Remove ucloud_common and put its files under ucloud.common subpackage.

Remove individual config.py used by every component and put them into single config.py ucloud/config.py
Use /etc/ucloud/ucloud.conf for Environment Variables
Refactoring and a lot of it
Make ucloud repo a package and different components of ucloud a subpackage for avoiding code duplication.
Improved logging.
This commit is contained in:
ahmadbilalkhalid 2019-11-18 22:39:57 +05:00
parent 1d2b980c74
commit 6fa77bce4d
51 changed files with 890 additions and 567 deletions

3
.gitignore vendored
View file

@ -1,8 +1,7 @@
.idea
.vscode
.env
__pycache__
docs/build
*/log.txt
logs.txt

View file

@ -5,6 +5,7 @@ verify_ssl = true
[dev-packages]
prospector = {extras = ["with_everything"],version = "*"}
pylama = "*"
[packages]
python-decouple = "*"
@ -12,7 +13,6 @@ requests = "*"
flask = "*"
flask-restful = "*"
bitmath = "*"
ucloud-common = {editable = true,git = "git+https://code.ungleich.ch/ucloud/ucloud_common.git",ref = "wip"}
etcd3-wrapper = {editable = true,git = "git+https://code.ungleich.ch/ungleich-public/etcd3_wrapper.git",ref = "wip"}
python-etcd3 = {editable = true,git = "git+https://github.com/kragniz/python-etcd3.git"}
pyotp = "*"

23
Pipfile.lock generated
View file

@ -1,7 +1,7 @@
{
"_meta": {
"hash": {
"sha256": "5e4aa65086afdf9ac2f1479e9e35684f767dfbbd13877c4e4a23dd471aef6c13"
"sha256": "f43a93c020eb20212b437fcc62882db03bfa93f4678eb930e31343d687c805ed"
},
"pipfile-spec": 6,
"requires": {
@ -379,10 +379,10 @@
},
"pynetbox": {
"hashes": [
"sha256:e171380b36bedb7e0cd6a735fe8193d5809b373897b6905a2de43342761426c7"
"sha256:09525a29f1ac8c1a54772d6e2b94a55b1db6ba6a1c5b07f7af6a6ce232b1f7d5"
],
"index": "pypi",
"version": "==4.0.8"
"version": "==4.1.0"
},
"pyotp": {
"hashes": [
@ -401,10 +401,10 @@
},
"python-decouple": {
"hashes": [
"sha256:1317df14b43efee4337a4aa02914bf004f010cd56d6c4bd894e6474ec8c4fe2d"
"sha256:55c546b85b0c47a15a47a4312d451a437f7344a9be3e001660bccd93b637de95"
],
"index": "pypi",
"version": "==3.1"
"version": "==3.3"
},
"python-etcd3": {
"editable": true,
@ -522,11 +522,6 @@
],
"version": "==6.0.0"
},
"ucloud-common": {
"editable": true,
"git": "https://code.ungleich.ch/ucloud/ucloud_common.git",
"ref": "eba92e5d6723093a3cc2999ae1f5c284e65dc809"
},
"urllib3": {
"hashes": [
"sha256:a8a318824cc77d1fd4b2bec2ded92646630d7fe8619497b142c84a9e6f5a7293",
@ -775,6 +770,14 @@
],
"version": "==2.4.2"
},
"pylama": {
"hashes": [
"sha256:9bae53ef9c1a431371d6a8dca406816a60d547147b60a4934721898f553b7d8f",
"sha256:fd61c11872d6256b019ef1235be37b77c922ef37ac9797df6bd489996dddeb15"
],
"index": "pypi",
"version": "==7.7.1"
},
"pylint": {
"hashes": [
"sha256:5d77031694a5fb97ea95e828c8d10fc770a1df6eb3906067aaed42201a8a6a09",

0
__init__.py Normal file
View file

3
api/__init__.py Normal file
View file

@ -0,0 +1,3 @@
import logging
logger = logging.getLogger(__name__)

View file

@ -1,7 +1,6 @@
import os
from config import etcd_client as client
from config import VM_PREFIX
from config import etcd_client, env_vars
class Optional:
@ -49,6 +48,6 @@ class VmUUIDField(Field):
self.validation = self.vm_uuid_validation
def vm_uuid_validation(self):
r = client.get(os.path.join(VM_PREFIX, self.uuid))
r = etcd_client.get(os.path.join(env_vars.get('VM_PREFIX'), self.uuid))
if not r:
self.add_error("VM with uuid {} does not exists".format(self.uuid))

View file

@ -1,33 +0,0 @@
import logging
from etcd3_wrapper import Etcd3Wrapper
from decouple import config
from ucloud_common.vm import VmPool
from ucloud_common.host import HostPool
from ucloud_common.request import RequestPool
logging.basicConfig(
level=logging.DEBUG,
filename="log.txt",
filemode="a",
format="%(asctime)s: %(levelname)s - %(message)s",
datefmt="%d-%b-%y %H:%M:%S",
)
WITHOUT_CEPH = config("WITHOUT_CEPH", False, cast=bool)
VM_PREFIX = config("VM_PREFIX")
HOST_PREFIX = config("HOST_PREFIX")
REQUEST_PREFIX = config("REQUEST_PREFIX")
FILE_PREFIX = config("FILE_PREFIX")
IMAGE_PREFIX = config("IMAGE_PREFIX")
IMAGE_STORE_PREFIX = config("IMAGE_STORE_PREFIX")
NETWORK_PREFIX = config("NETWORK_PREFIX")
etcd_client = Etcd3Wrapper(host=config("ETCD_URL"))
VM_POOL = VmPool(etcd_client, VM_PREFIX)
HOST_POOL = HostPool(etcd_client, HOST_PREFIX)
REQUEST_POOL = RequestPool(etcd_client, REQUEST_PREFIX)

View file

@ -1,10 +1,8 @@
import json
import os
from uuid import uuid4
from config import etcd_client as client
from config import IMAGE_STORE_PREFIX
from config import etcd_client, env_vars
data = {
"is_public": True,
@ -14,4 +12,4 @@ data = {
"attributes": {"list": [], "key": [], "pool": "images"},
}
client.put(os.path.join(IMAGE_STORE_PREFIX, uuid4().hex), json.dumps(data))
etcd_client.put(os.path.join(env_vars.get('IMAGE_STORE_PREFIX'), uuid4().hex), json.dumps(data))

