");
finally:
bigdata.release('node-%s' % cc.hmaster.hostname)
rssh.disconnect()
elif self.role == "slave":
if len(cc.hmasters) > 0:
rssh = SSHConnection(cc.hmaster.hostname, out)
try:
self.config_description = "updating hmaster files"
bigdata.acquire('node-%s' % cc.hmaster.hostname)
rssh.connect()
# hbase:
# copy regionservers file including new node to hmaster
for content, remote_filename, mode in populate_templates('hbase', 'hmaster', template_params):
# send configurations to new node
rssh.send_file_to(content, remote_filename, mode=mode)
# start hbase services in node
# optional: hbase splitting or other reblancing activities?
# hadoop/hdfs:
# copy slaves file including new node to hmaster
# start hadoop services in node
# optional: use balacer tool for re-distributing blocks to upscaled cluster
## Name node restart
#ssh.execute("service hadoop-hdfs-namenode restart")
finally:
bigdata.release('node-%s' % cc.hmaster.hostname)
rssh.disconnect()
else:
out.info("No masters available, skipping sending region servers file at this time")
self.put_config_state("attached")
out.info("Node configuration finished SUCCESSFULLY.")
except socket.error as e:
raise MgmtException("SSH connection failed: %s" % str(e), e)
except SSHException as e:
raise MgmtException("SSH error: %s" % str(e), e)
except Exception as e:
self.put_config_state("error", str(e))
raise
finally:
if ssh != None:
ssh.disconnect()
out.info("Closed SSH connection to %s" % self.hostname)
return True
def recreate_config_context(self, options):
try:
bigdata.acquire('cluster')
return bigdata.create_config_context(options)
finally:
bigdata.release('cluster')
def detach(self, cc, out):
# Check exclude/decommission status
if not cc.options.force:
out.info("Cheking decommission status of %s" % self.hostname)
if self.is_decommission_in_progress(cc, out):
out.status = "pending"
raise MgmtException("Excluding (decommission) is in progress. The node can't be detached safely, aborting. Use --force paramter to bypass this check.")
# Connect by SSH
hostname = self.hostname
bigdata.acquire("node-%s-files" % hostname)
ssh = SSHConnection(self.hostname, out)
try:
# Set state during the uninstall
out.info("Detaching node %s from the bigdata cluster" % (self))
self.put_config_state("detaching")
# SSH connect
out.info("Connecting to %s" % self.hostname)
ssh.connect()
# Populate /tmp/mongodb templates and send them to the node
template_params = make_template_params(cc, self)
out.info("Populating and applying configuration templates")
for content, remote_filename, mode in populate_templates('mongodb', self.role, template_params):
if remote_filename[0:12] == "/tmp/mongodb":
ssh.send_file_to(content, remote_filename, mode=mode)
# Remove shard from config server
if self.role == "shard":
if cc.options.force:
out.info("Shard draining skipped because of the force switch")
else:
try:
bigdata.acquire('cluster')
rset = cc.find_replicaset_members(self.replsetnum)
if len(rset) >= cc.replica_set_size:
out.info("Removing %s (very slow)" % template_params["SHARD_NAME"])
ssh.execute("mongo %s:27017 /tmp/mongodb/mongos-shard-remove.js" % (cc.first_mongos.hostname))
for node in rset:
node.put_config_state("drained", "")
finally:
bigdata.release('cluster')
# Stop services
out.info("Stopping services")
if self.role == "shard":
out.info(" mongod...")
ssh.execute("/etc/init.d/mongod stop", raise_on_non_zero=False)
elif self.role == "config":
out.info(" mongos...")
ssh.execute("/etc/init.d/mongos stop", raise_on_non_zero=False)
out.info(" config server...")
ssh.execute("/etc/init.d/mongo-cfgsrv stop", raise_on_non_zero=False);
# Uninstall RPMS
out.info("Uninstalling RPM files")
ssh.uninstall(['mongodb-org-server', 'mongodb-org'])
# Run the final cleanup script
out.info("Run cleanup script")
ssh.execute("cd /tmp/mongodb && ./cleanup.sh", raise_on_non_zero=False)
# Delete /etc/bigdata file from the server
ssh.remove_file("/etc/bigdata")
# Drop the node from the config context
if self.role == "shard":
del cc.shards[cc.shards.index(self)]
elif self.role == "config":
del cc.configs[cc.configs.index(self)]
# Make template params
template_params = make_template_params(cc, self)
# Remove post install and status check scripts
out.info("Remove installation related scripts from /tmp")
ssh.execute("rm -fR /tmp/mongodb", raise_on_non_zero=False, read_out_and_err=False)
# Drop from cluster
try:
bigdata.acquire('cluster')
self.drop()
finally:
bigdata.release('cluster')
out.info("The node was dropped SUCCESSFULLY.")
