def create_private():
data = request.get_json()
pod, _, version = _get_instances(**data)
ncore, nshare = pod.get_core_allocation(float(data['ncore']))
ports = data.get('ports', [])
args = data.get('args', [])
strategy = data.get('strategy', 'average')
callback_url = data.get('callback_url', '')
if callback_url and not is_strict_url(callback_url):
abort(400, 'callback_url must start with http:// or https://')
ncontainer = int(data['ncontainer'])
if not ncontainer:
abort(400, 'ncontainer must be > 0')
networks = [ipam.get_pool(n) for n in data.get('networks', [])]
spec_ips = data.get('spec_ips', [])
appconfig = version.appconfig
entrypoint = data['entrypoint']
if entrypoint not in appconfig.entrypoints:
abort(400, 'Entrypoint %s not in app.yaml' % entrypoint)
hostname = data.get('hostname', '')
host = hostname and Host.get_by_name(hostname) or None
task_ids, watch_keys = [], []
host_cores = _get_strategy(strategy)(pod, ncontainer, ncore, nshare, host)
if not host_cores:
abort(400, 'Not enough core resources')
for (host, container_count), cores in host_cores.iteritems():
t = _create_task(
version,
host,
container_count,
cores,
nshare,
networks,
ports,
args,
spec_ips,
entrypoint,
data['env'],
image=data.get('image', ''),
callback_url=callback_url,
)
if not t:
continue
host.occupy_cores(cores, nshare)
task_ids.append(t.id)
watch_keys.append(t.result_key)
return {'tasks': task_ids, 'watch_keys': watch_keys}
def check_addr(addr):
"""addr is like 10.20.0.1/16 or 10.100.3.12/24"""
try:
interface = IPNetwork(addr)
except AddrFormatError:
abort(400, 'Not valid interface')
net = ipam.get_pool(str(interface.cidr))
if not net:
abort(400, 'Interface not found')
return {'available': interface.ip in net}
def create_public():
data = request.get_json()
pod, _, version = _get_instances(**data)
ports = data.get('ports', [])
args = data.get('args', [])
callback_url = data.get('callback_url', '')
if callback_url and not is_strict_url(callback_url):
abort(400, 'callback_url must starts with http:// or https://')
networks = [ipam.get_pool(n) for n in data.get('networks', [])]
spec_ips = data.get('spec_ips', [])
appconfig = version.appconfig
ncontainer = int(data['ncontainer'])
if not ncontainer:
abort(400, 'ncontainer must be > 0')
entrypoint = data['entrypoint']
if entrypoint not in appconfig.entrypoints:
abort(400, 'Entrypoint %s not in app.yaml' % entrypoint)
task_ids, watch_keys = [], []
hosts = pod.get_free_public_hosts(ncontainer)
for host in itertools.islice(itertools.cycle(hosts), ncontainer):
t = _create_task(
version,
host,
1,
{},
0,
networks,
ports,
args,
spec_ips,
data['entrypoint'],
data['env'],
image=data.get('image', ''),
callback_url=callback_url,
)
if not t:
continue
task_ids.append(t.id)
watch_keys.append(t.result_key)
return {'tasks': task_ids, 'watch_keys': watch_keys}
def macvlan(id_or_name):
host = _get_host(id_or_name)
if request.method == 'GET':
return host.list_vlans(g.start, g.limit)
data = request.get_json()
netname = data.get('network', '')
network = ipam.get_pool(netname)
if not network:
abort(404, 'Network not found')
vg = VLanGateway.get_by_host_and_network(host.id, network.id)
if request.method == 'POST':
return vg or network.acquire_gateway_ip(host)
elif request.method == 'DELETE':
if vg:
vg.release()
return DEFAULT_RETURN_VALUE
def bind_network(id_or_cid):
c = _get_container(id_or_cid)
data = request.get_json()
appname = data['appname']
if not c.is_alive:
abort(404, 'Container %s not alive' % c.container_id)
if c.appname != appname:
abort(404, 'Container does not belong to app')
if c.network_mode == 'host':
abort(400, 'Container use host network mode')
network_names = data.get('networks', [])
networks = [ipam.get_pool(n) for n in network_names]
if not networks:
abort(400, 'network empty')
cidrs = [n.cidr for n in networks if n]
bind_container_ip(c, cidrs)
return {'cidrs': cidrs}
def bind_network(id_or_cid):
c = _get_container(id_or_cid)
data = request.get_json()
appname = data['appname']
if not c.is_alive:
abort(404, 'Container %s not alive' % c.container_id)
if c.appname != appname:
abort(404, 'Container does not belong to app')
if c.network_mode == 'host':
abort(400, 'Container use host network mode')
network_names = data.get('networks', [])
networks = [ipam.get_pool(n) for n in network_names]
if not networks:
abort(400, 'network empty')
cidrs = [n.cidr for n in networks if n]
bind_container_ip(c, cidrs)
return {'cidrs': cidrs}