View file

@ -1,20 +1,20 @@
import binascii
import requests
import ipaddress
import random
import subprocess as sp
import ipaddress
from decouple import config
import requests
from pyotp import TOTP
from config import VM_POOL, etcd_client, IMAGE_PREFIX
from config import vm_pool, env_vars
def check_otp(name, realm, token):
try:
data = {
"auth_name": config("AUTH_NAME", ""),
"auth_token": TOTP(config("AUTH_SEED", "")).now(),
"auth_realm": config("AUTH_REALM", ""),
"auth_name": env_vars.get("AUTH_NAME"),
"auth_token": TOTP(env_vars.get("AUTH_SEED")).now(),
"auth_realm": env_vars.get("AUTH_REALM"),
"name": name,
"realm": realm,
"token": token,
@ -24,8 +24,8 @@ def check_otp(name, realm, token):
response = requests.get(
"{OTP_SERVER}{OTP_VERIFY_ENDPOINT}".format(
OTP_SERVER=config("OTP_SERVER", ""),
OTP_VERIFY_ENDPOINT=config("OTP_VERIFY_ENDPOINT", "verify"),
OTP_SERVER=env_vars.get("OTP_SERVER", ""),
OTP_VERIFY_ENDPOINT=env_vars.get("OTP_VERIFY_ENDPOINT", "verify"),
),
json=data,
)
@ -41,7 +41,7 @@ def resolve_vm_name(name, owner):
result = next(
filter(
lambda vm: vm.value["owner"] == owner and vm.value["name"] == name,
VM_POOL.vms,
vm_pool.vms,
),
None,
)
@ -61,7 +61,7 @@ def resolve_image_name(name, etcd_client):
"""
seperator = ":"
# Ensure, user/program passed valid name that is of type string
try:
store_name_and_image_name = name.split(seperator)
@ -79,7 +79,7 @@ def resolve_image_name(name, etcd_client):
except Exception:
raise ValueError("Image name not in correct format i.e {store_name}:{image_name}")
images = etcd_client.get_prefix(IMAGE_PREFIX, value_in_json=True)
images = etcd_client.get_prefix(env_vars.get('IMAGE_PREFIX'), value_in_json=True)
# Try to find image with name == image_name and store_name == store_name
try:
@ -89,7 +89,7 @@ def resolve_image_name(name, etcd_client):
raise KeyError("No image with name {} found.".format(name))
else:
image_uuid = image.key.split('/')[-1]
return image_uuid
@ -102,16 +102,16 @@ def generate_mac(uaa=False, multicast=False, oui=None, separator=':', byte_fmt='
if oui:
if type(oui) == str:
oui = [int(chunk) for chunk in oui.split(separator)]
mac = oui + random_bytes(num=6-len(oui))
mac = oui + random_bytes(num=6 - len(oui))
else:
if multicast:
mac[0] |= 1 # set bit 0
mac[0] |= 1 # set bit 0
else:
mac[0] &= ~1 # clear bit 0
mac[0] &= ~1 # clear bit 0
if uaa:
mac[0] &= ~(1 << 1) # clear bit 1
mac[0] &= ~(1 << 1) # clear bit 1
else:
mac[0] |= 1 << 1 # set bit 1
mac[0] |= 1 << 1 # set bit 1
return separator.join(byte_fmt % b for b in mac)
@ -128,7 +128,7 @@ def get_ip_addr(mac_address, device):
and is connected/neighbor of arg:device
"""
try:
output = sp.check_output(['ip','-6','neigh', 'show', 'dev', device], stderr=sp.PIPE)
output = sp.check_output(['ip', '-6', 'neigh', 'show', 'dev', device], stderr=sp.PIPE)
except sp.CalledProcessError:
return None
else:
@ -145,25 +145,6 @@ def get_ip_addr(mac_address, device):
return result
def increment_etcd_counter(etcd_client, key):
kv = etcd_client.get(key)
if kv:
counter = int(kv.value)
counter = counter + 1
else:
counter = 1
etcd_client.put(key, str(counter))
return counter
def get_etcd_counter(etcd_client, key):
kv = etcd_client.get(key)
if kv:
return int(kv.value)
return None
def mac2ipv6(mac, prefix):
# only accept MACs separated by a colon
parts = mac.split(":")
@ -174,10 +155,10 @@ def mac2ipv6(mac, prefix):
parts[0] = "%x" % (int(parts[0], 16) ^ 2)
# format output
ipv6Parts = [str(0)]*4
ipv6Parts = [str(0)] * 4
for i in range(0, len(parts), 2):
ipv6Parts.append("".join(parts[i:i+2]))
ipv6Parts.append("".join(parts[i:i + 2]))
lower_part = ipaddress.IPv6Address(":".join(ipv6Parts))
prefix = ipaddress.IPv6Address(prefix)
return str(prefix + int(lower_part))