except Exception as e:
self.put_config_state("error", str(e))
if cc.options.force:
out.warn("An error detected but forcing detach still")
self.drop()
raise
finally:
ssh.disconnect()
out.info("Closed SSH the connection")
bigdata.release("node-%s-files" % hostname)
raise MgmtException("Invalid HBase role: %s" % role)
# Configure
out.role = role
if not cc.options.roleonly:
node = hbase.HBaseNode(role, cc.type)
node.hostname = hostname
node.ip_address = ip_address
cclist.append(node)
# Update local /etc/hosts
if not cc.options.roleonly and not cc.options.dns:
update_local_etc_hosts(cc)
# Release cluster lock and start attaching
bigdata.release("cluster")
node.attach(cc, out)
elif cc.type == "mongodb":
# Decide role
role = "unknown"
if options.role != None:
role = options.role
else:
(roles, rolemap) = mongodb.generate_role_list(cc, len(cc.everything) + 1)
if len(cc.configs) < rolemap["configs"]:
role = "config"
else:
role = "shard"
# Choose the node list of config context
cclist = None
def attach(self, cc, out):
out.info("Started installing %s in %s (%s) ---------------------------------------" % (self.role, self.hostname, self.ip_address))
# Regenerate host files of all the managed nodes (before acquiring the node lock)
try:
self.put_config_state("starting")
self.config_description = "regerating /etc/host files"
out.info("Regerating /etc/host files")
bigdata.acquire('cluster')
cc = bigdata.create_config_context(cc.options)
self.regenerate_etc_hosts_files(cc, out)
finally:
bigdata.release('cluster')
# Connect by SSH
bigdata.acquire("node-%s-files" % self.hostname)
ssh = SSHConnection(self.hostname, out)
try:
out.info("Connecting to %s" % self.hostname)
ssh.connect()
# Check operating system type and version and remote hostname
self.check_operating_system_version(cc, out, ssh)
self.check_remote_hostname(cc, out, ssh)
# Set state
self.put_config_state("attaching")
# Decide replica set name
try:
bigdata.acquire('cluster')
cc = bigdata.create_config_context(cc.options)
current_replica_set_size = 0
primary_replicaset_member = None
current_replica_set = []
current_is_the_last_in_rs = False
if self.role == 'shard':
max_rn = 0
replsets = {}
for snode in cc.shards:
if snode.replsetnum != None:
rn = int(snode.replsetnum)
max_rn = max(rn, max_rn)
if rn in replsets:
replsets[rn].append(snode)
else:
replsets[rn] = [snode]
self.replsetnum = None
for rn in replsets:
# Try to find a non-complete replica set and assign this node to it
if len(replsets[rn]) < cc.replica_set_size:
if len(replsets[rn]) + 1 == cc.replica_set_size:
current_is_the_last_in_rs = True
self.replsetnum = rn
primary_replicaset_member = replsets[rn][0]
current_replica_set_size = len(replsets[rn])
current_replica_set = replsets[rn]
if self.replsetnum == None:
self.replsetnum = max_rn + 1
else:
self.replsetnum = None
finally:
bigdata.release('cluster')
# Make template params
template_params = make_template_params(cc, self)
# Check if there is an installation in the node already
self.check_possible_big_data_installation(cc, out, ssh)
# Install RPM files from the custom repository
self.config_description = "installing RPMs"
out.info("Installing RPM files from repository")
ssh.install(['mongodb-org-server', 'mongodb-org'])
self.config_description = None
# Populate templates and send them to the remote server
out.info("Populating and applying configuration templates")
for content, remote_filename, mode in populate_templates('mongodb', self.role, template_params):
ssh.send_file_to(content, remote_filename, mode=mode)
# Run post install script
out.info("Executing post-install script")
ssh.execute("cd /tmp/mongodb && ./post-install.sh", raise_on_non_zero=True)
# Process startups
attached_when_finished = True
if self.role == "config":
# Start cfg server
out.info("Starting config server")
ssh.execute("/etc/rc.d/init.d/mongo-cfgsrv restart")
# Wait for cfg server
out.info("Waiting for mongo config server to get ready and its port bound (can take minutes)")
self.config_description = "initializing"
ssh.execute("cd /tmp/mongodb && ./wait-for-mongo-cfgsrv-startup.sh")
self.config_description = None
# Reconfigure config nodes
if len(cc.configs) >= cc.number_of_config_servers and cc.configs[-1].hostname == self.hostname:
# Since parallel configuration is possible, let's ensure,
# that the other config servers are ready
out.info("Waiting for the other config nodes to become ready")
self.put_config_state("waiting", "waiting for config nodes")
all_config_nodes_ok = False
while not all_config_nodes_ok:
try:
bigdata.acquire('cluster')
cc = bigdata.create_config_context(cc.options)
all_config_nodes_ok = True
for cnode in cc.configs:
if cnode.hostname != self.hostname:
if cnode.config_state != 'attached':
all_config_nodes_ok = False
break
if cnode.config_state == 'error':
# We can't continue if any of the config nodes failed
raise MgmtException("Config node %s has an error. Aborting configuration." % cnode.hostname)
