import os
import os.path
-import sha
import re
import logging
+import tempfile
+import time
from ganeti import rpc
from ganeti import ssh
from ganeti import constants
from ganeti import objects
from ganeti import ssconf
+from ganeti import serializer
+from ganeti import hypervisor
-def _InitSSHSetup(node):
- """Setup the SSH configuration for the cluster.
+def _InitSSHSetup():
+ """Setup the SSH configuration for the cluster.
This generates a dsa keypair for root, adds the pub key to the
permitted hosts and adds the hostkey to its own known hosts.
- Args:
- node: the name of this host as a fqdn
-
"""
priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS)
raise errors.OpExecError("Could not generate ssh keypair, error %s" %
result.output)
- f = open(pub_key, 'r')
+ utils.AddAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
+
+
+def GenerateSelfSignedSslCert(file_name, validity=(365 * 5)):
+ """Generates a self-signed SSL certificate.
+
+ @type file_name: str
+ @param file_name: Path to output file
+ @type validity: int
+ @param validity: Validity for certificate in days
+
+ """
+ (fd, tmp_file_name) = tempfile.mkstemp(dir=os.path.dirname(file_name))
try:
- utils.AddAuthorizedKey(auth_keys, f.read(8192))
+ try:
+ # Set permissions before writing key
+ os.chmod(tmp_file_name, 0600)
+
+ result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024",
+ "-days", str(validity), "-nodes", "-x509",
+ "-keyout", tmp_file_name, "-out", tmp_file_name,
+ "-batch"])
+ if result.failed:
+ raise errors.OpExecError("Could not generate SSL certificate, command"
+ " %s had exitcode %s and error message %s" %
+ (result.cmd, result.exit_code, result.output))
+
+ # Make read-only
+ os.chmod(tmp_file_name, 0400)
+
+ os.rename(tmp_file_name, file_name)
+ finally:
+ utils.RemoveFile(tmp_file_name)
finally:
- f.close()
+ os.close(fd)
+
+
+def GenerateHmacKey(file_name):
+ """Writes a new HMAC key.
+ @type file_name: str
+ @param file_name: Path to output file
-def _InitGanetiServerSetup():
+ """
+ utils.WriteFile(file_name, data=utils.GenerateSecret(), mode=0400)
+
+
+def _InitGanetiServerSetup(master_name):
"""Setup the necessary configuration for the initial node daemon.
This creates the nodepass file containing the shared password for
the cluster and also generates the SSL certificate.
"""
- result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024",
- "-days", str(365*5), "-nodes", "-x509",
- "-keyout", constants.SSL_CERT_FILE,
- "-out", constants.SSL_CERT_FILE, "-batch"])
- if result.failed:
- raise errors.OpExecError("could not generate server ssl cert, command"
- " %s had exitcode %s and error message %s" %
- (result.cmd, result.exit_code, result.output))
+ GenerateSelfSignedSslCert(constants.SSL_CERT_FILE)
+
+ # Don't overwrite existing file
+ if not os.path.exists(constants.RAPI_CERT_FILE):
+ GenerateSelfSignedSslCert(constants.RAPI_CERT_FILE)
- os.chmod(constants.SSL_CERT_FILE, 0400)
+ if not os.path.exists(constants.HMAC_CLUSTER_KEY):
+ GenerateHmacKey(constants.HMAC_CLUSTER_KEY)
result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"])
" had exitcode %s and error %s" %
(result.cmd, result.exit_code, result.output))
+ # Wait for node daemon to become responsive
+ def _CheckNodeDaemon():
+ result = rpc.RpcRunner.call_version([master_name])[master_name]
+ if result.fail_msg:
+ raise utils.RetryAgain()
-def InitCluster(cluster_name, mac_prefix, def_bridge,
- master_netdev, file_storage_dir,
- secondary_ip=None,
- vg_name=None, beparams=None, hvparams=None,
- enabled_hypervisors=None, default_hypervisor=None):
+ try:
+ utils.Retry(_CheckNodeDaemon, 1.0, 10.0)
+ except utils.RetryTimeout:
+ raise errors.OpExecError("Node daemon didn't answer queries within"
+ " 10 seconds")
+
+def InitCluster(cluster_name, mac_prefix,
+ master_netdev, file_storage_dir, candidate_pool_size,
+ secondary_ip=None, vg_name=None, beparams=None,
+ nicparams=None, hvparams=None, enabled_hypervisors=None,
+ modify_etc_hosts=True, modify_ssh_setup=True):
"""Initialise the cluster.
