root / lib / bootstrap.py @ f6bd6e98
History | View | Annotate | Download (12.5 kB)
1 |
#
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2006, 2007, 2008 Google Inc.
|
5 |
#
|
6 |
# This program is free software; you can redistribute it and/or modify
|
7 |
# it under the terms of the GNU General Public License as published by
|
8 |
# the Free Software Foundation; either version 2 of the License, or
|
9 |
# (at your option) any later version.
|
10 |
#
|
11 |
# This program is distributed in the hope that it will be useful, but
|
12 |
# WITHOUT ANY WARRANTY; without even the implied warranty of
|
13 |
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
14 |
# General Public License for more details.
|
15 |
#
|
16 |
# You should have received a copy of the GNU General Public License
|
17 |
# along with this program; if not, write to the Free Software
|
18 |
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
|
19 |
# 02110-1301, USA.
|
20 |
|
21 |
|
22 |
"""Functions to bootstrap a new cluster.
|
23 |
|
24 |
"""
|
25 |
|
26 |
import os |
27 |
import os.path |
28 |
import sha |
29 |
import re |
30 |
import logging |
31 |
|
32 |
from ganeti import rpc |
33 |
from ganeti import ssh |
34 |
from ganeti import utils |
35 |
from ganeti import errors |
36 |
from ganeti import config |
37 |
from ganeti import constants |
38 |
from ganeti import objects |
39 |
from ganeti import ssconf |
40 |
|
41 |
|
42 |
def _InitSSHSetup(node): |
43 |
"""Setup the SSH configuration for the cluster.
|
44 |
|
45 |
|
46 |
This generates a dsa keypair for root, adds the pub key to the
|
47 |
permitted hosts and adds the hostkey to its own known hosts.
|
48 |
|
49 |
Args:
|
50 |
node: the name of this host as a fqdn
|
51 |
|
52 |
"""
|
53 |
priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS) |
54 |
|
55 |
for name in priv_key, pub_key: |
56 |
if os.path.exists(name):
|
57 |
utils.CreateBackup(name) |
58 |
utils.RemoveFile(name) |
59 |
|
60 |
result = utils.RunCmd(["ssh-keygen", "-t", "dsa", |
61 |
"-f", priv_key,
|
62 |
"-q", "-N", ""]) |
63 |
if result.failed:
|
64 |
raise errors.OpExecError("Could not generate ssh keypair, error %s" % |
65 |
result.output) |
66 |
|
67 |
f = open(pub_key, 'r') |
68 |
try:
|
69 |
utils.AddAuthorizedKey(auth_keys, f.read(8192))
|
70 |
finally:
|
71 |
f.close() |
72 |
|
73 |
|
74 |
def _InitGanetiServerSetup(ss): |
75 |
"""Setup the necessary configuration for the initial node daemon.
|
76 |
|
77 |
This creates the nodepass file containing the shared password for
|
78 |
the cluster and also generates the SSL certificate.
|
79 |
|
80 |
Args:
|
81 |
ss: A WritableSimpleStore
|
82 |
|
83 |
"""
|
84 |
# Create pseudo random password
|
85 |
randpass = utils.GenerateSecret() |
86 |
# and write it into sstore
|
87 |
ss.SetKey(ss.SS_NODED_PASS, randpass) |
88 |
|
89 |
result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024", |
90 |
"-days", str(365*5), "-nodes", "-x509", |
91 |
"-keyout", constants.SSL_CERT_FILE,
|
92 |
"-out", constants.SSL_CERT_FILE, "-batch"]) |
93 |
if result.failed:
|
94 |
raise errors.OpExecError("could not generate server ssl cert, command" |
95 |
" %s had exitcode %s and error message %s" %
|
96 |
(result.cmd, result.exit_code, result.output)) |
97 |
|
98 |
os.chmod(constants.SSL_CERT_FILE, 0400)
|
99 |
|
100 |
result = utils.RunCmd([constants.NODE_INITD_SCRIPT, "restart"])
|
101 |
|
102 |
if result.failed:
|
103 |
raise errors.OpExecError("Could not start the node daemon, command %s" |
104 |
" had exitcode %s and error %s" %
|
105 |
(result.cmd, result.exit_code, result.output)) |
106 |
|
107 |
|
108 |
def InitCluster(cluster_name, hypervisor_type, mac_prefix, def_bridge, |
109 |
master_netdev, file_storage_dir, |
110 |
secondary_ip=None,
|
111 |
vg_name=None):
|
112 |
"""Initialise the cluster.
