root / lib / bootstrap.py @ bfbbc223
History | View | Annotate | Download (22.4 kB)
1 | a0c9f010 | Michael Hanselmann | #
|
---|---|---|---|
2 | a0c9f010 | Michael Hanselmann | #
|
3 | a0c9f010 | Michael Hanselmann | |
4 | a0c9f010 | Michael Hanselmann | # Copyright (C) 2006, 2007, 2008 Google Inc.
|
5 | a0c9f010 | Michael Hanselmann | #
|
6 | a0c9f010 | Michael Hanselmann | # This program is free software; you can redistribute it and/or modify
|
7 | a0c9f010 | Michael Hanselmann | # it under the terms of the GNU General Public License as published by
|
8 | a0c9f010 | Michael Hanselmann | # the Free Software Foundation; either version 2 of the License, or
|
9 | a0c9f010 | Michael Hanselmann | # (at your option) any later version.
|
10 | a0c9f010 | Michael Hanselmann | #
|
11 | a0c9f010 | Michael Hanselmann | # This program is distributed in the hope that it will be useful, but
|
12 | a0c9f010 | Michael Hanselmann | # WITHOUT ANY WARRANTY; without even the implied warranty of
|
13 | a0c9f010 | Michael Hanselmann | # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
14 | a0c9f010 | Michael Hanselmann | # General Public License for more details.
|
15 | a0c9f010 | Michael Hanselmann | #
|
16 | a0c9f010 | Michael Hanselmann | # You should have received a copy of the GNU General Public License
|
17 | a0c9f010 | Michael Hanselmann | # along with this program; if not, write to the Free Software
|
18 | a0c9f010 | Michael Hanselmann | # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
|
19 | a0c9f010 | Michael Hanselmann | # 02110-1301, USA.
|
20 | a0c9f010 | Michael Hanselmann | |
21 | a0c9f010 | Michael Hanselmann | |
22 | a0c9f010 | Michael Hanselmann | """Functions to bootstrap a new cluster.
|
23 | a0c9f010 | Michael Hanselmann |
|
24 | a0c9f010 | Michael Hanselmann | """
|
25 | a0c9f010 | Michael Hanselmann | |
26 | a0c9f010 | Michael Hanselmann | import os |
27 | a0c9f010 | Michael Hanselmann | import os.path |
28 | a0c9f010 | Michael Hanselmann | import re |
29 | b1b6ea87 | Iustin Pop | import logging |
30 | c4415fd5 | Michael Hanselmann | import tempfile |
31 | d693c864 | Iustin Pop | import time |
32 | a0c9f010 | Michael Hanselmann | |
33 | a0c9f010 | Michael Hanselmann | from ganeti import rpc |
34 | a0c9f010 | Michael Hanselmann | from ganeti import ssh |
35 | a0c9f010 | Michael Hanselmann | from ganeti import utils |
36 | a0c9f010 | Michael Hanselmann | from ganeti import errors |
37 | a0c9f010 | Michael Hanselmann | from ganeti import config |
38 | a0c9f010 | Michael Hanselmann | from ganeti import constants |
39 | b9eeeb02 | Michael Hanselmann | from ganeti import objects |
40 | a0c9f010 | Michael Hanselmann | from ganeti import ssconf |
41 | a33848a5 | Guido Trotter | from ganeti import serializer |
42 | a5728081 | Guido Trotter | from ganeti import hypervisor |
43 | a0c9f010 | Michael Hanselmann | |
44 | e38220e4 | Michael Hanselmann | |
45 | 531baf8e | Iustin Pop | def _InitSSHSetup(): |
46 | a0c9f010 | Michael Hanselmann | """Setup the SSH configuration for the cluster.
|
47 | a0c9f010 | Michael Hanselmann |
|
48 | a0c9f010 | Michael Hanselmann | This generates a dsa keypair for root, adds the pub key to the
|
49 | a0c9f010 | Michael Hanselmann | permitted hosts and adds the hostkey to its own known hosts.
|
50 | a0c9f010 | Michael Hanselmann |
|
51 | a0c9f010 | Michael Hanselmann | """
|
52 | a0c9f010 | Michael Hanselmann | priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.GANETI_RUNAS) |
53 | a0c9f010 | Michael Hanselmann | |
54 | a0c9f010 | Michael Hanselmann | for name in priv_key, pub_key: |
55 | a0c9f010 | Michael Hanselmann | if os.path.exists(name):
|
56 | a0c9f010 | Michael Hanselmann | utils.CreateBackup(name) |
57 | a0c9f010 | Michael Hanselmann | utils.RemoveFile(name) |
58 | a0c9f010 | Michael Hanselmann | |
59 | a0c9f010 | Michael Hanselmann | result = utils.RunCmd(["ssh-keygen", "-t", "dsa", |
60 | a0c9f010 | Michael Hanselmann | "-f", priv_key,
|
61 | a0c9f010 | Michael Hanselmann | "-q", "-N", ""]) |
62 | a0c9f010 | Michael Hanselmann | if result.failed:
|
63 | a0c9f010 | Michael Hanselmann | raise errors.OpExecError("Could not generate ssh keypair, error %s" % |
64 | a0c9f010 | Michael Hanselmann | result.output) |
65 | a0c9f010 | Michael Hanselmann | |
66 | 7a0156dc | Luca Bigliardi | utils.AddAuthorizedKey(auth_keys, utils.ReadFile(pub_key)) |
67 | a0c9f010 | Michael Hanselmann | |
68 | a0c9f010 | Michael Hanselmann | |
69 | cd34faf2 | Michael Hanselmann | def GenerateSelfSignedSslCert(file_name, validity=(365 * 5)): |
70 | 40a97d80 | Michael Hanselmann | """Generates a self-signed SSL certificate.