finally:
bigdata.release('cluster')
sleep(2.0)
self.put_config_state("attaching")
out.info("Reconfigurating and restarting mongos processes in config nodes")
try:
bigdata.acquire('cluster')
for cnode in cc.configs:
bigdata.acquire("node-%s-files" % cnode.hostname)
rssh = SSHConnection(cnode.hostname, out)
try:
rssh.connect()
# Update config node templates
for content, remote_filename, mode in populate_templates('mongodb', 'config', template_params):
rssh.send_file_to(content, remote_filename, mode=mode)
# Mongos (re)start
rssh.execute("/etc/init.d/mongos restart")
finally:
rssh.disconnect()
bigdata.release("node-%s-files" % cnode.hostname)
finally:
bigdata.release('cluster')
elif self.role == "shard":
# Start the services
out.info("Starting services of shard node")
ssh.execute("/etc/rc.d/init.d/mongod restart")
out.info("Waiting for mongod to get ready and its port bound (can take minutes)")
self.config_description = "mongod initializing"
ssh.execute("cd /tmp/mongodb && ./wait-for-mongod-startup.sh")
self.config_description = None
# Since parallel configuration is possible, let's ensure, all
# of the config servers are up and running
out.info("Waiting for the config nodes to become ready")
self.put_config_state("waiting", "waiting for config nodes")
all_config_nodes_ok = False
while not all_config_nodes_ok:
try:
bigdata.acquire('cluster')
cc = bigdata.create_config_context(cc.options)
if len(cc.configs) >= cc.number_of_config_servers:
all_config_nodes_ok = True
for cnode in cc.configs:
if cnode.config_state != 'attached':
all_config_nodes_ok = False
break
if cnode.config_state == 'error':
# We can't continue if any of the config nodes failed
raise MgmtException("Config node %s has an error. Aborting configuration." % cnode.hostname)
finally:
bigdata.release('cluster')
sleep(2.0)
self.put_config_state("attaching")
# Set state of the non-last rs members
self.put_config_state("pending", "configuration completed but waiting to be added to the replica set")
attached_when_finished = False
# Operations for the last replica set member
if current_is_the_last_in_rs:
self.put_config_state("attaching")
# Wait until other replica set members are ready
out.info("Waiting for the other members of the replicaset rs%s to start" % self.replsetnum)
self.put_config_state("waiting", "waiting for the replica set members")
all_replica_set_memebers_ready = False
while not all_replica_set_memebers_ready:
all_replica_set_memebers_ready = True
for rnode in current_replica_set:
if rnode.hostname != self.hostname:
if rnode.config_state != 'pending':
all_replica_set_memebers_ready = False
if rnode.config_state in ['error', None]:
raise MgmtException("Aborting replica set configuration because node %s failed" % rnode.hostname)
sleep(2.0)
self.put_config_state("attaching")
# Initiate replica set
try:
bigdata.acquire('cluster')
self.put_config_state("attaching", "initiating replica set")
out.info("Initiating the replica set rs%s " % self.replsetnum)
ssh.execute("mongo localhost:27018 /tmp/mongodb/mongod-replicaset-initiate.js")
self.put_config_state("attaching")
finally:
bigdata.release('cluster')
try:
bigdata.acquire('cluster')
# Repopulate templates and send them to the remote server (to update mongos-shard-add.js)
out.info("Populating and applying configuration templates")
cc = bigdata.create_config_context(cc.options)
template_params = make_template_params(cc, self)
for content, remote_filename, mode in populate_templates('mongodb', self.role, template_params):
ssh.send_file_to(content, remote_filename, mode=mode)
# Create a shard from the replica set
bigdata.acquire('cluster')
self.put_config_state("attaching", "creating shard")
out.info("Creating a shard for replica set rs%s " % self.replsetnum)
ssh.execute("mongo %s:27017 /tmp/mongodb/mongos-shard-add.js" % cc.first_mongos.hostname)
rset = cc.find_replicaset_members(self.replsetnum)
self.put_config_state("attaching")
for node in rset:
node.put_config_state("attached", "")
finally:
bigdata.release('cluster')
attached_when_finished = True
else:
raise "Unknown role: %s" % (self.role)
# Run status check script
out.info("Run status check script")
ssh.execute("cd /tmp/mongodb && ./status-check.sh")
# Remove post install and status check scripts
#out.info("Remove installation related scripts from /tmp")
#ssh.execute("rm -fR /tmp/mongodb")
if attached_when_finished:
self.put_config_state("attached")
out.info("Node configuration finished SUCCESSFULLY.")
except socket.error as e:
raise MgmtException("SSH connection failed: %s" % str(e), e)
except SSHException as e:
raise MgmtException("SSH error: %s" % str(e), e)
except Exception as e:
self.put_config_state("error", str(e))
raise
finally:
if ssh != None:
ssh.disconnect()
out.info("Closed SSH connection to %s" % self.hostname)
bigdata.release("node-%s-files" % self.hostname)
return True