+ @type candidate_pool_size: int
+ @param candidate_pool_size: master candidate pool size
+
"""
+ # TODO: complete the docstring
if config.ConfigWriter.IsCluster():
- raise errors.OpPrereqError("Cluster is already initialised")
+ raise errors.OpPrereqError("Cluster is already initialised",
+ errors.ECODE_STATE)
+
+ if not enabled_hypervisors:
+ raise errors.OpPrereqError("Enabled hypervisors list must contain at"
+ " least one member", errors.ECODE_INVAL)
+ invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
+ if invalid_hvs:
+ raise errors.OpPrereqError("Enabled hypervisors contains invalid"
+ " entries: %s" % invalid_hvs,
+ errors.ECODE_INVAL)
hostname = utils.HostInfo()
if hostname.ip.startswith("127."):
raise errors.OpPrereqError("This host's IP resolves to the private"
" range (%s). Please fix DNS or %s." %
- (hostname.ip, constants.ETC_HOSTS))
+ (hostname.ip, constants.ETC_HOSTS),
+ errors.ECODE_ENVIRON)
if not utils.OwnIpAddress(hostname.ip):
raise errors.OpPrereqError("Inconsistency: this host's name resolves"
" to %s,\nbut this ip address does not"
- " belong to this host."
- " Aborting." % hostname.ip)
+ " belong to this host. Aborting." %
+ hostname.ip, errors.ECODE_ENVIRON)
clustername = utils.HostInfo(cluster_name)
if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
timeout=5):
- raise errors.OpPrereqError("Cluster IP already active. Aborting.")
+ raise errors.OpPrereqError("Cluster IP already active. Aborting.",
+ errors.ECODE_NOTUNIQUE)
if secondary_ip:
if not utils.IsValidIP(secondary_ip):
- raise errors.OpPrereqError("Invalid secondary ip given")
+ raise errors.OpPrereqError("Invalid secondary ip given",
+ errors.ECODE_INVAL)
if (secondary_ip != hostname.ip and
not utils.OwnIpAddress(secondary_ip)):
raise errors.OpPrereqError("You gave %s as secondary IP,"
" but it does not belong to this host." %
- secondary_ip)
+ secondary_ip, errors.ECODE_ENVIRON)
else:
secondary_ip = hostname.ip
constants.MIN_VG_SIZE)
if vgstatus:
raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if"
- " you are not using lvm" % vgstatus)
+ " you are not using lvm" % vgstatus,
+ errors.ECODE_INVAL)
file_storage_dir = os.path.normpath(file_storage_dir)
if not os.path.isabs(file_storage_dir):
raise errors.OpPrereqError("The file storage directory you passed is"
- " not an absolute path.")