|
71 | a0c9f010 | Michael Hanselmann |
|
72 | 40a97d80 | Michael Hanselmann | @type file_name: str
|
73 | 40a97d80 | Michael Hanselmann | @param file_name: Path to output file
|
74 | 40a97d80 | Michael Hanselmann | @type validity: int
|
75 | 40a97d80 | Michael Hanselmann | @param validity: Validity for certificate in days
|
76 | a0c9f010 | Michael Hanselmann |
|
77 | a0c9f010 | Michael Hanselmann | """
|
78 | c4415fd5 | Michael Hanselmann | (fd, tmp_file_name) = tempfile.mkstemp(dir=os.path.dirname(file_name)) |
79 | c4415fd5 | Michael Hanselmann | try:
|
80 | 88828491 | Michael Hanselmann | try:
|
81 | 88828491 | Michael Hanselmann | # Set permissions before writing key
|
82 | 88828491 | Michael Hanselmann | os.chmod(tmp_file_name, 0600)
|
83 | 88828491 | Michael Hanselmann | |
84 | 88828491 | Michael Hanselmann | result = utils.RunCmd(["openssl", "req", "-new", "-newkey", "rsa:1024", |
85 | 88828491 | Michael Hanselmann | "-days", str(validity), "-nodes", "-x509", |
86 | 88828491 | Michael Hanselmann | "-keyout", tmp_file_name, "-out", tmp_file_name, |
87 | 88828491 | Michael Hanselmann | "-batch"])
|
88 | 88828491 | Michael Hanselmann | if result.failed:
|
89 | 88828491 | Michael Hanselmann | raise errors.OpExecError("Could not generate SSL certificate, command" |
90 | 88828491 | Michael Hanselmann | " %s had exitcode %s and error message %s" %
|
91 | 88828491 | Michael Hanselmann | (result.cmd, result.exit_code, result.output)) |
92 | 88828491 | Michael Hanselmann | |
93 | 88828491 | Michael Hanselmann | # Make read-only
|
94 | 88828491 | Michael Hanselmann | os.chmod(tmp_file_name, 0400)
|
95 | 88828491 | Michael Hanselmann | |
96 | 88828491 | Michael Hanselmann | os.rename(tmp_file_name, file_name) |
97 | 88828491 | Michael Hanselmann | finally:
|
98 | 88828491 | Michael Hanselmann | utils.RemoveFile(tmp_file_name) |
99 | c4415fd5 | Michael Hanselmann | finally:
|
100 | 88828491 | Michael Hanselmann | os.close(fd) |
101 | 40a97d80 | Michael Hanselmann | |
102 | 40a97d80 | Michael Hanselmann | |
103 | c008906b | Michael Hanselmann | def GenerateHmacKey(file_name): |
104 | c008906b | Michael Hanselmann | """Writes a new HMAC key.
|
105 | c008906b | Michael Hanselmann |
|
106 | c008906b | Michael Hanselmann | @type file_name: str
|
107 | c008906b | Michael Hanselmann | @param file_name: Path to output file
|
108 | c008906b | Michael Hanselmann |
|
109 | c008906b | Michael Hanselmann | """
|
110 | 43575108 | Michael Hanselmann | utils.WriteFile(file_name, data="%s\n" % utils.GenerateSecret(), mode=0400, |
111 | 43575108 | Michael Hanselmann | backup=True)
|
112 | 43575108 | Michael Hanselmann | |
113 | 43575108 | Michael Hanselmann | |
114 | 6b7d5878 | Michael Hanselmann | def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_confd_hmac_key, |
115 | 43575108 | Michael Hanselmann | rapi_cert_pem=None):
|
116 | 43575108 | Michael Hanselmann | """Updates the cluster certificates, keys and secrets.
|
117 | 43575108 | Michael Hanselmann |
|
118 | 43575108 | Michael Hanselmann | @type new_cluster_cert: bool
|
119 | 43575108 | Michael Hanselmann | @param new_cluster_cert: Whether to generate a new cluster certificate
|
120 | 43575108 | Michael Hanselmann | @type new_rapi_cert: bool
|
121 | 43575108 | Michael Hanselmann | @param new_rapi_cert: Whether to generate a new RAPI certificate
|
122 | 6b7d5878 | Michael Hanselmann | @type new_confd_hmac_key: bool
|
123 | 6b7d5878 | Michael Hanselmann | @param new_confd_hmac_key: Whether to generate a new HMAC key
|
124 | 43575108 | Michael Hanselmann | @type rapi_cert_pem: string
|
125 | 43575108 | Michael Hanselmann | @param rapi_cert_pem: New RAPI certificate in PEM format
|
126 | 43575108 | Michael Hanselmann |
|
127 | 43575108 | Michael Hanselmann | """
|
128 | 168c1de2 | Michael Hanselmann | # noded SSL certificate
|
129 | 168c1de2 | Michael Hanselmann | cluster_cert_exists = os.path.exists(constants.NODED_CERT_FILE) |
130 | 43575108 | Michael Hanselmann | if new_cluster_cert or not cluster_cert_exists: |
131 | 43575108 | Michael Hanselmann | if cluster_cert_exists:
|
132 | 168c1de2 | Michael Hanselmann | utils.CreateBackup(constants.NODED_CERT_FILE) |
133 | 43575108 | Michael Hanselmann | |
134 | 43575108 | Michael Hanselmann | logging.debug("Generating new cluster certificate at %s",
|
135 | 168c1de2 | Michael Hanselmann | constants.NODED_CERT_FILE) |
136 | 168c1de2 | Michael Hanselmann | GenerateSelfSignedSslCert(constants.NODED_CERT_FILE) |
137 | 43575108 | Michael Hanselmann | |
138 | 6b7d5878 | Michael Hanselmann | # confd HMAC key
|
139 | 6b7d5878 | Michael Hanselmann | if new_confd_hmac_key or not os.path.exists(constants.CONFD_HMAC_KEY): |
140 | 6b7d5878 | Michael Hanselmann | logging.debug("Writing new confd HMAC key to %s", constants.CONFD_HMAC_KEY)
|
141 | 6b7d5878 | Michael Hanselmann | GenerateHmacKey(constants.CONFD_HMAC_KEY) |
142 | 43575108 | Michael Hanselmann | |
143 | 43575108 | Michael Hanselmann | # RAPI
|
144 | 43575108 | Michael Hanselmann | rapi_cert_exists = os.path.exists(constants.RAPI_CERT_FILE) |
145 | 43575108 | Michael Hanselmann | |
146 | 43575108 | Michael Hanselmann | if rapi_cert_pem:
|
147 | 43575108 | Michael Hanselmann | # Assume rapi_pem contains a valid PEM-formatted certificate and key
|
148 | 43575108 | Michael Hanselmann | logging.debug("Writing RAPI certificate at %s",
|
149 | 43575108 | Michael Hanselmann | constants.RAPI_CERT_FILE) |
150 | 43575108 | Michael Hanselmann | utils.WriteFile(constants.RAPI_CERT_FILE, data=rapi_cert_pem, backup=True)
|
151 | 43575108 | Michael Hanselmann | |
152 | 43575108 | Michael Hanselmann | elif new_rapi_cert or not rapi_cert_exists: |
153 | 43575108 | Michael Hanselmann | if rapi_cert_exists:
|
154 | 43575108 | Michael Hanselmann | utils.CreateBackup(constants.RAPI_CERT_FILE) |
155 | 43575108 | Michael Hanselmann | |
156 | 43575108 | Michael Hanselmann | logging.debug("Generating new RAPI certificate at %s",
|
157 | 43575108 | Michael Hanselmann | constants.RAPI_CERT_FILE) |
158 | 43575108 | Michael Hanselmann | GenerateSelfSignedSslCert(constants.RAPI_CERT_FILE) |
159 | c008906b | Michael Hanselmann | |
160 | c008906b | Michael Hanselmann | |
161 | 8f215968 | Michael Hanselmann | def _InitGanetiServerSetup(master_name): |
162 | 40a97d80 | Michael Hanselmann | """Setup the necessary configuration for the initial node daemon.