|
113 |
|
114 |
"""
|
115 |
if config.ConfigWriter.IsCluster():
|
116 |
raise errors.OpPrereqError("Cluster is already initialised") |
117 |
|
118 |
if hypervisor_type == constants.HT_XEN_HVM31:
|
119 |
if not os.path.exists(constants.VNC_PASSWORD_FILE): |
120 |
raise errors.OpPrereqError("Please prepare the cluster VNC" |
121 |
"password file %s" %
|
122 |
constants.VNC_PASSWORD_FILE) |
123 |
|
124 |
hostname = utils.HostInfo() |
125 |
|
126 |
if hostname.ip.startswith("127."): |
127 |
raise errors.OpPrereqError("This host's IP resolves to the private" |
128 |
" range (%s). Please fix DNS or %s." %
|
129 |
(hostname.ip, constants.ETC_HOSTS)) |
130 |
|
131 |
if not utils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT, |
132 |
source=constants.LOCALHOST_IP_ADDRESS): |
133 |
raise errors.OpPrereqError("Inconsistency: this host's name resolves" |
134 |
" to %s,\nbut this ip address does not"
|
135 |
" belong to this host."
|
136 |
" Aborting." % hostname.ip)
|
137 |
|
138 |
clustername = utils.HostInfo(cluster_name) |
139 |
|
140 |
if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
|
141 |
timeout=5):
|
142 |
raise errors.OpPrereqError("Cluster IP already active. Aborting.") |
143 |
|
144 |
if secondary_ip:
|
145 |
if not utils.IsValidIP(secondary_ip): |
146 |
raise errors.OpPrereqError("Invalid secondary ip given") |
147 |
if (secondary_ip != hostname.ip and |
148 |
(not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
|
149 |
source=constants.LOCALHOST_IP_ADDRESS))): |
150 |
raise errors.OpPrereqError("You gave %s as secondary IP," |
151 |
" but it does not belong to this host." %
|
152 |
secondary_ip) |
153 |
else:
|
154 |
secondary_ip = hostname.ip |
155 |
|
156 |
if vg_name is not None: |
157 |
# Check if volume group is valid
|
158 |
vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name, |
159 |
constants.MIN_VG_SIZE) |
160 |
if vgstatus:
|
161 |
raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if" |
162 |
" you are not using lvm" % vgstatus)
|
163 |
|
164 |
file_storage_dir = os.path.normpath(file_storage_dir) |
165 |
|
166 |
if not os.path.isabs(file_storage_dir): |
167 |
raise errors.OpPrereqError("The file storage directory you passed is" |
168 |
" not an absolute path.")