+ " not an absolute path.", errors.ECODE_INVAL)
if not os.path.exists(file_storage_dir):
try:
os.makedirs(file_storage_dir, 0750)
except OSError, err:
raise errors.OpPrereqError("Cannot create file storage directory"
- " '%s': %s" %
- (file_storage_dir, err))
+ " '%s': %s" % (file_storage_dir, err),
+ errors.ECODE_ENVIRON)
if not os.path.isdir(file_storage_dir):
raise errors.OpPrereqError("The file storage directory '%s' is not"
- " a directory." % file_storage_dir)
+ " a directory." % file_storage_dir,
+ errors.ECODE_ENVIRON)
if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix):
- raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix)
+ raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix,
+ errors.ECODE_INVAL)
result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev])
if result.failed:
raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" %
(master_netdev,
- result.output.strip()))
+ result.output.strip()), errors.ECODE_INVAL)
if not (os.path.isfile(constants.NODE_INITD_SCRIPT) and
os.access(constants.NODE_INITD_SCRIPT, os.X_OK)):
raise errors.OpPrereqError("Init.d script '%s' missing or not"
- " executable." % constants.NODE_INITD_SCRIPT)
+ " executable." % constants.NODE_INITD_SCRIPT,
+ errors.ECODE_ENVIRON)
+
+ dirs = [(constants.RUN_GANETI_DIR, constants.RUN_DIRS_MODE)]
+ utils.EnsureDirs(dirs)
- utils.CheckBEParams(beparams)
+ utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES)
+ utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES)
+ objects.NIC.CheckParameterSyntax(nicparams)
+
+ # hvparams is a mapping of hypervisor->hvparams dict
+ for hv_name, hv_params in hvparams.iteritems():
+ utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
+ hv_class = hypervisor.GetHypervisor(hv_name)
+ hv_class.CheckParameterSyntax(hv_params)
# set up the inter-node password and certificate
- _InitGanetiServerSetup()
+ _InitGanetiServerSetup(hostname.name)
# set up ssh config and /etc/hosts
- f = open(constants.SSH_HOST_RSA_PUB, 'r')
- try:
- sshline = f.read()
- finally:
- f.close()
+ sshline = utils.ReadFile(constants.SSH_HOST_RSA_PUB)
sshkey = sshline.split(" ")[1]
- utils.AddHostToEtcHosts(hostname.name)
- _InitSSHSetup(hostname.name)
+ if modify_etc_hosts:
+ utils.AddHostToEtcHosts(hostname.name)
+
+ if modify_ssh_setup:
+ _InitSSHSetup()
+
+ now = time.time()
# init of cluster config file
cluster_config = objects.Cluster(
highest_used_port=(constants.FIRST_DRBD_PORT - 1),
mac_prefix=mac_prefix,
volume_group_name=vg_name,
- default_bridge=def_bridge,
tcpudp_port_pool=set(),
master_node=hostname.name,
master_ip=clustername.ip,
cluster_name=clustername.name,
file_storage_dir=file_storage_dir,
enabled_hypervisors=enabled_hypervisors,
- default_hypervisor=default_hypervisor,
- beparams={constants.BEGR_DEFAULT: beparams},
+ beparams={constants.PP_DEFAULT: beparams},
+ nicparams={constants.PP_DEFAULT: nicparams},
hvparams=hvparams,
+ candidate_pool_size=candidate_pool_size,
+ modify_etc_hosts=modify_etc_hosts,
+ modify_ssh_setup=modify_ssh_setup,
+ ctime=now,
+ mtime=now,
+ uuid=utils.NewUUID(),
)
master_node_config = objects.Node(name=hostname.name,
primary_ip=hostname.ip,
secondary_ip=secondary_ip,
- serial_no=1)
-
- cfg = InitConfig(constants.CONFIG_VERSION,
- cluster_config, master_node_config)
+ serial_no=1,
+ master_candidate=True,
+ offline=False, drained=False,
+ )
+ InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config)
+ cfg = config.ConfigWriter()
ssh.WriteKnownHostsFile(cfg, constants.SSH_KNOWN_HOSTS_FILE)
+ cfg.Update(cfg.GetClusterInfo(), logging.error)
# start the master ip
# TODO: Review rpc call from bootstrap
- rpc.RpcRunner.call_node_start_master(hostname.name, True)
+ # TODO: Warn on failed start master
+ rpc.RpcRunner.call_node_start_master(hostname.name, True, False)
def InitConfig(version, cluster_config, master_node_config,
node, and no instances.
@type version: int
- @param version: Configuration version
- @type cluster_config: objects.Cluster
- @param cluster_config: Cluster configuration
- @type master_node_config: objects.Node
- @param master_node_config: Master node configuration
- @type file_name: string
- @param file_name: Configuration file path
-
- @rtype: ssconf.SimpleConfigWriter
- @returns: Initialized config instance
+ @param version: configuration version
+ @type cluster_config: L{objects.Cluster}
+ @param cluster_config: cluster configuration
+ @type master_node_config: L{objects.Node}
+ @param master_node_config: master node configuration
+ @type cfg_file: string
+ @param cfg_file: configuration file path
"""
nodes = {
master_node_config.name: master_node_config,
}
+ now = time.time()
config_data = objects.ConfigData(version=version,
cluster=cluster_config,
nodes=nodes,
instances={},
- serial_no=1)
- cfg = ssconf.SimpleConfigWriter.FromDict(config_data.ToDict(), cfg_file)
- cfg.Save()
-
- return cfg
+ serial_no=1,
+ ctime=now, mtime=now)
+ utils.WriteFile(cfg_file,
+ data=serializer.Dump(config_data.ToDict()),
+ mode=0600)
def FinalizeClusterDestroy(master):
begun in cmdlib.LUDestroyOpcode.