|
163 | 40a97d80 | Michael Hanselmann |
|
164 | 40a97d80 | Michael Hanselmann | This creates the nodepass file containing the shared password for
|
165 | 40a97d80 | Michael Hanselmann | the cluster and also generates the SSL certificate.
|
166 | 40a97d80 | Michael Hanselmann |
|
167 | 40a97d80 | Michael Hanselmann | """
|
168 | 43575108 | Michael Hanselmann | # Generate cluster secrets
|
169 | 43575108 | Michael Hanselmann | GenerateClusterCrypto(True, False, False) |
170 | 4a34c5cf | Guido Trotter | |
171 | f154a7a3 | Michael Hanselmann | result = utils.RunCmd([constants.DAEMON_UTIL, "start", constants.NODED])
|
172 | a0c9f010 | Michael Hanselmann | if result.failed:
|
173 | a0c9f010 | Michael Hanselmann | raise errors.OpExecError("Could not start the node daemon, command %s" |
174 | a0c9f010 | Michael Hanselmann | " had exitcode %s and error %s" %
|
175 | a0c9f010 | Michael Hanselmann | (result.cmd, result.exit_code, result.output)) |
176 | a0c9f010 | Michael Hanselmann | |
177 | 5627f375 | Michael Hanselmann | _WaitForNodeDaemon(master_name) |
178 | 5627f375 | Michael Hanselmann | |
179 | 5627f375 | Michael Hanselmann | |
180 | 5627f375 | Michael Hanselmann | def _WaitForNodeDaemon(node_name): |
181 | 5627f375 | Michael Hanselmann | """Wait for node daemon to become responsive.
|
182 | 5627f375 | Michael Hanselmann |
|
183 | 5627f375 | Michael Hanselmann | """
|
184 | d3833ebd | Michael Hanselmann | def _CheckNodeDaemon(): |
185 | 5627f375 | Michael Hanselmann | result = rpc.RpcRunner.call_version([node_name])[node_name] |
186 | d3833ebd | Michael Hanselmann | if result.fail_msg:
|
187 | d3833ebd | Michael Hanselmann | raise utils.RetryAgain()
|
188 | 8f215968 | Michael Hanselmann | |
189 | d3833ebd | Michael Hanselmann | try:
|
190 | d3833ebd | Michael Hanselmann | utils.Retry(_CheckNodeDaemon, 1.0, 10.0) |
191 | d3833ebd | Michael Hanselmann | except utils.RetryTimeout:
|
192 | 5627f375 | Michael Hanselmann | raise errors.OpExecError("Node daemon on %s didn't answer queries within" |
193 | 5627f375 | Michael Hanselmann | " 10 seconds" % node_name)
|
194 | 5627f375 | Michael Hanselmann | |
195 | a0c9f010 | Michael Hanselmann | |
196 | ec0652ad | Guido Trotter | def InitCluster(cluster_name, mac_prefix, |
197 | ce735215 | Guido Trotter | master_netdev, file_storage_dir, candidate_pool_size, |
198 | b6a30b0d | Guido Trotter | secondary_ip=None, vg_name=None, beparams=None, |
199 | b6a30b0d | Guido Trotter | nicparams=None, hvparams=None, enabled_hypervisors=None, |
200 | b989b9d9 | Ken Wehr | modify_etc_hosts=True, modify_ssh_setup=True): |
201 | a0c9f010 | Michael Hanselmann | """Initialise the cluster.
|
202 | a0c9f010 | Michael Hanselmann |
|
203 | ce735215 | Guido Trotter | @type candidate_pool_size: int
|
204 | ce735215 | Guido Trotter | @param candidate_pool_size: master candidate pool size
|
205 | ce735215 | Guido Trotter |
|
206 | a0c9f010 | Michael Hanselmann | """
|
207 | ce735215 | Guido Trotter | # TODO: complete the docstring
|
208 | a0c9f010 | Michael Hanselmann | if config.ConfigWriter.IsCluster():
|
209 | debac808 | Iustin Pop | raise errors.OpPrereqError("Cluster is already initialised", |
210 | debac808 | Iustin Pop | errors.ECODE_STATE) |
211 | a0c9f010 | Michael Hanselmann | |
212 | b119bccb | Guido Trotter | if not enabled_hypervisors: |
213 | b119bccb | Guido Trotter | raise errors.OpPrereqError("Enabled hypervisors list must contain at" |
214 | debac808 | Iustin Pop | " least one member", errors.ECODE_INVAL)
|
215 | b119bccb | Guido Trotter | invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
|
216 | b119bccb | Guido Trotter | if invalid_hvs:
|
217 | b119bccb | Guido Trotter | raise errors.OpPrereqError("Enabled hypervisors contains invalid" |
218 | debac808 | Iustin Pop | " entries: %s" % invalid_hvs,
|
219 | debac808 | Iustin Pop | errors.ECODE_INVAL) |
220 | b119bccb | Guido Trotter | |
221 | 104f4ca1 | Iustin Pop | hostname = utils.GetHostInfo() |
222 | a0c9f010 | Michael Hanselmann | |
223 | a0c9f010 | Michael Hanselmann | if hostname.ip.startswith("127."): |
224 | a0c9f010 | Michael Hanselmann | raise errors.OpPrereqError("This host's IP resolves to the private" |
225 | a0c9f010 | Michael Hanselmann | " range (%s). Please fix DNS or %s." %
|
226 | debac808 | Iustin Pop | (hostname.ip, constants.ETC_HOSTS), |
227 | debac808 | Iustin Pop | errors.ECODE_ENVIRON) |
228 | a0c9f010 | Michael Hanselmann | |
229 | caad16e2 | Iustin Pop | if not utils.OwnIpAddress(hostname.ip): |
230 | a0c9f010 | Michael Hanselmann | raise errors.OpPrereqError("Inconsistency: this host's name resolves" |
231 | a0c9f010 | Michael Hanselmann | " to %s,\nbut this ip address does not"
|
232 | debac808 | Iustin Pop | " belong to this host. Aborting." %
|
233 | debac808 | Iustin Pop | hostname.ip, errors.ECODE_ENVIRON) |
234 | a0c9f010 | Michael Hanselmann | |
235 | 44caf5a8 | Iustin Pop | clustername = utils.GetHostInfo(utils.HostInfo.NormalizeName(cluster_name)) |
236 | a0c9f010 | Michael Hanselmann | |
237 | a0c9f010 | Michael Hanselmann | if utils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT,
|
238 | a0c9f010 | Michael Hanselmann | timeout=5):
|
239 | debac808 | Iustin Pop | raise errors.OpPrereqError("Cluster IP already active. Aborting.", |