View file

@ -1,35 +1,19 @@
import json
import subprocess
import os
import pynetbox
import decouple
import schemas
import subprocess
from uuid import uuid4
import pynetbox
from flask import Flask, request
from flask_restful import Resource, Api
from ucloud_common.vm import VMStatus
from ucloud_common.request import RequestEntry, RequestType
from helper import (generate_mac, get_ip_addr, get_etcd_counter,
increment_etcd_counter, mac2ipv6)
from config import (
etcd_client,
WITHOUT_CEPH,
VM_PREFIX,
HOST_PREFIX,
FILE_PREFIX,
IMAGE_PREFIX,
NETWORK_PREFIX,
logging,
REQUEST_POOL,
VM_POOL,
HOST_POOL,
)
from common import counters
from common.request import RequestEntry, RequestType
from common.vm import VMStatus
from config import (etcd_client, request_pool, vm_pool, host_pool, env_vars)
from . import schemas
from .helper import generate_mac, mac2ipv6
from api import logger
app = Flask(__name__)
api = Api(app)
@ -42,12 +26,12 @@ class CreateVM(Resource):
validator = schemas.CreateVMSchema(data)
if validator.is_valid():
vm_uuid = uuid4().hex
vm_key = os.path.join(VM_PREFIX, vm_uuid)
vm_key = os.path.join(env_vars.get("VM_PREFIX"), vm_uuid)
specs = {
'cpu': validator.specs['cpu'],
'ram': validator.specs['ram'],
'os-ssd': validator.specs['os-ssd'],
'hdd': validator.specs['hdd']
"cpu": validator.specs["cpu"],
"ram": validator.specs["ram"],
"os-ssd": validator.specs["os-ssd"],
"hdd": validator.specs["hdd"],
}
macs = [generate_mac() for i in range(len(data["network"]))]
vm_entry = {
@ -61,15 +45,16 @@ class CreateVM(Resource):
"log": [],
"vnc_socket": "",
"network": list(zip(data["network"], macs)),
"metadata": {
"ssh-keys": []
},
"metadata": {"ssh-keys": []},
}
etcd_client.put(vm_key, vm_entry, value_in_json=True)
# Create ScheduleVM Request
r = RequestEntry.from_scratch(type=RequestType.ScheduleVM, uuid=vm_uuid)
REQUEST_POOL.put(r)
r = RequestEntry.from_scratch(
type=RequestType.ScheduleVM, uuid=vm_uuid,
request_prefix=env_vars.get("REQUEST_PREFIX")
)
request_pool.put(r)
return {"message": "VM Creation Queued"}, 200
return validator.get_errors(), 400
@ -81,13 +66,21 @@ class VmStatus(Resource):
data = request.json
validator = schemas.VMStatusSchema(data)
if validator.is_valid():
vm = VM_POOL.get(os.path.join(VM_PREFIX, data["uuid"]))
vm = vm_pool.get(
os.path.join(env_vars.get("VM_PREFIX"), data["uuid"])
)
vm_value = vm.value.copy()
vm_value["ip"] = []
for network_and_mac in vm.network:
network_name, mac = network_and_mac
network = etcd_client.get(os.path.join(NETWORK_PREFIX, data["name"], network_name),
value_in_json=True)
network = etcd_client.get(
os.path.join(
env_vars.get("NETWORK_PREFIX"),
data["name"],
network_name,
),
value_in_json=True,
)
ipv6_addr = network.value.get("ipv6").split("::")[0] + "::"
vm_value["ip"].append(mac2ipv6(mac, ipv6_addr))
vm.value = vm_value
@ -102,7 +95,9 @@ class CreateImage(Resource):
data = request.json
validator = schemas.CreateImageSchema(data)
if validator.is_valid():
file_entry = etcd_client.get(os.path.join(FILE_PREFIX, data["uuid"]))
file_entry = etcd_client.get(
os.path.join(env_vars.get("FILE_PREFIX"), data["uuid"])
)
file_entry_value = json.loads(file_entry.value)
image_entry_json = {
@ -114,7 +109,8 @@ class CreateImage(Resource):
"visibility": "public",
}
etcd_client.put(
os.path.join(IMAGE_PREFIX, data["uuid"]), json.dumps(image_entry_json)
os.path.join(env_vars.get("IMAGE_PREFIX"), data["uuid"]),
json.dumps(image_entry_json),
)
return {"message": "Image queued for creation."}
@ -124,15 +120,18 @@ class CreateImage(Resource):
class ListPublicImages(Resource):
@staticmethod
def get():
images = etcd_client.get_prefix(IMAGE_PREFIX, value_in_json=True)
images = etcd_client.get_prefix(
env_vars.get("IMAGE_PREFIX"), value_in_json=True
)
r = {}
r["images"] = []
for image in images:
image_key = "{}:{}".format(image.value["store_name"], image.value["name"])
r["images"].append({
"name":image_key,
"status": image.value["status"]
})
image_key = "{}:{}".format(
image.value["store_name"], image.value["name"]
)
r["images"].append(
{"name": image_key, "status": image.value["status"]}
)
return r, 200
@ -143,34 +142,47 @@ class VMAction(Resource):
validator = schemas.VmActionSchema(data)
if validator.is_valid():
vm_entry = VM_POOL.get(os.path.join(VM_PREFIX, data["uuid"]))
vm_entry = vm_pool.get(
os.path.join(env_vars.get("VM_PREFIX"), data["uuid"])
)
action = data["action"]
if action == "start":
vm_entry.status = VMStatus.requested_start
VM_POOL.put(vm_entry)
vm_pool.put(vm_entry)
action = "schedule"
if action == "delete" and vm_entry.hostname == "":
try:
path_without_protocol = vm_entry.path[vm_entry.path.find(":") + 1 :]
path_without_protocol = vm_entry.path[
vm_entry.path.find(":") + 1:
]
if WITHOUT_CEPH:
if env_vars.get("WITHOUT_CEPH"):
command_to_delete = [
"rm", "-rf",
"rm",
"-rf",
os.path.join("/var/vm", vm_entry.uuid),
]
else:
command_to_delete = ["rbd", "rm", path_without_protocol]
command_to_delete = [
"rbd",
"rm",
path_without_protocol,
]
subprocess.check_output(command_to_delete, stderr=subprocess.PIPE)
subprocess.check_output(
command_to_delete, stderr=subprocess.PIPE
)
except subprocess.CalledProcessError as e:
if "No such file" in e.stderr.decode("utf-8"):
etcd_client.client.delete(vm_entry.key)
return {"message": "VM successfully deleted"}
else:
logging.exception(e)
return {"message": "Some error occurred while deleting VM"}
logger.exception(e)
return {
"message": "Some error occurred while deleting VM"
}
else:
etcd_client.client.delete(vm_entry.key)
return {"message": "VM successfully deleted"}
@ -179,8 +191,9 @@ class VMAction(Resource):
type="{}VM".format(action.title()),
uuid=data["uuid"],
hostname=vm_entry.hostname,
request_prefix=env_vars.get("REQUEST_PREFIX")
)
REQUEST_POOL.put(r)
request_pool.put(r)
return {"message": "VM {} Queued".format(action.title())}, 200
else:
return validator.get_errors(), 400
@ -193,15 +206,18 @@ class VMMigration(Resource):
validator = schemas.VmMigrationSchema(data)
if validator.is_valid():
vm = VM_POOL.get(data["uuid"])
vm = vm_pool.get(data["uuid"])
r = RequestEntry.from_scratch(
type=RequestType.ScheduleVM,
uuid=vm.uuid,
destination=os.path.join(HOST_PREFIX, data["destination"]),
destination=os.path.join(
env_vars.get("HOST_PREFIX"), data["destination"]
),
migration=True,
request_prefix=env_vars.get("REQUEST_PREFIX")
)
REQUEST_POOL.put(r)
request_pool.put(r)
return {"message": "VM Migration Initialization Queued"}, 200
else:
return validator.get_errors(), 400
@ -214,7 +230,9 @@ class ListUserVM(Resource):
validator = schemas.OTPSchema(data)
if validator.is_valid():
vms = etcd_client.get_prefix(VM_PREFIX, value_in_json=True)
vms = etcd_client.get_prefix(
env_vars.get("VM_PREFIX"), value_in_json=True
)
return_vms = []
user_vms = filter(lambda v: v.value["owner"] == data["name"], vms)
for vm in user_vms:
@ -246,9 +264,13 @@ class ListUserFiles(Resource):
validator = schemas.OTPSchema(data)
if validator.is_valid():
files = etcd_client.get_prefix(FILE_PREFIX, value_in_json=True)
files = etcd_client.get_prefix(
env_vars.get("FILE_PREFIX"), value_in_json=True
)
return_files = []
user_files = list(filter(lambda f: f.value["owner"] == data["name"], files))
user_files = list(
filter(lambda f: f.value["owner"] == data["name"], files)
)
for file in user_files:
return_files.append(
{
@ -267,7 +289,7 @@ class CreateHost(Resource):
data = request.json
validator = schemas.CreateHostSchema(data)
if validator.is_valid():
host_key = os.path.join(HOST_PREFIX, uuid4().hex)
host_key = os.path.join(env_vars.get("HOST_PREFIX"), uuid4().hex)
host_entry = {
"specs": data["specs"],
"hostname": data["hostname"],
@ -284,7 +306,7 @@ class CreateHost(Resource):
class ListHost(Resource):
@staticmethod
def get():
hosts = HOST_POOL.hosts
hosts = host_pool.hosts
r = {
host.key: {
"status": host.status,
@ -305,21 +327,38 @@ class GetSSHKeys(Resource):
if not validator.key_name.value:
# {user_prefix}/{realm}/{name}/key/
etcd_key = os.path.join(decouple.config("USER_PREFIX"), data["realm"],
data["name"], "key")
etcd_entry = etcd_client.get_prefix(etcd_key, value_in_json=True)
keys = {key.key.split("/")[-1]: key.value for key in etcd_entry}
etcd_key = os.path.join(
env_vars.get('USER_PREFIX'),
data["realm"],
data["name"],
"key",
)
etcd_entry = etcd_client.get_prefix(
etcd_key, value_in_json=True
)
keys = {
key.key.split("/")[-1]: key.value for key in etcd_entry
}
return {"keys": keys}
else:
# {user_prefix}/{realm}/{name}/key/{key_name}
etcd_key = os.path.join(decouple.config("USER_PREFIX"), data["realm"],
data["name"], "key", data["key_name"])
etcd_key = os.path.join(
env_vars.get('USER_PREFIX'),
data["realm"],
data["name"],
"key",
data["key_name"],
)
etcd_entry = etcd_client.get(etcd_key, value_in_json=True)
if etcd_entry:
return {"keys": {etcd_entry.key.split("/")[-1]: etcd_entry.value}}
return {
"keys": {
etcd_entry.key.split("/")[-1]: etcd_entry.value
}
}
else:
return {"keys": {}}
else:
@ -332,13 +371,22 @@ class AddSSHKey(Resource):
data = request.json
validator = schemas.AddSSHSchema(data)
if validator.is_valid():
# {user_prefix}/{realm}/{name}/key/{key_name}
etcd_key = os.path.join(USER_PREFIX, data["realm"], data["name"],
"key", data["key_name"])
etcd_key = os.path.join(
env_vars.get("USER_PREFIX"),
data["realm"],
data["name"],
"key",
data["key_name"],
)
etcd_entry = etcd_client.get(etcd_key, value_in_json=True)
if etcd_entry:
return {"message": "Key with name '{}' already exists".format(data["key_name"])}
return {
"message": "Key with name '{}' already exists".format(
data["key_name"]
)
}
else:
# Key Not Found. It implies user' haven't added any key yet.
etcd_client.put(etcd_key, data["key"], value_in_json=True)
@ -353,16 +401,25 @@ class RemoveSSHKey(Resource):
data = request.json
validator = schemas.RemoveSSHSchema(data)
if validator.is_valid():
# {user_prefix}/{realm}/{name}/key/{key_name}
etcd_key = os.path.join(USER_PREFIX, data["realm"], data["name"],
"key", data["key_name"])
etcd_key = os.path.join(
env_vars.get("USER_PREFIX"),
data["realm"],
data["name"],
"key",
data["key_name"],
)
etcd_entry = etcd_client.get(etcd_key, value_in_json=True)
if etcd_entry:
etcd_client.client.delete(etcd_key)
return {"message": "Key successfully removed."}
else:
return {"message": "No Key with name '{}' Exists at all.".format(data["key_name"])}
return {
"message": "No Key with name '{}' Exists at all.".format(
data["key_name"]
)
}
else:
return validator.get_errors(), 400
@ -374,29 +431,42 @@ class CreateNetwork(Resource):
validator = schemas.CreateNetwork(data)
if validator.is_valid():
network_entry = {
"id": increment_etcd_counter(etcd_client, "/v1/counter/vxlan"),
"id": counters.increment_etcd_counter(
etcd_client, "/v1/counter/vxlan"
),
"type": data["type"],
}
if validator.user.value:
nb = pynetbox.api(url=decouple.config("NETBOX_URL"),
token=decouple.config("NETBOX_TOKEN"))
nb_prefix = nb.ipam.prefixes.get(prefix=decouple.config("PREFIX"))
nb = pynetbox.api(
url=env_vars.get("NETBOX_URL"),
token=env_vars.get("NETBOX_TOKEN"),
)
nb_prefix = nb.ipam.prefixes.get(
prefix=env_vars.get("PREFIX")
)
prefix = nb_prefix.available_prefixes.create(data=
{
"prefix_length": decouple.config("PREFIX_LENGTH", cast=int),
"description": "{}'s network \"{}\"".format(data["name"],
data["network_name"]),
"is_pool": True
prefix = nb_prefix.available_prefixes.create(
data={
"prefix_length": env_vars.get(
"PREFIX_LENGTH", cast=int
),
"description": '{}\'s network "{}"'.format(
data["name"], data["network_name"]
),
"is_pool": True,
}
)
network_entry["ipv6"] = prefix["prefix"]
else:
network_entry["ipv6"] = "fd00::/64"
network_key = os.path.join(NETWORK_PREFIX, data["name"], data["network_name"])
network_key = os.path.join(
env_vars.get("NETWORK_PREFIX"),
data["name"],
data["network_name"],
)
etcd_client.put(network_key, network_entry, value_in_json=True)
return {"message": "Network successfully added."}
else:
@ -410,7 +480,9 @@ class ListUserNetwork(Resource):
validator = schemas.OTPSchema(data)
if validator.is_valid():
prefix = os.path.join(NETWORK_PREFIX, data["name"])
prefix = os.path.join(
env_vars.get("NETWORK_PREFIX"), data["name"]
)
networks = etcd_client.get_prefix(prefix, value_in_json=True)
user_networks = []
for net in networks:
@ -443,5 +515,20 @@ api.add_resource(ListHost, "/host/list")
api.add_resource(CreateNetwork, "/network/create")
if __name__ == "__main__":
def main():
data = {
"is_public": True,
"type": "ceph",
"name": "images",
"description": "first ever public image-store",
"attributes": {"list": [], "key": [], "pool": "images"},
}
etcd_client.put(os.path.join(env_vars.get('IMAGE_STORE_PREFIX'), uuid4().hex), json.dumps(data))
app.run(host="::", debug=True)
if __name__ == "__main__":
main()