|
169 |
|
170 |
if not os.path.exists(file_storage_dir): |
171 |
try:
|
172 |
os.makedirs(file_storage_dir, 0750)
|
173 |
except OSError, err: |
174 |
raise errors.OpPrereqError("Cannot create file storage directory" |
175 |
" '%s': %s" %
|
176 |
(file_storage_dir, err)) |
177 |
|
178 |
if not os.path.isdir(file_storage_dir): |
179 |
raise errors.OpPrereqError("The file storage directory '%s' is not" |
180 |
" a directory." % file_storage_dir)
|
181 |
|
182 |
if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix): |
183 |
raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix) |
184 |
|
185 |
if hypervisor_type not in constants.HYPER_TYPES: |
186 |
raise errors.OpPrereqError("Invalid hypervisor type given '%s'" % |
187 |
hypervisor_type) |
188 |
|
189 |
result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev]) |
190 |
if result.failed:
|
191 |
raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" % |
192 |
(master_netdev, |
193 |
result.output.strip())) |
194 |
|
195 |
if not (os.path.isfile(constants.NODE_INITD_SCRIPT) and |
196 |
os.access(constants.NODE_INITD_SCRIPT, os.X_OK)): |
197 |
raise errors.OpPrereqError("Init.d script '%s' missing or not" |
198 |
" executable." % constants.NODE_INITD_SCRIPT)
|
199 |
|
200 |
# set up the simple store
|
201 |
ss = ssconf.WritableSimpleStore() |
202 |
ss.SetKey(ss.SS_HYPERVISOR, hypervisor_type) |
203 |
ss.SetKey(ss.SS_MASTER_NODE, hostname.name) |
204 |
ss.SetKey(ss.SS_MASTER_IP, clustername.ip) |
205 |
ss.SetKey(ss.SS_MASTER_NETDEV, master_netdev) |
206 |
ss.SetKey(ss.SS_CLUSTER_NAME, clustername.name) |
207 |
ss.SetKey(ss.SS_FILE_STORAGE_DIR, file_storage_dir) |
208 |
ss.SetKey(ss.SS_CONFIG_VERSION, constants.CONFIG_VERSION) |
209 |
|
210 |
# set up the inter-node password and certificate
|
211 |
_InitGanetiServerSetup(ss) |
212 |
|
213 |
# set up ssh config and /etc/hosts
|
214 |
f = open(constants.SSH_HOST_RSA_PUB, 'r') |
215 |
try:
|
216 |
sshline = f.read() |
217 |
finally:
|
218 |
f.close() |
219 |
sshkey = sshline.split(" ")[1] |
220 |
|
221 |
utils.AddHostToEtcHosts(hostname.name) |
222 |
_InitSSHSetup(hostname.name) |
223 |
|
224 |
# init of cluster config file
|
225 |
cluster_config = objects.Cluster( |
226 |
serial_no=1,
|
227 |
rsahostkeypub=sshkey, |
228 |
highest_used_port=(constants.FIRST_DRBD_PORT - 1),
|
229 |
mac_prefix=mac_prefix, |
230 |
volume_group_name=vg_name, |
231 |
default_bridge=def_bridge, |
232 |
tcpudp_port_pool=set(),
|
233 |
hypervisor=hypervisor_type, |
234 |
master_node=hostname.name, |
235 |
master_ip=clustername.ip, |
236 |
master_netdev=master_netdev, |
237 |
cluster_name=clustername.name, |
238 |
file_storage_dir=file_storage_dir, |
239 |
) |
240 |
master_node_config = objects.Node(name=hostname.name, |
241 |
primary_ip=hostname.ip, |
242 |
secondary_ip=secondary_ip) |
243 |
cfg = config.ConfigWriter() |
244 |
cfg.InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config) |
245 |
|
246 |
ssh.WriteKnownHostsFile(cfg, ss, constants.SSH_KNOWN_HOSTS_FILE) |
247 |
|
248 |
# start the master ip
|
249 |
# TODO: Review rpc call from bootstrap
|
250 |
rpc.call_node_start_master(hostname.name, True)
|
251 |
|
252 |
|
253 |
def FinalizeClusterDestroy(master): |
254 |
"""Execute the last steps of cluster destroy
|
255 |
|
256 |
This function shuts down all the daemons, completing the destroy
|
257 |
begun in cmdlib.LUDestroyOpcode.
|
258 |
|
259 |
"""
|
260 |
if not rpc.call_node_stop_master(master, True): |
261 |
logging.warning("Could not disable the master role")
|
262 |
if not rpc.call_node_leave_cluster(master): |
263 |
logging.warning("Could not shutdown the node daemon and cleanup the node")
|
264 |
|
265 |
|
266 |
def SetupNodeDaemon(node, ssh_key_check): |
267 |
"""Add a node to the cluster.
|
268 |
|
269 |
This function must be called before the actual opcode, and will ssh
|
270 |
to the remote node, copy the needed files, and start ganeti-noded,
|
271 |
allowing the master to do the rest via normal rpc calls.