240 | debac808 | Iustin Pop | errors.ECODE_NOTUNIQUE) |
241 | a0c9f010 | Michael Hanselmann | |
242 | a0c9f010 | Michael Hanselmann | if secondary_ip:
|
243 | a0c9f010 | Michael Hanselmann | if not utils.IsValidIP(secondary_ip): |
244 | debac808 | Iustin Pop | raise errors.OpPrereqError("Invalid secondary ip given", |
245 | debac808 | Iustin Pop | errors.ECODE_INVAL) |
246 | a0c9f010 | Michael Hanselmann | if (secondary_ip != hostname.ip and |
247 | caad16e2 | Iustin Pop | not utils.OwnIpAddress(secondary_ip)):
|
248 | a0c9f010 | Michael Hanselmann | raise errors.OpPrereqError("You gave %s as secondary IP," |
249 | a0c9f010 | Michael Hanselmann | " but it does not belong to this host." %
|
250 | debac808 | Iustin Pop | secondary_ip, errors.ECODE_ENVIRON) |
251 | b9eeeb02 | Michael Hanselmann | else:
|
252 | b9eeeb02 | Michael Hanselmann | secondary_ip = hostname.ip |
253 | a0c9f010 | Michael Hanselmann | |
254 | a0c9f010 | Michael Hanselmann | if vg_name is not None: |
255 | a0c9f010 | Michael Hanselmann | # Check if volume group is valid
|
256 | a0c9f010 | Michael Hanselmann | vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name, |
257 | a0c9f010 | Michael Hanselmann | constants.MIN_VG_SIZE) |
258 | a0c9f010 | Michael Hanselmann | if vgstatus:
|
259 | a0c9f010 | Michael Hanselmann | raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if" |
260 | debac808 | Iustin Pop | " you are not using lvm" % vgstatus,
|
261 | debac808 | Iustin Pop | errors.ECODE_INVAL) |
262 | a0c9f010 | Michael Hanselmann | |
263 | a0c9f010 | Michael Hanselmann | file_storage_dir = os.path.normpath(file_storage_dir) |
264 | a0c9f010 | Michael Hanselmann | |
265 | a0c9f010 | Michael Hanselmann | if not os.path.isabs(file_storage_dir): |
266 | a0c9f010 | Michael Hanselmann | raise errors.OpPrereqError("The file storage directory you passed is" |
267 | debac808 | Iustin Pop | " not an absolute path.", errors.ECODE_INVAL)
|
268 | a0c9f010 | Michael Hanselmann | |
269 | a0c9f010 | Michael Hanselmann | if not os.path.exists(file_storage_dir): |
270 | a0c9f010 | Michael Hanselmann | try:
|
271 | a0c9f010 | Michael Hanselmann | os.makedirs(file_storage_dir, 0750)
|
272 | a0c9f010 | Michael Hanselmann | except OSError, err: |
273 | a0c9f010 | Michael Hanselmann | raise errors.OpPrereqError("Cannot create file storage directory" |
274 | debac808 | Iustin Pop | " '%s': %s" % (file_storage_dir, err),
|
275 | debac808 | Iustin Pop | errors.ECODE_ENVIRON) |
276 | a0c9f010 | Michael Hanselmann | |
277 | a0c9f010 | Michael Hanselmann | if not os.path.isdir(file_storage_dir): |
278 | a0c9f010 | Michael Hanselmann | raise errors.OpPrereqError("The file storage directory '%s' is not" |
279 | debac808 | Iustin Pop | " a directory." % file_storage_dir,
|
280 | debac808 | Iustin Pop | errors.ECODE_ENVIRON) |
281 | a0c9f010 | Michael Hanselmann | |
282 | a0c9f010 | Michael Hanselmann | if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix): |
283 | debac808 | Iustin Pop | raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix, |
284 | debac808 | Iustin Pop | errors.ECODE_INVAL) |
285 | a0c9f010 | Michael Hanselmann | |
286 | a0c9f010 | Michael Hanselmann | result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev]) |
287 | a0c9f010 | Michael Hanselmann | if result.failed:
|
288 | a0c9f010 | Michael Hanselmann | raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" % |
289 | a0c9f010 | Michael Hanselmann | (master_netdev, |
290 | debac808 | Iustin Pop | result.output.strip()), errors.ECODE_INVAL) |
291 | a0c9f010 | Michael Hanselmann | |
292 | 9dae41ad | Guido Trotter | dirs = [(constants.RUN_GANETI_DIR, constants.RUN_DIRS_MODE)] |
293 | 9dae41ad | Guido Trotter | utils.EnsureDirs(dirs) |
294 | 9dae41ad | Guido Trotter | |
295 | a5728081 | Guido Trotter | utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES) |
296 | b6a30b0d | Guido Trotter | utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES) |
297 | b6a30b0d | Guido Trotter | objects.NIC.CheckParameterSyntax(nicparams) |
298 | b6a30b0d | Guido Trotter | |
299 | a5728081 | Guido Trotter | # hvparams is a mapping of hypervisor->hvparams dict
|
300 | a5728081 | Guido Trotter | for hv_name, hv_params in hvparams.iteritems(): |
301 | a5728081 | Guido Trotter | utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES) |
302 | a5728081 | Guido Trotter | hv_class = hypervisor.GetHypervisor(hv_name) |
303 | a5728081 | Guido Trotter | hv_class.CheckParameterSyntax(hv_params) |
304 | d4b72030 | Guido Trotter | |
305 | a0c9f010 | Michael Hanselmann | # set up the inter-node password and certificate
|
306 | 8f215968 | Michael Hanselmann | _InitGanetiServerSetup(hostname.name) |
307 | a0c9f010 | Michael Hanselmann | |
308 | a0c9f010 | Michael Hanselmann | # set up ssh config and /etc/hosts
|
309 | 13998ef2 | Michael Hanselmann | sshline = utils.ReadFile(constants.SSH_HOST_RSA_PUB) |
310 | a0c9f010 | Michael Hanselmann | sshkey = sshline.split(" ")[1] |
311 | a0c9f010 | Michael Hanselmann | |
312 | b86a6bcd | Guido Trotter | if modify_etc_hosts:
|
313 | b86a6bcd | Guido Trotter | utils.AddHostToEtcHosts(hostname.name) |
314 | b86a6bcd | Guido Trotter | |