View file

@ -16,21 +16,15 @@ inflexible for our purpose.
import json
import os
import bitmath
import helper
from ucloud_common.host import HostPool, HostStatus
from ucloud_common.vm import VmPool, VMStatus
from common_fields import Field, VmUUIDField, Optional
from helper import check_otp, resolve_vm_name
from config import etcd_client as client
from config import (HOST_PREFIX, VM_PREFIX, IMAGE_PREFIX,
FILE_PREFIX, IMAGE_STORE_PREFIX, NETWORK_PREFIX)
HOST_POOL = HostPool(client, HOST_PREFIX)
VM_POOL = VmPool(client, VM_PREFIX)
from common.host import HostStatus
from common.vm import VMStatus
from config import etcd_client, env_vars, vm_pool, host_pool
from . import helper
from .common_fields import Field, VmUUIDField
from .helper import check_otp, resolve_vm_name
class BaseSchema:
@ -108,14 +102,14 @@ class CreateImageSchema(BaseSchema):
super().__init__(data, fields)
def file_uuid_validation(self):
file_entry = client.get(os.path.join(FILE_PREFIX, self.uuid.value))
file_entry = etcd_client.get(os.path.join(env_vars.get('FILE_PREFIX'), self.uuid.value))
if file_entry is None:
self.add_error(
"Image File with uuid '{}' Not Found".format(self.uuid.value)
)
def image_store_name_validation(self):
image_stores = list(client.get_prefix(IMAGE_STORE_PREFIX))
image_stores = list(etcd_client.get_prefix(env_vars.get('IMAGE_STORE_PREFIX')))
image_store = next(
filter(
@ -205,7 +199,7 @@ class CreateHostSchema(OTPSchema):
class CreateVMSchema(OTPSchema):
def __init__(self, data):
self.parsed_specs = {}
# Fields
self.specs = Field("specs", dict, data.get("specs", KeyError))
self.vm_name = Field("vm_name", str, data.get("vm_name", KeyError))
@ -222,10 +216,9 @@ class CreateVMSchema(OTPSchema):
super().__init__(data=data, fields=fields)
def image_validation(self):
try:
image_uuid = helper.resolve_image_name(self.image.value, client)
image_uuid = helper.resolve_image_name(self.image.value, etcd_client)
except Exception as e:
self.add_error(str(e))
else:
@ -237,17 +230,17 @@ class CreateVMSchema(OTPSchema):
'VM with same name "{}" already exists'.format(self.vm_name.value)
)
def network_validation(self):
def network_validation(self):
_network = self.network.value
if _network:
for net in _network:
network = client.get(os.path.join(NETWORK_PREFIX,
self.name.value,
net), value_in_json=True)
network = etcd_client.get(os.path.join(env_vars.get('NETWORK_PREFIX'),
self.name.value,
net), value_in_json=True)
if not network:
self.add_error("Network with name {} does not exists"\
.format(net))
self.add_error("Network with name {} does not exists" \
.format(net))
def specs_validation(self):
ALLOWED_BASE = 10
@ -316,7 +309,7 @@ class VMStatusSchema(OTPSchema):
super().__init__(data, fields)
def validation(self):
vm = VM_POOL.get(self.uuid.value)
vm = vm_pool.get(self.uuid.value)
if not (
vm.value["owner"] == self.name.value or self.realm.value == "ungleich-admin"
):
@ -349,7 +342,7 @@ class VmActionSchema(OTPSchema):
)
def validation(self):
vm = VM_POOL.get(self.uuid.value)
vm = vm_pool.get(self.uuid.value)
if not (
vm.value["owner"] == self.name.value or self.realm.value == "ungleich-admin"
):
@ -389,14 +382,14 @@ class VmMigrationSchema(OTPSchema):
def destination_validation(self):
host_key = self.destination.value
host = HOST_POOL.get(host_key)
host = host_pool.get(host_key)
if not host:
self.add_error("No Such Host ({}) exists".format(self.destination.value))
elif host.status != HostStatus.alive:
self.add_error("Destination Host is dead")
def validation(self):
vm = VM_POOL.get(self.uuid.value)
vm = vm_pool.get(self.uuid.value)
if not (
vm.value["owner"] == self.name.value or self.realm.value == "ungleich-admin"
):
@ -405,7 +398,7 @@ class VmMigrationSchema(OTPSchema):
if vm.status != VMStatus.running:
self.add_error("Can't migrate non-running VM")
if vm.hostname == os.path.join(HOST_PREFIX, self.destination.value):
if vm.hostname == os.path.join(env_vars.get('HOST_PREFIX'), self.destination.value):
self.add_error("Destination host couldn't be same as Source Host")
@ -445,15 +438,15 @@ class CreateNetwork(OTPSchema):
fields = [self.network_name, self.type, self.user]
super().__init__(data, fields=fields)
def network_name_validation(self):
network = client.get(os.path.join(NETWORK_PREFIX,
self.name.value,
self.network_name.value),
value_in_json=True)
network = etcd_client.get(os.path.join(env_vars.get('NETWORK_PREFIX'),
self.name.value,
self.network_name.value),
value_in_json=True)
if network:
self.add_error("Network with name {} already exists"\
.format(self.network_name.value))
self.add_error("Network with name {} already exists" \
.format(self.network_name.value))
def network_type_validation(self):
supported_network_types = ["vxlan"]

0
common/__init__.py Normal file
View file

48
common/classes.py Normal file
View file

@ -0,0 +1,48 @@
from decouple import Config, RepositoryEnv, UndefinedValueError
from etcd3_wrapper import EtcdEntry
class EnvironmentVariables:
def __init__(self, env_file):
try:
env_config = Config(RepositoryEnv(env_file))
except FileNotFoundError:
print("{} does not exists".format(env_file))
exit(1)
else:
self.config = env_config
def get(self, *args, **kwargs):
"""Return value of var from env_vars"""
try:
value = self.config.get(*args, **kwargs)
except UndefinedValueError as e:
print(e)
exit(1)
else:
return value
class SpecificEtcdEntryBase:
def __init__(self, e: EtcdEntry):
self.key = e.key
for k in e.value.keys():
self.__setattr__(k, e.value[k])
def original_keys(self):
r = dict(self.__dict__)
if "key" in r:
del r["key"]
return r
@property
def value(self):
return self.original_keys()
@value.setter
def value(self, v):
self.__dict__ = v
def __repr__(self):
return str(dict(self.__dict__))

21
common/counters.py Normal file
View file

@ -0,0 +1,21 @@
from etcd3_wrapper import Etcd3Wrapper
def increment_etcd_counter(etcd_client: Etcd3Wrapper, key):
kv = etcd_client.get(key)
if kv:
counter = int(kv.value)
counter = counter + 1
else:
counter = 1
etcd_client.put(key, str(counter))
return counter
def get_etcd_counter(etcd_client: Etcd3Wrapper, key):
kv = etcd_client.get(key)
if kv:
return int(kv.value)
return None

39
common/helpers.py Normal file
View file

@ -0,0 +1,39 @@
import logging
import socket
from os.path import join as join_path
def create_package_loggers(packages, base_path, mode="a"):
loggers = {}
for pkg in packages:
logger = logging.getLogger(pkg)
logger_handler = logging.FileHandler(
join_path(base_path, "{}.txt".format(pkg)),
mode=mode
)
logger.setLevel(logging.DEBUG)
logger_handler.setFormatter(logging.Formatter(fmt="%(asctime)s: %(levelname)s - %(message)s",
datefmt="%d-%b-%y %H:%M:%S"))
logger.addHandler(logger_handler)
loggers[pkg] = logger
# TODO: Should be removed as soon as migration
# mechanism is finalized inside ucloud
def get_ipv4_address():
# If host is connected to internet
# Return IPv4 address of machine
# Otherwise, return 127.0.0.1
with socket.socket(socket.AF_INET, socket.SOCK_DGRAM) as s:
try:
s.connect(("8.8.8.8", 80))
except socket.timeout:
address = "127.0.0.1"
except Exception as e:
logging.getLogger().exception(e)
address = "127.0.0.1"
else:
address = s.getsockname()[0]
return address

