system-config/launch/launch-node.py
Monty Taylor 453051aafb Don't pass OS_CLOUD and OS_REGION_NAME to expand-groups
There is a bug in OCC that causes an envvars cloud to be created when
the only two env vars are the selectors OS_CLOUD and OS_REGION_NAME. So
exclude them from the envionment when running the group creation
command.

Also, there is a bug in the invocation of the hostname playbook, in that
it was passing in the UUID as the target to run against, but we're
writing out a name-based inventory.

Change-Id: I0b524dc43ec96c6645ae82a090744eab463e7fb9
2016-03-22 19:33:42 -05:00

272 lines
9.2 KiB
Python
Executable File

#!/usr/bin/env python
# Launch a new OpenStack project infrastructure node.
# Copyright (C) 2011-2012 OpenStack LLC.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
# implied.
#
# See the License for the specific language governing permissions and
# limitations under the License.
import argparse
import os
import subprocess
import sys
import tempfile
import time
import traceback
import dns
import utils
import os_client_config
import paramiko
import shade
SCRIPT_DIR = os.path.dirname(sys.argv[0])
try:
# This unactionable warning does not need to be printed over and over.
import requests.packages.urllib3
requests.packages.urllib3.disable_warnings()
except:
pass
def bootstrap_server(server, key, name, volume, keep):
ip = server.public_v4
ssh_kwargs = dict(pkey=key)
print 'Public IP', ip
for username in ['root', 'ubuntu', 'centos', 'admin']:
ssh_client = utils.ssh_connect(ip, username, ssh_kwargs, timeout=600)
if ssh_client:
break
if not ssh_client:
raise Exception("Unable to log in via SSH")
# cloud-init puts the "please log in as user foo" message and
# subsequent exit() in root's authorized_keys -- overwrite it with
# a normal version to get root login working again.
if username != 'root':
ssh_client.ssh("sudo cp ~/.ssh/authorized_keys"
" ~root/.ssh/authorized_keys")
ssh_client.ssh("sudo chmod 644 ~root/.ssh/authorized_keys")
ssh_client.ssh("sudo chown root.root ~root/.ssh/authorized_keys")
ssh_client = utils.ssh_connect(ip, 'root', ssh_kwargs, timeout=600)
if server.public_v6:
ssh_client.ssh('ping6 -c5 -Q 0x10 review.openstack.org '
'|| ping6 -c5 -Q 0x10 wiki.openstack.org')
ssh_client.scp(os.path.join(SCRIPT_DIR, '..', 'make_swap.sh'),
'make_swap.sh')
ssh_client.ssh('bash -x make_swap.sh')
if volume:
ssh_client.scp(os.path.join(SCRIPT_DIR, '..', 'mount_volume.sh'),
'mount_volume.sh')
ssh_client.ssh('bash -x mount_volume.sh')
# This next chunk should really exist as a playbook, but whatev
ssh_client.scp(os.path.join(SCRIPT_DIR, '..', 'install_puppet.sh'),
'install_puppet.sh')
ssh_client.ssh('bash -x install_puppet.sh')
# Write out the private SSH key we generated
key_file = tempfile.NamedTemporaryFile(delete=not keep)
key.write_private_key(key_file)
key_file.flush()
# Write out inventory
inventory_file = tempfile.NamedTemporaryFile(delete=not keep)
inventory_file.write("{host} ansible_host={ip} ansible_user=root".format(
host=name, ip=server.interface_ip))
inventory_file.flush()
ansible_cmd = [
'ansible-playbook',
'-i', inventory_file.name, '-l', name,
'--private-key={key}'.format(key=key_file.name),
"--ssh-common-args='-o StrictHostKeyChecking=no'",
'-e', 'target={name}'.format(name=name),
]
# Run the remote puppet apply playbook limited to just this server
# we just created
try:
for playbook in [
'set_hostnames.yml',
'remote_puppet_adhoc.yaml']:
print subprocess.check_output(
ansible_cmd + [
os.path.join(
SCRIPT_DIR, '..', 'playbooks', playbook)],
stderr=subprocess.STDOUT)
except subprocess.CalledProcessError as e:
print "Subprocess failed"
print e.output
raise
try:
ssh_client.ssh("reboot")
except Exception as e:
# Some init system kill the connection too fast after reboot.
# Deal with it by ignoring ssh errors when rebooting.
if e.rc == -1:
pass
else:
raise
def build_server(cloud, name, image, flavor,