315 | b989b9d9 | Ken Wehr | if modify_ssh_setup:
|
316 | b989b9d9 | Ken Wehr | _InitSSHSetup() |
317 | a0c9f010 | Michael Hanselmann | |
318 | 430b923c | Iustin Pop | now = time.time() |
319 | 430b923c | Iustin Pop | |
320 | a0c9f010 | Michael Hanselmann | # init of cluster config file
|
321 | b9eeeb02 | Michael Hanselmann | cluster_config = objects.Cluster( |
322 | b9eeeb02 | Michael Hanselmann | serial_no=1,
|
323 | b9eeeb02 | Michael Hanselmann | rsahostkeypub=sshkey, |
324 | b9eeeb02 | Michael Hanselmann | highest_used_port=(constants.FIRST_DRBD_PORT - 1),
|
325 | b9eeeb02 | Michael Hanselmann | mac_prefix=mac_prefix, |
326 | b9eeeb02 | Michael Hanselmann | volume_group_name=vg_name, |
327 | b9eeeb02 | Michael Hanselmann | tcpudp_port_pool=set(),
|
328 | f6bd6e98 | Michael Hanselmann | master_node=hostname.name, |
329 | f6bd6e98 | Michael Hanselmann | master_ip=clustername.ip, |
330 | f6bd6e98 | Michael Hanselmann | master_netdev=master_netdev, |
331 | f6bd6e98 | Michael Hanselmann | cluster_name=clustername.name, |
332 | f6bd6e98 | Michael Hanselmann | file_storage_dir=file_storage_dir, |
333 | ea3a925f | Alexander Schreiber | enabled_hypervisors=enabled_hypervisors, |
334 | 4ef7f423 | Guido Trotter | beparams={constants.PP_DEFAULT: beparams}, |
335 | b6a30b0d | Guido Trotter | nicparams={constants.PP_DEFAULT: nicparams}, |
336 | ea3a925f | Alexander Schreiber | hvparams=hvparams, |
337 | ce735215 | Guido Trotter | candidate_pool_size=candidate_pool_size, |
338 | 022c3a0b | Guido Trotter | modify_etc_hosts=modify_etc_hosts, |
339 | b989b9d9 | Ken Wehr | modify_ssh_setup=modify_ssh_setup, |
340 | 430b923c | Iustin Pop | ctime=now, |
341 | 430b923c | Iustin Pop | mtime=now, |
342 | 430b923c | Iustin Pop | uuid=utils.NewUUID(), |
343 | b9eeeb02 | Michael Hanselmann | ) |
344 | b9eeeb02 | Michael Hanselmann | master_node_config = objects.Node(name=hostname.name, |
345 | b9eeeb02 | Michael Hanselmann | primary_ip=hostname.ip, |
346 | b9222f32 | Guido Trotter | secondary_ip=secondary_ip, |
347 | c044f32c | Guido Trotter | serial_no=1,
|
348 | c044f32c | Guido Trotter | master_candidate=True,
|
349 | af64c0ea | Iustin Pop | offline=False, drained=False, |
350 | c044f32c | Guido Trotter | ) |
351 | 9e1333b9 | Guido Trotter | InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config) |
352 | 05cc153f | Guido Trotter | cfg = config.ConfigWriter() |
353 | 9e1333b9 | Guido Trotter | ssh.WriteKnownHostsFile(cfg, constants.SSH_KNOWN_HOSTS_FILE) |
354 | a4eae71f | Michael Hanselmann | cfg.Update(cfg.GetClusterInfo(), logging.error) |
355 | 827f753e | Guido Trotter | |
356 | b3f1cf6f | Iustin Pop | # start the master ip
|
357 | b3f1cf6f | Iustin Pop | # TODO: Review rpc call from bootstrap
|
358 | b726aff0 | Iustin Pop | # TODO: Warn on failed start master
|
359 | 3583908a | Guido Trotter | rpc.RpcRunner.call_node_start_master(hostname.name, True, False) |
360 | b3f1cf6f | Iustin Pop | |
361 | b1b6ea87 | Iustin Pop | |
362 | 02f99608 | Oleksiy Mishchenko | def InitConfig(version, cluster_config, master_node_config, |
363 | 02f99608 | Oleksiy Mishchenko | cfg_file=constants.CLUSTER_CONF_FILE): |
364 | 7b3a8fb5 | Iustin Pop | """Create the initial cluster configuration.
|
365 | 7b3a8fb5 | Iustin Pop |
|
366 | 7b3a8fb5 | Iustin Pop | It will contain the current node, which will also be the master
|
367 | 7b3a8fb5 | Iustin Pop | node, and no instances.
|
368 | 7b3a8fb5 | Iustin Pop |
|
369 | 7b3a8fb5 | Iustin Pop | @type version: int
|
370 | c41eea6e | Iustin Pop | @param version: configuration version
|
371 | c41eea6e | Iustin Pop | @type cluster_config: L{objects.Cluster}
|
372 | c41eea6e | Iustin Pop | @param cluster_config: cluster configuration
|
373 | c41eea6e | Iustin Pop | @type master_node_config: L{objects.Node}
|
374 | c41eea6e | Iustin Pop | @param master_node_config: master node configuration
|
375 | c41eea6e | Iustin Pop | @type cfg_file: string
|
376 | c41eea6e | Iustin Pop | @param cfg_file: configuration file path
|
377 | c41eea6e | Iustin Pop |
|
378 | 7b3a8fb5 | Iustin Pop | """
|
379 | 7b3a8fb5 | Iustin Pop | nodes = { |
380 | 7b3a8fb5 | Iustin Pop | master_node_config.name: master_node_config, |
381 | 7b3a8fb5 | Iustin Pop | } |
382 | 7b3a8fb5 | Iustin Pop | |
383 | d693c864 | Iustin Pop | now = time.time() |
384 | 7b3a8fb5 | Iustin Pop | config_data = objects.ConfigData(version=version, |
385 | 7b3a8fb5 | Iustin Pop | cluster=cluster_config, |
386 | 7b3a8fb5 | Iustin Pop | nodes=nodes, |
387 | 7b3a8fb5 | Iustin Pop | instances={}, |
388 | d693c864 | Iustin Pop | serial_no=1,
|
389 | d693c864 | Iustin Pop | ctime=now, mtime=now) |
390 | a33848a5 | Guido Trotter | utils.WriteFile(cfg_file, |
391 | a33848a5 | Guido Trotter | data=serializer.Dump(config_data.ToDict()), |
392 | a33848a5 | Guido Trotter | mode=0600)
|
393 | 02f99608 | Oleksiy Mishchenko | |
394 | 02f99608 | Oleksiy Mishchenko | |
395 | 140aa4a8 | Iustin Pop | def FinalizeClusterDestroy(master): |
396 | 140aa4a8 | Iustin Pop | """Execute the last steps of cluster destroy
|
397 | 140aa4a8 | Iustin Pop |
|
398 | 140aa4a8 | Iustin Pop | This function shuts down all the daemons, completing the destroy
|
399 | 140aa4a8 | Iustin Pop | begun in cmdlib.LUDestroyOpcode.