67
common/host.py Normal file
View file

@ -0,0 +1,67 @@
import time
from datetime import datetime
from os.path import join
from typing import List
from .classes import SpecificEtcdEntryBase
class HostStatus:
"""Possible Statuses of ucloud host."""
alive = "ALIVE"
dead = "DEAD"
class HostEntry(SpecificEtcdEntryBase):
"""Represents Host Entry Structure and its supporting methods."""
def __init__(self, e):
self.specs = None # type: dict
self.hostname = None # type: str
self.status = None # type: str
self.last_heartbeat = None # type: str
super().__init__(e)
def update_heartbeat(self):
self.status = HostStatus.alive
self.last_heartbeat = time.strftime("%Y-%m-%d %H:%M:%S")
def is_alive(self):
last_heartbeat = datetime.strptime(self.last_heartbeat, "%Y-%m-%d %H:%M:%S")
delta = datetime.now() - last_heartbeat
if delta.total_seconds() > 60:
return False
return True
def declare_dead(self):
self.status = HostStatus.dead
self.last_heartbeat = time.strftime("%Y-%m-%d %H:%M:%S")
class HostPool:
def __init__(self, etcd_client, host_prefix):
self.client = etcd_client
self.prefix = host_prefix
@property
def hosts(self) -> List[HostEntry]:
_hosts = self.client.get_prefix(self.prefix, value_in_json=True)
return [HostEntry(host) for host in _hosts]
def get(self, key):
if not key.startswith(self.prefix):
key = join(self.prefix, key)
v = self.client.get(key, value_in_json=True)
if v:
return HostEntry(v)
return None
def put(self, obj: HostEntry):
self.client.put(obj.key, obj.value, value_in_json=True)
def by_status(self, status, _hosts=None):
if _hosts is None:
_hosts = self.hosts
return list(filter(lambda x: x.status == status, _hosts))

46
common/request.py Normal file
View file

@ -0,0 +1,46 @@
import json
from os.path import join
from uuid import uuid4
from etcd3_wrapper.etcd3_wrapper import PsuedoEtcdEntry
from .classes import SpecificEtcdEntryBase
class RequestType:
CreateVM = "CreateVM"
ScheduleVM = "ScheduleVM"
StartVM = "StartVM"
StopVM = "StopVM"
InitVMMigration = "InitVMMigration"
TransferVM = "TransferVM"
DeleteVM = "DeleteVM"
class RequestEntry(SpecificEtcdEntryBase):
def __init__(self, e):
self.type = None # type: str
self.migration = None # type: bool
self.destination = None # type: str
self.uuid = None # type: str
self.hostname = None # type: str
super().__init__(e)
@classmethod
def from_scratch(cls, request_prefix, **kwargs):
e = PsuedoEtcdEntry(join(request_prefix, uuid4().hex),
value=json.dumps(kwargs).encode("utf-8"), value_in_json=True)
return cls(e)
class RequestPool:
def __init__(self, etcd_client, request_prefix):
self.client = etcd_client
self.prefix = request_prefix
def put(self, obj: RequestEntry):
if not obj.key.startswith(self.prefix):
obj.key = join(self.prefix, obj.key)
self.client.put(obj.key, obj.value, value_in_json=True)

110
common/vm.py Normal file
View file

@ -0,0 +1,110 @@
from contextlib import contextmanager
from datetime import datetime
from os.path import join
from .classes import SpecificEtcdEntryBase
class VMStatus:
# Must be only assigned to brand new VM
requested_new = "REQUESTED_NEW"
# Only Assigned to already created vm
requested_start = "REQUESTED_START"
# These all are for running vms
requested_shutdown = "REQUESTED_SHUTDOWN"
requested_migrate = "REQUESTED_MIGRATE"
requested_delete = "REQUESTED_DELETE"
# either its image is not found or user requested
# to delete it
deleted = "DELETED"
stopped = "STOPPED" # After requested_shutdown
killed = "KILLED" # either host died or vm died itself
running = "RUNNING"
error = "ERROR" # An error occurred that cannot be resolved automatically
class VMEntry(SpecificEtcdEntryBase):
def __init__(self, e):
self.owner = None # type: str
self.specs = None # type: dict
self.hostname = None # type: str
self.status = None # type: str
self.image_uuid = None # type: str
self.log = None # type: list
self.in_migration = None # type: bool
super().__init__(e)
@property
def uuid(self):
return self.key.split("/")[-1]
def declare_killed(self):
self.hostname = ""
self.in_migration = False
if self.status == VMStatus.running:
self.status = VMStatus.killed
def declare_stopped(self):
self.hostname = ""
self.in_migration = False
self.status = VMStatus.stopped
def add_log(self, msg):
self.log = self.log[:5]
self.log.append("{} - {}".format(datetime.now().isoformat(), msg))
@property
def path(self):
return "rbd:uservms/{}".format(self.uuid)
class VmPool:
def __init__(self, etcd_client, vm_prefix):
self.client = etcd_client
self.prefix = vm_prefix
@property
def vms(self):
_vms = self.client.get_prefix(self.prefix, value_in_json=True)
return [VMEntry(vm) for vm in _vms]
def by_host(self, host, _vms=None):
if _vms is None:
_vms = self.vms
return list(filter(lambda x: x.hostname == host, _vms))
def by_status(self, status, _vms=None):
if _vms is None:
_vms = self.vms
return list(filter(lambda x: x.status == status, _vms))
def except_status(self, status, _vms=None):
if _vms is None:
_vms = self.vms
return list(filter(lambda x: x.status != status, _vms))
def get(self, key):
if not key.startswith(self.prefix):
key = join(self.prefix, key)
v = self.client.get(key, value_in_json=True)
if v:
return VMEntry(v)
return None
def put(self, obj: VMEntry):
self.client.put(obj.key, obj.value, value_in_json=True)
@contextmanager
def get_put(self, key) -> VMEntry:
# Updates object at key on exit
obj = self.get(key)
yield obj
if obj:
self.put(obj)

19
config.py Normal file
View file

@ -0,0 +1,19 @@
from etcd3_wrapper import Etcd3Wrapper
from common.classes import EnvironmentVariables
from common.host import HostPool
from common.request import RequestPool
from common.vm import VmPool
env_vars = EnvironmentVariables('/etc/ucloud/ucloud.conf')
etcd_wrapper_args = ()
etcd_wrapper_kwargs = {"host": env_vars.get("ETCD_URL")}
etcd_client = Etcd3Wrapper(*etcd_wrapper_args, **etcd_wrapper_kwargs)
host_pool = HostPool(etcd_client, env_vars.get('HOST_PREFIX'))
vm_pool = VmPool(etcd_client, env_vars.get('VM_PREFIX'))
request_pool = RequestPool(etcd_client, env_vars.get('REQUEST_PREFIX'))
running_vms = []

0
docs/__init__.py Normal file
View file

0
docs/source/__init__.py Normal file
View file

View file

@ -21,7 +21,6 @@ project = 'ucloud'
copyright = '2019, ungleich'
author = 'ungleich'
# -- General configuration ---------------------------------------------------
# Add any Sphinx extension module names here, as strings. They can be
@ -39,7 +38,6 @@ templates_path = ['_templates']
# This pattern also affects html_static_path and html_extra_path.
exclude_patterns = []
# -- Options for HTML output -------------------------------------------------
# The theme to use for HTML and HTML Help pages. See the documentation for
@ -50,4 +48,4 @@ html_theme = 'alabaster'
# Add any paths that contain custom static files (such as style sheets) here,
# relative to this directory. They are copied after the builtin static files,
# so a file named "default.css" will overwrite the builtin "default.css".
html_static_path = ['_static']
html_static_path = ['_static']

View file

@ -1,7 +1,7 @@
.. ucloud documentation master file, created by
sphinx-quickstart on Mon Nov 11 19:08:16 2019.
You can adapt this file completely to your liking, but it should at least
contain the root `toctree` directive.
sphinx-quickstart on Mon Nov 11 19:08:16 2019.
You can adapt this file completely to your liking, but it should at least
contain the root `toctree` directive.
Welcome to ucloud's documentation!
==================================
@ -15,7 +15,7 @@ Welcome to ucloud's documentation!
usage/usage-for-admins
usage/usage-for-users
usage/how-to-create-an-os-image-for-ucloud
misc/todo
Indices and tables
==================