"""
- if not rpc.RpcRunner.call_node_stop_master(master, True):
- logging.warning("Could not disable the master role")
- if not rpc.RpcRunner.call_node_leave_cluster(master):
- logging.warning("Could not shutdown the node daemon and cleanup the node")
+ cfg = config.ConfigWriter()
+ modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup
+ result = rpc.RpcRunner.call_node_stop_master(master, True)
+ msg = result.fail_msg
+ if msg:
+ logging.warning("Could not disable the master role: %s", msg)
+ result = rpc.RpcRunner.call_node_leave_cluster(master, modify_ssh_setup)
+ msg = result.fail_msg
+ if msg:
+ logging.warning("Could not shutdown the node daemon and cleanup"
+ " the node: %s", msg)
def SetupNodeDaemon(cluster_name, node, ssh_key_check):
"""
sshrunner = ssh.SshRunner(cluster_name)
- gntpem = utils.ReadFile(constants.SSL_CERT_FILE)
+
+ noded_cert = utils.ReadFile(constants.SSL_CERT_FILE)
+ rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE)
+ hmac_key = utils.ReadFile(constants.HMAC_CLUSTER_KEY)
+
# in the base64 pem encoding, neither '!' nor '.' are valid chars,
# so we use this to detect an invalid certificate; as long as the
# cert doesn't contain this, the here-document will be correctly
- # parsed by the shell sequence below
- if re.search('^!EOF\.', gntpem, re.MULTILINE):
- raise errors.OpExecError("invalid PEM encoding in the SSL certificate")
- if not gntpem.endswith("\n"):
- raise errors.OpExecError("PEM must end with newline")
+ # parsed by the shell sequence below. HMAC keys are hexadecimal strings,
+ # so the same restrictions apply.
+ for content in (noded_cert, rapi_cert, hmac_key):
+ if re.search('^!EOF\.', content, re.MULTILINE):
+ raise errors.OpExecError("invalid SSL certificate or HMAC key")
+
+ if not noded_cert.endswith("\n"):
+ noded_cert += "\n"
+ if not rapi_cert.endswith("\n"):
+ rapi_cert += "\n"
+ if not hmac_key.endswith("\n"):
+ hmac_key += "\n"
# set up inter-node password and certificate and restarts the node daemon
# and then connect with ssh to set password and start ganeti-noded
# either by being constants or by the checks above
mycommand = ("umask 077 && "
"cat > '%s' << '!EOF.' && \n"
- "%s!EOF.\n%s restart" %
- (constants.SSL_CERT_FILE, gntpem,
+ "%s!EOF.\n"
+ "cat > '%s' << '!EOF.' && \n"
+ "%s!EOF.\n"
+ "cat > '%s' << '!EOF.' && \n"
+ "%s!EOF.\n"
+ "chmod 0400 %s %s %s && "
+ "%s restart" %
+ (constants.SSL_CERT_FILE, noded_cert,
+ constants.RAPI_CERT_FILE, rapi_cert,
+ constants.HMAC_CLUSTER_KEY, hmac_key,
+ constants.SSL_CERT_FILE, constants.RAPI_CERT_FILE,
+ constants.HMAC_CLUSTER_KEY,
constants.NODE_INITD_SCRIPT))
result = sshrunner.Run(node, 'root', mycommand, batch=False,
(node, result.fail_reason, result.output))
-def MasterFailover():
+def MasterFailover(no_voting=False):
"""Failover the master node.
This checks that we are not already the master, and will cause the
current master to cease being master, and the non-master to become
new master.
+ @type no_voting: boolean
+ @param no_voting: force the operation without remote nodes agreement
+ (dangerous)
+
"""
sstore = ssconf.SimpleStore()
raise errors.OpPrereqError("This commands must be run on the node"
" where you want the new master to be."