|
400 | 140aa4a8 | Iustin Pop |
|
401 | 140aa4a8 | Iustin Pop | """
|
402 | b989b9d9 | Ken Wehr | cfg = config.ConfigWriter() |
403 | b989b9d9 | Ken Wehr | modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup |
404 | 781de953 | Iustin Pop | result = rpc.RpcRunner.call_node_stop_master(master, True)
|
405 | 3cebe102 | Michael Hanselmann | msg = result.fail_msg |
406 | 6c00d19a | Iustin Pop | if msg:
|
407 | 099c52ad | Iustin Pop | logging.warning("Could not disable the master role: %s", msg)
|
408 | b989b9d9 | Ken Wehr | result = rpc.RpcRunner.call_node_leave_cluster(master, modify_ssh_setup) |
409 | 3cebe102 | Michael Hanselmann | msg = result.fail_msg |
410 | 0623d351 | Iustin Pop | if msg:
|
411 | 0623d351 | Iustin Pop | logging.warning("Could not shutdown the node daemon and cleanup"
|
412 | 0623d351 | Iustin Pop | " the node: %s", msg)
|
413 | 140aa4a8 | Iustin Pop | |
414 | 140aa4a8 | Iustin Pop | |
415 | 87622829 | Iustin Pop | def SetupNodeDaemon(cluster_name, node, ssh_key_check): |
416 | 827f753e | Guido Trotter | """Add a node to the cluster.
|
417 | 827f753e | Guido Trotter |
|
418 | b1b6ea87 | Iustin Pop | This function must be called before the actual opcode, and will ssh
|
419 | b1b6ea87 | Iustin Pop | to the remote node, copy the needed files, and start ganeti-noded,
|
420 | b1b6ea87 | Iustin Pop | allowing the master to do the rest via normal rpc calls.
|
421 | 827f753e | Guido Trotter |
|
422 | 87622829 | Iustin Pop | @param cluster_name: the cluster name
|
423 | 87622829 | Iustin Pop | @param node: the name of the new node
|
424 | 87622829 | Iustin Pop | @param ssh_key_check: whether to do a strict key check
|
425 | 827f753e | Guido Trotter |
|
426 | 827f753e | Guido Trotter | """
|
427 | 87622829 | Iustin Pop | sshrunner = ssh.SshRunner(cluster_name) |
428 | 5557b04c | Michael Hanselmann | |
429 | 168c1de2 | Michael Hanselmann | noded_cert = utils.ReadFile(constants.NODED_CERT_FILE) |
430 | 2438c157 | Michael Hanselmann | rapi_cert = utils.ReadFile(constants.RAPI_CERT_FILE) |
431 | 6b7d5878 | Michael Hanselmann | confd_hmac_key = utils.ReadFile(constants.CONFD_HMAC_KEY) |
432 | 5557b04c | Michael Hanselmann | |
433 | 827f753e | Guido Trotter | # in the base64 pem encoding, neither '!' nor '.' are valid chars,
|
434 | 827f753e | Guido Trotter | # so we use this to detect an invalid certificate; as long as the
|
435 | 827f753e | Guido Trotter | # cert doesn't contain this, the here-document will be correctly
|
436 | 77b076ca | Guido Trotter | # parsed by the shell sequence below. HMAC keys are hexadecimal strings,
|
437 | 77b076ca | Guido Trotter | # so the same restrictions apply.
|
438 | 6b7d5878 | Michael Hanselmann | for content in (noded_cert, rapi_cert, confd_hmac_key): |
439 | 77b076ca | Guido Trotter | if re.search('^!EOF\.', content, re.MULTILINE): |
440 | 77b076ca | Guido Trotter | raise errors.OpExecError("invalid SSL certificate or HMAC key") |
441 | 5557b04c | Michael Hanselmann | |
442 | 5557b04c | Michael Hanselmann | if not noded_cert.endswith("\n"): |
443 | 5557b04c | Michael Hanselmann | noded_cert += "\n"
|
444 | 2438c157 | Michael Hanselmann | if not rapi_cert.endswith("\n"): |
445 | 2438c157 | Michael Hanselmann | rapi_cert += "\n"
|
446 | 6b7d5878 | Michael Hanselmann | if not confd_hmac_key.endswith("\n"): |
447 | 6b7d5878 | Michael Hanselmann | confd_hmac_key += "\n"
|
448 | 827f753e | Guido Trotter | |
449 | 827f753e | Guido Trotter | # set up inter-node password and certificate and restarts the node daemon
|
450 | 827f753e | Guido Trotter | # and then connect with ssh to set password and start ganeti-noded
|
451 | 827f753e | Guido Trotter | # note that all the below variables are sanitized at this point,
|
452 | 827f753e | Guido Trotter | # either by being constants or by the checks above
|
453 | 827f753e | Guido Trotter | mycommand = ("umask 077 && "
|
454 | 827f753e | Guido Trotter | "cat > '%s' << '!EOF.' && \n"
|
455 | 2438c157 | Michael Hanselmann | "%s!EOF.\n"
|
456 | 2438c157 | Michael Hanselmann | "cat > '%s' << '!EOF.' && \n"
|
457 | 2438c157 | Michael Hanselmann | "%s!EOF.\n"
|
458 | 77b076ca | Guido Trotter | "cat > '%s' << '!EOF.' && \n"
|
459 | 77b076ca | Guido Trotter | "%s!EOF.\n"
|
460 | 77b076ca | Guido Trotter | "chmod 0400 %s %s %s && "
|
461 | f154a7a3 | Michael Hanselmann | "%s start %s" %
|
462 | 168c1de2 | Michael Hanselmann | (constants.NODED_CERT_FILE, noded_cert, |
463 | 2438c157 | Michael Hanselmann | constants.RAPI_CERT_FILE, rapi_cert, |
464 | 6b7d5878 | Michael Hanselmann | constants.CONFD_HMAC_KEY, confd_hmac_key, |
465 | 168c1de2 | Michael Hanselmann | constants.NODED_CERT_FILE, constants.RAPI_CERT_FILE, |
466 | 6b7d5878 | Michael Hanselmann | constants.CONFD_HMAC_KEY, |
467 | f154a7a3 | Michael Hanselmann | constants.DAEMON_UTIL, constants.NODED)) |
468 | 827f753e | Guido Trotter | |
469 | c4b6c29c | Michael Hanselmann | result = sshrunner.Run(node, 'root', mycommand, batch=False, |
470 | c4b6c29c | Michael Hanselmann | ask_key=ssh_key_check, |
471 | c4b6c29c | Michael Hanselmann | use_cluster_key=False,
|
472 | c4b6c29c | Michael Hanselmann | strict_host_check=ssh_key_check) |
473 | 827f753e | Guido Trotter | if result.failed:
|
474 | 827f753e | Guido Trotter | raise errors.OpExecError("Remote command on node %s, error: %s," |
475 | 827f753e | Guido Trotter | " output: %s" %
|
476 | 827f753e | Guido Trotter | (node, result.fail_reason, result.output)) |
477 | 827f753e | Guido Trotter | |
478 | 5627f375 | Michael Hanselmann | _WaitForNodeDaemon(node) |
479 | 5627f375 | Michael Hanselmann | |
480 | b1b6ea87 | Iustin Pop | |
481 | 8e2524c3 | Guido Trotter | def MasterFailover(no_voting=False): |
482 | b1b6ea87 | Iustin Pop | """Failover the master node.