View file

@ -36,6 +36,9 @@ Enable Edge Repos, Update and Upgrade
Install Dependencies
~~~~~~~~~~~~~~~~~~~~
.. note::
The installation and configuration of a production grade etcd cluster
is out of scope of this manual. So, we will install etcd with default configuration.
.. code-block:: sh
:linenos:

View file

@ -0,0 +1,7 @@
TODO
====
* Check for :code:`etcd3.exceptions.ConnectionFailedError` when calling some etcd operation to
avoid crashing whole application.
* Throw KeyError instead of returning None when some key is not found in etcd.
* Expose more details in ListUserFiles.

View file

@ -55,7 +55,28 @@ To start host we created earlier, execute the following command
Create OS Image
---------------
First, we need to upload the file.
Create ucloud-init ready OS image (Optional)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
This step is optional if you just want to test ucloud. However, sooner or later
you want to create OS images with ucloud-init to properly
contexualize VMs.
1. Start a VM with OS image on which you want to install ucloud-init
2. Execute the following command on the started VM
.. code-block:: sh
apk add git
git clone https://code.ungleich.ch/ucloud/ucloud-init.git
cd ucloud-init
sh ./install.sh
3. Congratulations. Your image is now ucloud-init ready.
Upload Sample OS Image
~~~~~~~~~~~~~~~~~~~~~~
Execute the following to upload the sample OS image file.
.. code-block:: sh
@ -63,7 +84,7 @@ First, we need to upload the file.
(cd /var/www/admin && wget http://[2a0a:e5c0:2:12:0:f0ff:fea9:c3d9]/alpine-untouched.qcow2)
Run File Scanner and Image Scanner
------------------------------------
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Currently, our uploaded file *alpine-untouched.qcow2* is not tracked by ucloud. We can only make
images from tracked files. So, we need to track the file by running File Scanner

0
filescanner/__init__.py Normal file
View file

View file

@ -1,14 +1,13 @@
import os
import glob
import os
import pathlib
import time
import hashlib
import subprocess as sp
from decouple import config
from etcd3_wrapper import Etcd3Wrapper
import time
from uuid import uuid4
from etcd3_wrapper import Etcd3Wrapper
from config import env_vars
def getxattr(file, attr):
@ -22,7 +21,7 @@ def getxattr(file, attr):
value = value.decode("utf-8")
except sp.CalledProcessError:
value = None
return value
@ -32,8 +31,8 @@ def setxattr(file, attr, value):
attr = "user." + attr
sp.check_output(['setfattr', file,
'--name', attr,
'--value', str(value)])
'--name', attr,
'--value', str(value)])
def sha512sum(file: str):
@ -68,12 +67,13 @@ except Exception as e:
print('Make sure you have getfattr and setfattr available')
exit(1)
def main():
BASE_DIR = config("BASE_DIR")
BASE_DIR = env_vars.get("BASE_DIR")
FILE_PREFIX = config("FILE_PREFIX")
FILE_PREFIX = env_vars.get("FILE_PREFIX")
etcd_client = Etcd3Wrapper(host=config("ETCD_URL"))
etcd_client = Etcd3Wrapper(host=env_vars.get("ETCD_URL"))
# Recursively Get All Files and Folder below BASE_DIR
files = glob.glob("{}/**".format(BASE_DIR), recursive=True)
@ -125,4 +125,4 @@ def main():
if __name__ == "__main__":
main()
main()

3
host/__init__.py Normal file
View file

@ -0,0 +1,3 @@
import logging
logger = logging.getLogger(__name__)

View file

@ -1,36 +0,0 @@
import logging
from etcd3_wrapper import Etcd3Wrapper
from ucloud_common.vm import VmPool
from ucloud_common.host import HostPool
from ucloud_common.request import RequestPool
from decouple import config
WITHOUT_CEPH = config("WITHOUT_CEPH", False, cast=bool)
logging.basicConfig(
level=logging.DEBUG,
filename="log.txt",
filemode="a",
format="%(asctime)s: %(levelname)s - %(message)s",
datefmt="%d-%b-%y %H:%M:%S",
)
etcd_wrapper_args = ()
etcd_wrapper_kwargs = {"host": config("ETCD_URL")}
etcd_client = Etcd3Wrapper(*etcd_wrapper_args, **etcd_wrapper_kwargs)
HOST_PREFIX = config("HOST_PREFIX")
NETWORK_PREFIX = config("NETWORK_PREFIX")
VM_PREFIX = config("VM_PREFIX")
REQUEST_PREFIX = config("REQUEST_PREFIX")
VM_DIR = config("VM_DIR")
IMAGE_DIR = config("IMAGE_DIR")
host_pool = HostPool(etcd_client, HOST_PREFIX)
vm_pool = VmPool(etcd_client, VM_PREFIX)
request_pool = RequestPool(etcd_client, REQUEST_PREFIX)
running_vms = []

View file

@ -1,31 +1,29 @@
import argparse
import time
import os
import sys
import virtualmachine
import multiprocessing as mp
import os
import time
from ucloud_common.host import HostEntry
from ucloud_common.request import RequestEntry, RequestType
from config import (vm_pool, host_pool, request_pool,
etcd_client, logging, running_vms,
etcd_wrapper_args, etcd_wrapper_kwargs,
REQUEST_PREFIX, HOST_PREFIX,
WITHOUT_CEPH, VM_DIR, HostPool)
from etcd3_wrapper import Etcd3Wrapper
import etcd3
from common.request import RequestEntry, RequestType
from config import (vm_pool, request_pool,
etcd_client, running_vms,
etcd_wrapper_args, etcd_wrapper_kwargs,
HostPool, env_vars)
from . import virtualmachine
from host import logger
def update_heartbeat(host):
client = Etcd3Wrapper(*etcd_wrapper_args, **etcd_wrapper_kwargs)
host_pool = HostPool(client, HOST_PREFIX)
host_pool = HostPool(client, env_vars.get('HOST_PREFIX'))
this_host = next(filter(lambda h: h.hostname == host, host_pool.hosts), None)
while True:
this_host.update_heartbeat()
host_pool.put(this_host)
time.sleep(10)
def maintenance(host):
# To capture vm running according to running_vms list
@ -65,31 +63,25 @@ def maintenance(host):
running_vms.remove(_vm)
def main():
argparser = argparse.ArgumentParser()
argparser.add_argument("hostname", help="Name of this host. e.g /v1/host/1")
args = argparser.parse_args()
def main(hostname):
assert env_vars.get('WITHOUT_CEPH') and os.path.isdir(env_vars.get('VM_DIR')), (
"You have set env_vars.get('WITHOUT_CEPH') to True. So, the vm directory mentioned"
" in .env file must exists. But, it don't.")
assert WITHOUT_CEPH and os.path.isdir(VM_DIR), (
"You have set WITHOUT_CEPH to True. So, the vm directory mentioned"
" in .env file must exists. But, it don't." )
mp.set_start_method('spawn')
heartbeat_updating_process = mp.Process(target=update_heartbeat, args=(args.hostname,))
heartbeat_updating_process = mp.Process(target=update_heartbeat, args=(hostname,))
host_pool = HostPool(etcd_client, HOST_PREFIX)
host = next(filter(lambda h: h.hostname == args.hostname, host_pool.hosts), None)
host_pool = HostPool(etcd_client, env_vars.get('HOST_PREFIX'))
host = next(filter(lambda h: h.hostname == hostname, host_pool.hosts), None)
assert host is not None, "No such host"
try:
heartbeat_updating_process.start()
except Exception as e:
logging.info("No Need To Go Further. Our heartbeat updating mechanism is not working")
logging.exception(e)
logger.info("No Need To Go Further. Our heartbeat updating mechanism is not working")
logger.exception(e)
exit(-1)
logging.info("%s Session Started %s", '*' * 5, '*' * 5)
logger.info("%s Session Started %s", '*' * 5, '*' * 5)
# It is seen that under heavy load, timeout event doesn't come
# in a predictive manner (which is intentional because we give
@ -99,22 +91,21 @@ def main():
# update the heart beat in a predictive manner we start Heart
# beat updating mechanism in separated thread
for events_iterator in [
etcd_client.get_prefix(REQUEST_PREFIX, value_in_json=True),
etcd_client.watch_prefix(REQUEST_PREFIX, timeout=10, value_in_json=True),
etcd_client.get_prefix(env_vars.get('REQUEST_PREFIX'), value_in_json=True),
etcd_client.watch_prefix(env_vars.get('REQUEST_PREFIX'), timeout=10, value_in_json=True),
]:
for request_event in events_iterator:
request_event = RequestEntry(request_event)
if request_event.type == "TIMEOUT":
logging.info("Timeout Event")
logger.info("Timeout Event")
maintenance(host)
continue
# If the event is directed toward me OR I am destination of a InitVMMigration
if (request_event.hostname == host.key or request_event.destination == host.key):
logging.debug("EVENT: %s", request_event)
if request_event.hostname == host.key or request_event.destination == host.key:
logger.debug("EVENT: %s", request_event)
request_pool.client.client.delete(request_event.key)
vm_entry = vm_pool.get(request_event.uuid)
@ -135,11 +126,14 @@ def main():
elif request_event.type == RequestType.TransferVM:
virtualmachine.transfer(request_event)
else:
logging.info("VM Entry missing")
logging.info("Running VMs %s", running_vms)
logger.info("VM Entry missing")
logger.info("Running VMs %s", running_vms)
if __name__ == "__main__":
main()
argparser = argparse.ArgumentParser()
argparser.add_argument("hostname", help="Name of this host. e.g /v1/host/1")
args = argparser.parse_args()
mp.set_start_method('spawn')
main(args.hostname)