" %s is already the master" %
- old_master)
+ old_master, errors.ECODE_INVAL)
if new_master not in mc_list:
mc_no_master = [name for name in mc_list if name != old_master]
" as master candidates. Only these nodes"
" can become masters. Current list of"
" master candidates is:\n"
- "%s" % ('\n'.join(mc_no_master)))
-
- vote_list = GatherMasterVotes(node_list)
-
- if vote_list:
- voted_master = vote_list[0][0]
- if voted_master is None:
- raise errors.OpPrereqError("Cluster is inconsistent, most nodes did not"
- " respond.")
- elif voted_master != old_master:
- raise errors.OpPrereqError("I have wrong configuration, I believe the"
- " master is %s but the other nodes voted for"
- " %s. Please resync the configuration of"
- " this node." % (old_master, voted_master))
+ "%s" % ('\n'.join(mc_no_master)),
+ errors.ECODE_STATE)
+
+ if not no_voting:
+ vote_list = GatherMasterVotes(node_list)
+
+ if vote_list:
+ voted_master = vote_list[0][0]
+ if voted_master is None:
+ raise errors.OpPrereqError("Cluster is inconsistent, most nodes did"
+ " not respond.", errors.ECODE_ENVIRON)
+ elif voted_master != old_master:
+ raise errors.OpPrereqError("I have a wrong configuration, I believe"
+ " the master is %s but the other nodes"
+ " voted %s. Please resync the configuration"
+ " of this node." %
+ (old_master, voted_master),
+ errors.ECODE_STATE)
# end checks
rcode = 0
logging.info("Setting master to %s, old master: %s", new_master, old_master)
- if not rpc.RpcRunner.call_node_stop_master(old_master, True):
+ result = rpc.RpcRunner.call_node_stop_master(old_master, True)
+ msg = result.fail_msg
+ if msg:
logging.error("Could not disable the master role on the old master"
- " %s, please disable manually", old_master)
+ " %s, please disable manually: %s", old_master, msg)
# Here we have a phase where no master should be running
cluster_info.master_node = new_master
# this will also regenerate the ssconf files, since we updated the
# cluster info
- cfg.Update(cluster_info)
+ cfg.Update(cluster_info, logging.error)
- if not rpc.RpcRunner.call_node_start_master(new_master, True):
+ result = rpc.RpcRunner.call_node_start_master(new_master, True, no_voting)
+ msg = result.fail_msg
+ if msg:
logging.error("Could not start the master role on the new master"
- " %s, please check", new_master)
+ " %s, please check: %s", new_master, msg)
rcode = 1
return rcode
+def GetMaster():
+ """Returns the current master node.
+
+ This is a separate function in bootstrap since it's needed by
+ gnt-cluster, and instead of importing directly ssconf, it's better
+ to abstract it in bootstrap, where we do use ssconf in other
+ functions too.
+
+ """
+ sstore = ssconf.SimpleStore()
+
+ old_master, _ = ssconf.GetMasterAndMyself(sstore)
+
+ return old_master
+
+
def GatherMasterVotes(node_list):
"""Check the agreement on who is the master.
@type node_list: list
@param node_list: the list of nodes to query for master info; the current
- node wil be removed if it is in the list
+ node will be removed if it is in the list
@rtype: list
@return: list of (node, votes)
# this should not happen (unless internal error in rpc)
logging.critical("Can't complete rpc call, aborting master startup")
return [(None, len(node_list))]
- positive = negative = 0
- other_masters = {}
votes = {}
for node in results:
- if not isinstance(results[node], (tuple, list)) or len(results[node]) < 3:
- # here the rpc layer should have already logged errors
+ nres = results[node]
+ data = nres.payload
+ msg = nres.fail_msg
+ fail = False
+ if msg:
+ logging.warning("Error contacting node %s: %s", node, msg)
+ fail = True
+ elif not isinstance(data, (tuple, list)) or len(data) < 3:
+ logging.warning("Invalid data received from node %s: %s", node, data)
+ fail = True
+ if fail:
if None not in votes:
votes[None] = 0
votes[None] += 1
continue
- master_node = results[node][2]
+ master_node = data[2]
if master_node not in votes:
votes[master_node] = 0
votes[master_node] += 1