volume, keep, network, boot_from_volume, config_drive):
key = None
server = None
create_kwargs = dict(image=image, flavor=flavor, name=name,
reuse_ips=False, wait=True,
boot_from_volume=boot_from_volume,
network=network,
config_drive=config_drive)
if volume:
create_kwargs['volumes'] = [volume]
key_name = 'launch-%i' % (time.time())
key = paramiko.RSAKey.generate(2048)
public_key = key.get_name() + ' ' + key.get_base64()
cloud.create_keypair(key_name, public_key)
create_kwargs['key_name'] = key_name
try:
server = cloud.create_server(**create_kwargs)
except Exception:
try:
cloud.delete_keypair(key_name)
except Exception:
print "Exception encountered deleting keypair:"
traceback.print_exc()
raise
try:
cloud.delete_keypair(key_name)
server = cloud.get_openstack_vars(server)
bootstrap_server(server, key, name, volume, keep)
print('UUID=%s\nIPV4=%s\nIPV6=%s\n' % (
server.id, server.public_v4, server.public_v6))
except Exception:
try:
if keep:
print "Server failed to build, keeping as requested."
else:
cloud.delete_server(server.id, delete_ips=True)
except Exception:
print "Exception encountered deleting server:"
traceback.print_exc()
# Raise the important exception that started this
raise
return server
def main():
parser = argparse.ArgumentParser()
parser.add_argument("name", help="server name")
parser.add_argument("--cloud", dest="cloud", required=True,
help="cloud name")
parser.add_argument("--region", dest="region",
help="cloud region")
parser.add_argument("--flavor", dest="flavor", default='1GB',
help="name (or substring) of flavor")
parser.add_argument("--image", dest="image",
default="Ubuntu 14.04 LTS (Trusty Tahr) (PVHVM)",
help="image name")
parser.add_argument("--volume", dest="volume",
help="UUID of volume to attach to the new server.",
default=None)
parser.add_argument("--boot-from-volume", dest="boot_from_volume",
help="Create a boot volume for the server and use it.",
action='store_true',
default=False)
parser.add_argument("--keep", dest="keep",
help="Don't clean up or delete the server on error.",
action='store_true',
default=False)
parser.add_argument("--verbose", dest="verbose", default=False,
action='store_true',
help="Be verbose about logging cloud actions")
parser.add_argument("--network", dest="network", default=None,
help="network label to attach instance to")
parser.add_argument("--config-drive", dest="config_drive",
help="Boot with config_drive attached.",
action='store_true',
default=True)
options = parser.parse_args()
shade.simple_logging(debug=options.verbose)
cloud_kwargs = {}
if options.region:
cloud_kwargs['region_name'] = options.region
cloud_config = os_client_config.OpenStackConfig().get_one_cloud(
options.cloud, **cloud_kwargs)
cloud = shade.OpenStackCloud(cloud_config)
flavor = cloud.get_flavor(options.flavor)
if flavor:
print "Found flavor", flavor.name
else:
print "Unable to find matching flavor; flavor list:"
for i in cloud.list_flavors():
print i.name
sys.exit(1)
image = cloud.get_image_exclude(options.image, 'deprecated')
if image:
print "Found image", image.name
else:
print "Unable to find matching image; image list:"
for i in cloud.list_images():
print i.name
sys.exit(1)
server = build_server(cloud, options.name, image, flavor,
options.volume, options.keep,
options.network, options.boot_from_volume,
options.config_drive)
dns.print_dns(cloud, server)
# Zero the ansible inventory cache so that next run finds the new server
inventory_cache = '/var/cache/ansible-inventory/ansible-inventory.cache'
if os.path.exists(inventory_cache):
with open(inventory_cache, 'w'):
pass
# Remove cloud and region from the environment to work around a bug in occ
expand_env = os.environ.copy()
expand_env.pop('OS_CLOUD', None)
expand_env.pop('OS_REGION_NAME', None)
print subprocess.check_output(
'/usr/local/bin/expand-groups.sh',
env=expand_env,
stderr=subprocess.STDOUT)
if __name__ == '__main__':
main()