View file

@ -15,24 +15,23 @@
import errno
import logging
import os
import subprocess
import re
import shutil
import socket
import subprocess
import tempfile
from . import qmp
LOG = logging.getLogger(__name__)
# Mapping host architecture to any additional architectures it can
# support which often includes its 32 bit cousin.
ADDITIONAL_ARCHES = {
"x86_64" : "i386",
"aarch64" : "armhf"
"x86_64": "i386",
"aarch64": "armhf"
}
def kvm_available(target_arch=None):
host_arch = os.uname()[4]
if target_arch and target_arch != host_arch:
@ -56,10 +55,12 @@ class QEMUMachineAddDeviceError(QEMUMachineError):
failures reported by the QEMU binary itself.
"""
class MonitorResponseError(qmp.QMPError):
"""
Represents erroneous QMP monitor reply
"""
def __init__(self, reply):
try:
desc = reply["error"]["desc"]
@ -108,7 +109,7 @@ class QEMUMachine(object):
self._qemu_log_file = None
self._popen = None
self._binary = binary
self._args = list(args) # Force copy args in case we modify them
self._args = list(args) # Force copy args in case we modify them
self._wrapper = wrapper
self._events = []
self._iolog = None
@ -453,10 +454,11 @@ class QEMUMachine(object):
See event_match for details.
timeout: QEMUMonitorProtocol.pull_event timeout parameter.
"""
def _match(event):
for name, match in events:
if (event['event'] == name and
self.event_match(event, match)):
self.event_match(event, match)):
return True
return False

View file

@ -8,10 +8,10 @@
# This work is licensed under the terms of the GNU GPL, version 2. See
# the COPYING file in the top-level directory.
import json
import errno
import socket
import json
import logging
import socket
class QMPError(Exception):
@ -31,7 +31,6 @@ class QMPTimeoutError(QMPError):
class QEMUMonitorProtocol(object):
#: Logger object for debugging messages
logger = logging.getLogger('QMP')
#: Socket's error class