def create_containers(task_id, ncontainer, nshare, cores, network_ids, spec_ips=None):
"""
执行task_id的任务. 部署ncontainer个容器, 占用*_core_ids这些核, 绑定到networks这些子网
"""
_log.info('Task<id=%s>: Started', task_id)
task = Task.get(task_id)
if not task:
_log.error('Task (id=%s) not found, quit', task_id)
return
if spec_ips is None:
spec_ips = []
need_network = bool(network_ids)
networks = [ipam.get_pool(n) for n in network_ids]
notifier = TaskNotifier(task)
host = task.host
version = task.version
entrypoint = task.props['entrypoint']
env = task.props['env']
ports = task.props['ports']
args = task.props['args']
# use raw
image = task.props['image']
callback_url = task.props['callback_url']
cpu_shares = int(float(nshare) / host.pod.core_share * 1024) if nshare else 1024
cids = []
backends = []
entry = version.appconfig.entrypoints[entrypoint]
for fcores, pcores in _iter_cores(cores, ncontainer):
cores_for_one_container = {'full': fcores, 'part': pcores}
# 在宿主机上创建容器
try:
cid, cname = dockerjob.create_one_container(host,
version,
entrypoint,
env,
fcores + pcores,
ports=ports, args=args,
cpu_shares=cpu_shares,
image=image,
need_network=need_network)
except Exception as e:
# 写给celery日志看
_log.exception(e)
host.release_cores(cores_for_one_container, nshare)
continue
# 容器记录下来
c = Container.create(cid, host, version, cname, entrypoint, cores_for_one_container, env, nshare, callback_url)
# 为容器创建网络栈
# 同时把各种信息都记录下来
# 如果失败, 清除掉所有记录和宿主机上的容器
# 循环下一次尝试
cidrs = [n.netspace for n in networks]
if not ipam.allocate_ips(cidrs, cid, spec_ips):
_clean_failed_containers(cid)
continue
notifier.notify_agent(c)
add_container_for_agent(host, c)
add_container_backends(c)
cids.append(cid)
backends.extend(c.get_backends())
c.callback_report(status='start')
health_check = entry.get('health_check', '')
if health_check and backends:
urls = [b + health_check for b in backends]
if not wait_health_check(urls):
# TODO 这里要么回滚要么报警
_log.info('Task<id=%s>: Done, but something went error', task_id)
return
publish_to_service_discovery(version.name)
task.finish(consts.TASK_SUCCESS)
task.reason = 'ok'
task.container_ids = cids
notifier.pub_success()
_log.info('Task<id=%s>: Done', task_id)
def create_containers(task_id,
ncontainer,
nshare,
cores,
network_ids,
spec_ips=None):
"""
执行task_id的任务. 部署ncontainer个容器, 占用*_core_ids这些核, 绑定到networks这些子网
"""
_log.info('Task<id=%s>: Started', task_id)
task = Task.get(task_id)
if not task:
_log.error('Task (id=%s) not found, quit', task_id)
return
if spec_ips is None:
spec_ips = []
need_network = bool(network_ids)
networks = [ipam.get_pool(n) for n in network_ids]
notifier = TaskNotifier(task)
host = task.host
version = task.version
entrypoint = task.props['entrypoint']
env = task.props['env']
ports = task.props['ports']
args = task.props['args']
# use raw
image = task.props['image']
callback_url = task.props['callback_url']
cpu_shares = int(float(nshare) / host.pod.core_share *
1024) if nshare else 1024
cids = []
backends = []
entry = version.appconfig.entrypoints[entrypoint]
for fcores, pcores in _iter_cores(cores, ncontainer):
cores_for_one_container = {'full': fcores, 'part': pcores}
# 在宿主机上创建容器
try:
cid, cname = dockerjob.create_one_container(
host,
version,
entrypoint,
env,
fcores + pcores,
ports=ports,
args=args,
cpu_shares=cpu_shares,
image=image,
need_network=need_network)
except Exception as e:
# 写给celery日志看
_log.exception(e)
host.release_cores(cores_for_one_container, nshare)
continue
# 容器记录下来
c = Container.create(cid, host, version, cname, entrypoint,
cores_for_one_container, env, nshare,
callback_url)
# 为容器创建网络栈
# 同时把各种信息都记录下来
# 如果失败, 清除掉所有记录和宿主机上的容器
# 循环下一次尝试
cidrs = [n.netspace for n in networks]
if not ipam.allocate_ips(cidrs, cid, spec_ips):
_clean_failed_containers(cid)
continue
notifier.notify_agent(c)
add_container_for_agent(host, c)
add_container_backends(c)
cids.append(cid)
backends.extend(c.get_backends())
c.callback_report(status='start')
health_check = entry.get('health_check', '')
if health_check and backends:
urls = [b + health_check for b in backends]
if not wait_health_check(urls):
# TODO 这里要么回滚要么报警
_log.info('Task<id=%s>: Done, but something went error', task_id)
return
publish_to_service_discovery(version.name)
task.finish(consts.TASK_SUCCESS)
task.reason = 'ok'
task.container_ids = cids
notifier.pub_success()
_log.info('Task<id=%s>: Done', task_id)
def get_network(id_or_name):
n = ipam.get_pool(id_or_name)
if not n:
abort(404, 'Network %s not found' % id_or_name)
return n