|
272 |
|
273 |
Args:
|
274 |
node: fully qualified domain name for the new node
|
275 |
|
276 |
"""
|
277 |
ss = ssconf.SimpleStore() |
278 |
sshrunner = ssh.SshRunner(ss) |
279 |
gntpass = ss.GetNodeDaemonPassword() |
280 |
if not re.match('^[a-zA-Z0-9.]{1,64}$', gntpass): |
281 |
raise errors.OpExecError("ganeti password corruption detected") |
282 |
f = open(constants.SSL_CERT_FILE)
|
283 |
try:
|
284 |
gntpem = f.read(8192)
|
285 |
finally:
|
286 |
f.close() |
287 |
# in the base64 pem encoding, neither '!' nor '.' are valid chars,
|
288 |
# so we use this to detect an invalid certificate; as long as the
|
289 |
# cert doesn't contain this, the here-document will be correctly
|
290 |
# parsed by the shell sequence below
|
291 |
if re.search('^!EOF\.', gntpem, re.MULTILINE): |
292 |
raise errors.OpExecError("invalid PEM encoding in the SSL certificate") |
293 |
if not gntpem.endswith("\n"): |
294 |
raise errors.OpExecError("PEM must end with newline") |
295 |
|
296 |
# set up inter-node password and certificate and restarts the node daemon
|
297 |
# and then connect with ssh to set password and start ganeti-noded
|
298 |
# note that all the below variables are sanitized at this point,
|
299 |
# either by being constants or by the checks above
|
300 |
mycommand = ("umask 077 && "
|
301 |
"echo '%s' > '%s' && "
|
302 |
"cat > '%s' << '!EOF.' && \n"
|
303 |
"%s!EOF.\n%s restart" %
|
304 |
(gntpass, ss.KeyToFilename(ss.SS_NODED_PASS), |
305 |
constants.SSL_CERT_FILE, gntpem, |
306 |
constants.NODE_INITD_SCRIPT)) |
307 |
|
308 |
result = sshrunner.Run(node, 'root', mycommand, batch=False, |
309 |
ask_key=ssh_key_check, |
310 |
use_cluster_key=False,
|
311 |
strict_host_check=ssh_key_check) |
312 |
if result.failed:
|
313 |
raise errors.OpExecError("Remote command on node %s, error: %s," |
314 |
" output: %s" %
|
315 |
(node, result.fail_reason, result.output)) |
316 |
|
317 |
return 0 |
318 |
|
319 |
|
320 |
def MasterFailover(): |
321 |
"""Failover the master node.
|
322 |
|
323 |
This checks that we are not already the master, and will cause the
|
324 |
current master to cease being master, and the non-master to become
|
325 |
new master.
|
326 |
|
327 |
"""
|
328 |
ss = ssconf.WritableSimpleStore() |
329 |
|
330 |
new_master = utils.HostInfo().name |
331 |
old_master = ss.GetMasterNode() |
332 |
|
333 |
if old_master == new_master:
|
334 |
raise errors.OpPrereqError("This commands must be run on the node" |
335 |
" where you want the new master to be."
|
336 |
" %s is already the master" %
|
337 |
old_master) |
338 |
# end checks
|
339 |
|
340 |
rcode = 0
|
341 |
|
342 |
logging.info("setting master to %s, old master: %s", new_master, old_master)
|
343 |
|
344 |
if not rpc.call_node_stop_master(old_master, True): |
345 |
logging.error("could disable the master role on the old master"
|
346 |
" %s, please disable manually", old_master)
|
347 |
|
348 |
ss.SetKey(ss.SS_MASTER_NODE, new_master) |
349 |
|
350 |
cfg = config.ConfigWriter() |
351 |
|
352 |
if not rpc.call_upload_file(cfg.GetNodeList(), |
353 |
ss.KeyToFilename(ss.SS_MASTER_NODE)): |
354 |
logging.error("could not distribute the new simple store master file"
|
355 |
" to the other nodes, please check.")
|
356 |
|
357 |
if not rpc.call_node_start_master(new_master, True): |
358 |
logging.error("could not start the master role on the new master"
|
359 |
" %s, please check", new_master)
|
360 |
rcode = 1
|
361 |
|
362 |
return rcode
|