|
483 | b1b6ea87 | Iustin Pop |
|
484 | b1b6ea87 | Iustin Pop | This checks that we are not already the master, and will cause the
|
485 | b1b6ea87 | Iustin Pop | current master to cease being master, and the non-master to become
|
486 | b1b6ea87 | Iustin Pop | new master.
|
487 | b1b6ea87 | Iustin Pop |
|
488 | 8e2524c3 | Guido Trotter | @type no_voting: boolean
|
489 | 8e2524c3 | Guido Trotter | @param no_voting: force the operation without remote nodes agreement
|
490 | 8e2524c3 | Guido Trotter | (dangerous)
|
491 | 8e2524c3 | Guido Trotter |
|
492 | b1b6ea87 | Iustin Pop | """
|
493 | 8135a2db | Iustin Pop | sstore = ssconf.SimpleStore() |
494 | b1b6ea87 | Iustin Pop | |
495 | 8135a2db | Iustin Pop | old_master, new_master = ssconf.GetMasterAndMyself(sstore) |
496 | 8135a2db | Iustin Pop | node_list = sstore.GetNodeList() |
497 | 8135a2db | Iustin Pop | mc_list = sstore.GetMasterCandidates() |
498 | b1b6ea87 | Iustin Pop | |
499 | b1b6ea87 | Iustin Pop | if old_master == new_master:
|
500 | b1b6ea87 | Iustin Pop | raise errors.OpPrereqError("This commands must be run on the node" |
501 | b1b6ea87 | Iustin Pop | " where you want the new master to be."
|
502 | b1b6ea87 | Iustin Pop | " %s is already the master" %
|
503 | debac808 | Iustin Pop | old_master, errors.ECODE_INVAL) |
504 | d5927e48 | Iustin Pop | |
505 | 8135a2db | Iustin Pop | if new_master not in mc_list: |
506 | 8135a2db | Iustin Pop | mc_no_master = [name for name in mc_list if name != old_master] |
507 | 8135a2db | Iustin Pop | raise errors.OpPrereqError("This node is not among the nodes marked" |
508 | 8135a2db | Iustin Pop | " as master candidates. Only these nodes"
|
509 | 8135a2db | Iustin Pop | " can become masters. Current list of"
|
510 | 8135a2db | Iustin Pop | " master candidates is:\n"
|
511 | debac808 | Iustin Pop | "%s" % ('\n'.join(mc_no_master)), |
512 | debac808 | Iustin Pop | errors.ECODE_STATE) |
513 | 8135a2db | Iustin Pop | |
514 | 8e2524c3 | Guido Trotter | if not no_voting: |
515 | 8e2524c3 | Guido Trotter | vote_list = GatherMasterVotes(node_list) |
516 | 8e2524c3 | Guido Trotter | |
517 | 8e2524c3 | Guido Trotter | if vote_list:
|
518 | 8e2524c3 | Guido Trotter | voted_master = vote_list[0][0] |
519 | 8e2524c3 | Guido Trotter | if voted_master is None: |
520 | 8e2524c3 | Guido Trotter | raise errors.OpPrereqError("Cluster is inconsistent, most nodes did" |
521 | debac808 | Iustin Pop | " not respond.", errors.ECODE_ENVIRON)
|
522 | 8e2524c3 | Guido Trotter | elif voted_master != old_master:
|
523 | 8e2524c3 | Guido Trotter | raise errors.OpPrereqError("I have a wrong configuration, I believe" |
524 | 8e2524c3 | Guido Trotter | " the master is %s but the other nodes"
|
525 | 8e2524c3 | Guido Trotter | " voted %s. Please resync the configuration"
|
526 | 8e2524c3 | Guido Trotter | " of this node." %
|
527 | debac808 | Iustin Pop | (old_master, voted_master), |
528 | debac808 | Iustin Pop | errors.ECODE_STATE) |
529 | b1b6ea87 | Iustin Pop | # end checks
|
530 | b1b6ea87 | Iustin Pop | |
531 | b1b6ea87 | Iustin Pop | rcode = 0
|
532 | b1b6ea87 | Iustin Pop | |
533 | d5927e48 | Iustin Pop | logging.info("Setting master to %s, old master: %s", new_master, old_master)
|
534 | b1b6ea87 | Iustin Pop | |
535 | 781de953 | Iustin Pop | result = rpc.RpcRunner.call_node_stop_master(old_master, True)
|
536 | 3cebe102 | Michael Hanselmann | msg = result.fail_msg |
537 | 6c00d19a | Iustin Pop | if msg:
|
538 | d5927e48 | Iustin Pop | logging.error("Could not disable the master role on the old master"
|
539 | 6c00d19a | Iustin Pop | " %s, please disable manually: %s", old_master, msg)
|
540 | b1b6ea87 | Iustin Pop | |
541 | d23ef431 | Michael Hanselmann | # Here we have a phase where no master should be running
|
542 | b1b6ea87 | Iustin Pop | |
543 | bbe19c17 | Iustin Pop | # instantiate a real config writer, as we now know we have the
|
544 | bbe19c17 | Iustin Pop | # configuration data
|
545 | bbe19c17 | Iustin Pop | cfg = config.ConfigWriter() |
546 | b1b6ea87 | Iustin Pop | |
547 | bbe19c17 | Iustin Pop | cluster_info = cfg.GetClusterInfo() |
548 | bbe19c17 | Iustin Pop | cluster_info.master_node = new_master |
549 | bbe19c17 | Iustin Pop | # this will also regenerate the ssconf files, since we updated the
|
550 | bbe19c17 | Iustin Pop | # cluster info
|
551 | a4eae71f | Michael Hanselmann | cfg.Update(cluster_info, logging.error) |
552 | d5927e48 | Iustin Pop | |
553 | 3583908a | Guido Trotter | result = rpc.RpcRunner.call_node_start_master(new_master, True, no_voting)
|
554 | 3cebe102 | Michael Hanselmann | msg = result.fail_msg |
555 | b726aff0 | Iustin Pop | if msg:
|
556 | d5927e48 | Iustin Pop | logging.error("Could not start the master role on the new master"
|
557 | b726aff0 | Iustin Pop | " %s, please check: %s", new_master, msg)
|
558 | b1b6ea87 | Iustin Pop | rcode = 1
|
559 | b1b6ea87 | Iustin Pop | |
560 | b1b6ea87 | Iustin Pop | return rcode
|
561 | d7cdb55d | Iustin Pop | |
562 | d7cdb55d | Iustin Pop | |
563 | 8eb148ae | Iustin Pop | def GetMaster(): |
564 | 8eb148ae | Iustin Pop | """Returns the current master node.
|
565 | 8eb148ae | Iustin Pop |
|
566 | 8eb148ae | Iustin Pop | This is a separate function in bootstrap since it's needed by
|
567 | 8eb148ae | Iustin Pop | gnt-cluster, and instead of importing directly ssconf, it's better
|
568 | 8eb148ae | Iustin Pop | to abstract it in bootstrap, where we do use ssconf in other
|
569 | 8eb148ae | Iustin Pop | functions too.
|
570 | 8eb148ae | Iustin Pop |
|
571 | 8eb148ae | Iustin Pop | """
|
572 | 8eb148ae | Iustin Pop | sstore = ssconf.SimpleStore() |
573 | 8eb148ae | Iustin Pop | |
574 | 8eb148ae | Iustin Pop | old_master, _ = ssconf.GetMasterAndMyself(sstore) |
575 | 8eb148ae | Iustin Pop | |
576 | 8eb148ae | Iustin Pop | return old_master
|
577 | 8eb148ae | Iustin Pop | |
578 | 8eb148ae | Iustin Pop | |
579 | d7cdb55d | Iustin Pop | def GatherMasterVotes(node_list): |
580 | d7cdb55d | Iustin Pop | """Check the agreement on who is the master.
|
581 | d7cdb55d | Iustin Pop |
|
582 | d7cdb55d | Iustin Pop | This function will return a list of (node, number of votes), ordered
|
583 | d7cdb55d | Iustin Pop | by the number of votes. Errors will be denoted by the key 'None'.
|
584 | d7cdb55d | Iustin Pop |
|
585 | d7cdb55d | Iustin Pop | Note that the sum of votes is the number of nodes this machine
|
586 | d7cdb55d | Iustin Pop | knows, whereas the number of entries in the list could be different
|
587 | d7cdb55d | Iustin Pop | (if some nodes vote for another master).
|
588 | d7cdb55d | Iustin Pop |
|
589 | d7cdb55d | Iustin Pop | We remove ourselves from the list since we know that (bugs aside)
|
590 | d7cdb55d | Iustin Pop | since we use the same source for configuration information for both
|
591 | d7cdb55d | Iustin Pop | backend and boostrap, we'll always vote for ourselves.
|
592 | d7cdb55d | Iustin Pop |
|
593 | d7cdb55d | Iustin Pop | @type node_list: list
|
594 | d7cdb55d | Iustin Pop | @param node_list: the list of nodes to query for master info; the current
|
595 | 5bbd3f7f | Michael Hanselmann | node will be removed if it is in the list
|
596 | d7cdb55d | Iustin Pop | @rtype: list
|
597 | d7cdb55d | Iustin Pop | @return: list of (node, votes)
|
598 | d7cdb55d | Iustin Pop |
|
599 | d7cdb55d | Iustin Pop | """
|
600 | d7cdb55d | Iustin Pop | myself = utils.HostInfo().name |
601 | d7cdb55d | Iustin Pop | try:
|
602 | d7cdb55d | Iustin Pop | node_list.remove(myself) |
603 | d7cdb55d | Iustin Pop | except ValueError: |
604 | d7cdb55d | Iustin Pop | pass
|
605 | d7cdb55d | Iustin Pop | if not node_list: |
606 | d7cdb55d | Iustin Pop | # no nodes left (eventually after removing myself)
|
607 | d7cdb55d | Iustin Pop | return []
|
608 | d7cdb55d | Iustin Pop | results = rpc.RpcRunner.call_master_info(node_list) |
609 | d7cdb55d | Iustin Pop | if not isinstance(results, dict): |
610 | d7cdb55d | Iustin Pop | # this should not happen (unless internal error in rpc)
|
611 | d7cdb55d | Iustin Pop | logging.critical("Can't complete rpc call, aborting master startup")
|
612 | d7cdb55d | Iustin Pop | return [(None, len(node_list))] |
613 | d7cdb55d | Iustin Pop | votes = {} |
614 | d7cdb55d | Iustin Pop | for node in results: |
615 | 781de953 | Iustin Pop | nres = results[node] |
616 | 2a52a064 | Iustin Pop | data = nres.payload |
617 | 3cebe102 | Michael Hanselmann | msg = nres.fail_msg |
618 | 2a52a064 | Iustin Pop | fail = False
|
619 | 2a52a064 | Iustin Pop | if msg:
|
620 | 2a52a064 | Iustin Pop | logging.warning("Error contacting node %s: %s", node, msg)
|
621 | 2a52a064 | Iustin Pop | fail = True
|
622 | 2a52a064 | Iustin Pop | elif not isinstance(data, (tuple, list)) or len(data) < 3: |
623 | 2a52a064 | Iustin Pop | logging.warning("Invalid data received from node %s: %s", node, data)
|
624 | 2a52a064 | Iustin Pop | fail = True
|
625 | 2a52a064 | Iustin Pop | if fail:
|
626 | d7cdb55d | Iustin Pop | if None not in votes: |
627 | d7cdb55d | Iustin Pop | votes[None] = 0 |
628 | d7cdb55d | Iustin Pop | votes[None] += 1 |
629 | d7cdb55d | Iustin Pop | continue
|
630 | 781de953 | Iustin Pop | master_node = data[2]
|
631 | d7cdb55d | Iustin Pop | if master_node not in votes: |
632 | d7cdb55d | Iustin Pop | votes[master_node] = 0
|
633 | d7cdb55d | Iustin Pop | votes[master_node] += 1
|
634 | d7cdb55d | Iustin Pop | |
635 | d7cdb55d | Iustin Pop | vote_list = [v for v in votes.items()] |
636 | d7cdb55d | Iustin Pop | # sort first on number of votes then on name, since we want None
|
637 | d7cdb55d | Iustin Pop | # sorted later if we have the half of the nodes not responding, and
|
638 | d7cdb55d | Iustin Pop | # half voting all for the same master
|
639 | d7cdb55d | Iustin Pop | vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True) |
640 | d7cdb55d | Iustin Pop | |
641 | d7cdb55d | Iustin Pop | return vote_list |