View file

@ -6,30 +6,24 @@
import errno
import os
import random
import subprocess as sp
import tempfile
import time
import random
import ipaddress
from functools import wraps
from os.path import join
from typing import Union
from string import Template
from typing import Union
import bitmath
import sshtunnel
import qmp
from decouple import config
from ucloud_common.helpers import get_ipv4_address
from ucloud_common.request import RequestEntry, RequestType
from ucloud_common.vm import VMEntry, VMStatus
from config import (WITHOUT_CEPH, VM_PREFIX, VM_DIR, IMAGE_DIR,
NETWORK_PREFIX, etcd_client, logging,
request_pool, running_vms, vm_pool)
from common.helpers import get_ipv4_address
from common.request import RequestEntry, RequestType
from common.vm import VMEntry, VMStatus
from config import etcd_client, request_pool, running_vms, vm_pool, env_vars
from . import qmp
from host import logger
class VM:
def __init__(self, key, handle, vnc_socket_file):
@ -43,7 +37,7 @@ class VM:
def create_dev(script, _id, dev, ip=None):
command = [script, _id, dev]
if ip:
if ip:
command.append(ip)
try:
output = sp.check_output(command, stderr=sp.PIPE)
@ -57,13 +51,13 @@ def create_dev(script, _id, dev, ip=None):
def create_vxlan_br_tap(_id, _dev, ip=None):
network_script_base = os.path.join(os.path.dirname(os.path.dirname(__file__)), 'network')
vxlan = create_dev(script=os.path.join(network_script_base, 'create-vxlan.sh'),
_id=_id, dev=_dev)
_id=_id, dev=_dev)
if vxlan:
bridge = create_dev(script=os.path.join(network_script_base, 'create-bridge.sh'),
_id=_id, dev=vxlan, ip=ip)
if bridge:
tap = create_dev(script=os.path.join(network_script_base, 'create-tap.sh'),
_id=str(random.randint(1, 100000)), dev=bridge)
_id=str(random.randint(1, 100000)), dev=bridge)
if tap:
return tap
@ -77,35 +71,35 @@ def generate_mac(uaa=False, multicast=False, oui=None, separator=':', byte_fmt='
if oui:
if type(oui) == str:
oui = [int(chunk) for chunk in oui.split(separator)]
mac = oui + random_bytes(num=6-len(oui))
mac = oui + random_bytes(num=6 - len(oui))
else:
if multicast:
mac[0] |= 1 # set bit 0
mac[0] |= 1 # set bit 0
else:
mac[0] &= ~1 # clear bit 0
mac[0] &= ~1 # clear bit 0
if uaa:
mac[0] &= ~(1 << 1) # clear bit 1
mac[0] &= ~(1 << 1) # clear bit 1
else:
mac[0] |= 1 << 1 # set bit 1
mac[0] |= 1 << 1 # set bit 1
return separator.join(byte_fmt % b for b in mac)
def update_radvd_conf(etcd_client):
network_script_base = os.path.join(os.path.dirname(os.path.dirname(__file__)), 'network')
networks = {
net.value['ipv6']:net.value['id']
for net in etcd_client.get_prefix('/v1/network/', value_in_json=True)
if net.value.get('ipv6')
}
networks = {
net.value['ipv6']: net.value['id']
for net in etcd_client.get_prefix('/v1/network/', value_in_json=True)
if net.value.get('ipv6')
}
radvd_template = open(os.path.join(network_script_base,
'radvd-template.conf'), 'r').read()
radvd_template = Template(radvd_template)
content = [radvd_template.safe_substitute(bridge='br{}'.format(networks[net]),
prefix=net)
prefix=net)
for net in networks if networks.get(net)]
with open('/etc/radvd.conf', 'w') as radvd_conf:
radvd_conf.writelines(content)
@ -113,7 +107,7 @@ def update_radvd_conf(etcd_client):
def get_start_command_args(
vm_entry, vnc_sock_filename: str, migration=False, migration_port=4444,
vm_entry, vnc_sock_filename: str, migration=False, migration_port=4444,
):
threads_per_core = 1
vm_memory = int(bitmath.parse_string(vm_entry.specs["ram"]).to_MB())
@ -121,9 +115,9 @@ def get_start_command_args(
vm_uuid = vm_entry.uuid
vm_networks = vm_entry.network
if WITHOUT_CEPH:
if env_vars.get('WITHOUT_CEPH'):
command = "-drive file={},format=raw,if=virtio,cache=none".format(
os.path.join(VM_DIR, vm_uuid)
os.path.join(env_vars.get('VM_DIR'), vm_uuid)
)
else:
command = "-drive file=rbd:uservms/{},format=raw,if=virtio,cache=none".format(
@ -138,24 +132,24 @@ def get_start_command_args(
if migration:
command += " -incoming tcp:0:{}".format(migration_port)
tap = None
for network_and_mac in vm_networks:
network_name, mac = network_and_mac
_key = os.path.join(NETWORK_PREFIX, vm_entry.owner, network_name)
_key = os.path.join(env_vars.get('NETWORK_PREFIX'), vm_entry.owner, network_name)
network = etcd_client.get(_key, value_in_json=True)
network_type = network.value["type"]
network_id = str(network.value["id"])
network_ipv6 = network.value["ipv6"]
if network_type == "vxlan":
tap = create_vxlan_br_tap(network_id, config("VXLAN_PHY_DEV"), network_ipv6)
tap = create_vxlan_br_tap(network_id, env_vars.get("VXLAN_PHY_DEV"), network_ipv6)
update_radvd_conf(etcd_client)
command += " -netdev tap,id=vmnet{net_id},ifname={tap},script=no,downscript=no"\
" -device virtio-net-pci,netdev=vmnet{net_id},mac={mac}"\
.format(tap=tap, net_id=network_id, mac=mac)
command += " -netdev tap,id=vmnet{net_id},ifname={tap},script=no,downscript=no" \
" -device virtio-net-pci,netdev=vmnet{net_id},mac={mac}" \
.format(tap=tap, net_id=network_id, mac=mac)
return command.split(" ")
@ -189,15 +183,15 @@ def need_running_vm(func):
if vm:
try:
status = vm.handle.command("query-status")
logging.debug("VM Status Check - %s", status)
logger.debug("VM Status Check - %s", status)
except Exception as exception:
logging.info("%s failed - VM %s %s", func.__name__, e, exception)
logger.info("%s failed - VM %s %s", func.__name__, e, exception)
else:
return func(e)
return None
else:
logging.info("%s failed because VM %s is not running", func.__name__, e.key)
logger.info("%s failed because VM %s is not running", func.__name__, e.key)
return None
return wrapper
@ -206,18 +200,18 @@ def need_running_vm(func):
def create(vm_entry: VMEntry):
vm_hdd = int(bitmath.parse_string(vm_entry.specs["os-ssd"]).to_MB())
if WITHOUT_CEPH:
if env_vars.get('WITHOUT_CEPH'):
_command_to_create = [
"cp",
os.path.join(IMAGE_DIR, vm_entry.image_uuid),
os.path.join(VM_DIR, vm_entry.uuid),
os.path.join(env_vars.get('IMAGE_DIR'), vm_entry.image_uuid),
os.path.join(env_vars.get('VM_DIR'), vm_entry.uuid),
]
_command_to_extend = [
"qemu-img",
"resize",
"-f", "raw",
os.path.join(VM_DIR, vm_entry.uuid),
os.path.join(env_vars.get('VM_DIR'), vm_entry.uuid),
"{}M".format(vm_hdd),
]
else:
@ -240,22 +234,22 @@ def create(vm_entry: VMEntry):
sp.check_output(_command_to_create)
except sp.CalledProcessError as e:
if e.returncode == errno.EEXIST:
logging.debug("Image for vm %s exists", vm_entry.uuid)
logger.debug("Image for vm %s exists", vm_entry.uuid)
# File Already exists. No Problem Continue
return
# This exception catches all other exceptions
# i.e FileNotFound (BaseImage), pool Does Not Exists etc.
logging.exception(e)
logger.exception(e)
vm_entry.status = "ERROR"
else:
try:
sp.check_output(_command_to_extend)
except Exception as e:
logging.exception(e)
logger.exception(e)
else:
logging.info("New VM Created")
logger.info("New VM Created")
def start(vm_entry: VMEntry):
@ -263,7 +257,7 @@ def start(vm_entry: VMEntry):
# VM already running. No need to proceed further.
if _vm:
logging.info("VM %s already running", vm_entry.uuid)
logger.info("VM %s already running", vm_entry.uuid)
return
else:
create(vm_entry)
@ -282,19 +276,19 @@ def stop(vm_entry):
def delete(vm_entry):
logging.info("Deleting VM | %s", vm_entry)
logger.info("Deleting VM | %s", vm_entry)
stop(vm_entry)
path_without_protocol = vm_entry.path[vm_entry.path.find(":") + 1 :]
path_without_protocol = vm_entry.path[vm_entry.path.find(":") + 1:]
if WITHOUT_CEPH:
vm_deletion_command = ["rm", os.path.join(VM_DIR, vm_entry.uuid)]
if env_vars.get('WITHOUT_CEPH'):
vm_deletion_command = ["rm", os.path.join(env_vars.get('VM_DIR'), vm_entry.uuid)]
else:
vm_deletion_command = ["rbd", "rm", path_without_protocol]
try:
sp.check_output(vm_deletion_command)
except Exception as e:
logging.exception(e)
logger.exception(e)
else:
etcd_client.client.delete(vm_entry.key)
@ -307,20 +301,20 @@ def transfer(request_event):
_host, _port = request_event.parameters["host"], request_event.parameters["port"]
_uuid = request_event.uuid
_destination = request_event.destination_host_key
vm = get_vm(running_vms, join(VM_PREFIX, _uuid))
vm = get_vm(running_vms, join(env_vars.get('VM_PREFIX'), _uuid))
if vm:
tunnel = sshtunnel.SSHTunnelForwarder(
(_host, 22),
ssh_username=config("ssh_username"),
ssh_pkey=config("ssh_pkey"),
ssh_private_key_password=config("ssh_private_key_password"),
ssh_username=env_vars.get("ssh_username"),
ssh_pkey=env_vars.get("ssh_pkey"),
ssh_private_key_password=env_vars.get("ssh_private_key_password"),
remote_bind_address=("127.0.0.1", _port),
)
try:
tunnel.start()
except sshtunnel.BaseSSHTunnelForwarderError:
logging.exception("Couldn't establish connection to (%s, 22)", _host)
logger.exception("Couldn't establish connection to (%s, 22)", _host)
else:
vm.handle.command(
"migrate", uri="tcp:{}:{}".format(_host, tunnel.local_bind_port)
@ -356,7 +350,7 @@ def init_migration(vm_entry, destination_host_key):
if _vm:
# VM already running. No need to proceed further.
logging.info("%s Already running", _vm.key)
logger.info("%s Already running", _vm.key)
return
launch_vm(vm_entry, migration=True, migration_port=4444,
@ -364,13 +358,13 @@ def init_migration(vm_entry, destination_host_key):
def launch_vm(vm_entry, migration=False, migration_port=None, destination_host_key=None):
logging.info("Starting %s", vm_entry.key)
logger.info("Starting %s", vm_entry.key)
vm = create_vm_object(vm_entry, migration=migration, migration_port=migration_port)
try:
vm.handle.launch()
except Exception as e:
logging.exception(e)
logger.exception(e)
if migration:
# We don't care whether MachineError or any other error occurred
@ -392,6 +386,7 @@ def launch_vm(vm_entry, migration=False, migration_port=None, destination_host_k
parameters={"host": get_ipv4_address(), "port": 4444},
uuid=vm_entry.uuid,
destination_host_key=destination_host_key,
request_prefix=env_vars.get("REQUEST_PREFIX")
)
request_pool.put(r)
else:
@ -400,4 +395,3 @@ def launch_vm(vm_entry, migration=False, migration_port=None, destination_host_k
vm_entry.add_log("Started successfully")
vm_pool.put(vm_entry)

3
imagescanner/__init__.py Normal file
View file

@ -0,0 +1,3 @@
import logging
logger = logging.getLogger(__name__)

View file

@ -1,22 +0,0 @@
import logging
from etcd3_wrapper import Etcd3Wrapper
from decouple import config
BASE_PATH = config("BASE_DIR", "/var/www")
WITHOUT_CEPH = config("WITHOUT_CEPH", False, cast=bool)
ETCD_URL = config("ETCD_URL")
IMAGE_PREFIX = config("IMAGE_PREFIX")
IMAGE_STORE_PREFIX = config("IMAGE_STORE_PREFIX")
IMAGE_DIR = config("IMAGE_DIR")
logging.basicConfig(
level=logging.DEBUG,
filename="log.txt",
filemode="a",
format="%(asctime)s: %(levelname)s - %(message)s",
datefmt="%d-%b-%y %H:%M:%S",
)
client = Etcd3Wrapper(host=ETCD_URL)

View file

@ -1,11 +1,10 @@
import os
import json
import os
import subprocess
import sys
from config import (logging, client, IMAGE_DIR,
BASE_PATH, WITHOUT_CEPH,
IMAGE_PREFIX, IMAGE_STORE_PREFIX)
from config import etcd_client, env_vars
from imagescanner import logger
def qemu_img_type(path):
@ -13,7 +12,7 @@ def qemu_img_type(path):
try:
qemu_img_info = subprocess.check_output(qemu_img_info_command)
except Exception as e:
logging.exception(e)
logger.exception(e)
return None
else:
qemu_img_info = json.loads(qemu_img_info.decode("utf-8"))
@ -21,12 +20,12 @@ def qemu_img_type(path):
def main():
# If you are using WITHOUT_CEPH FLAG in .env
# then please make sure that IMAGE_DIR directory
# If you are using env_vars.get('WITHOUT_CEPH') FLAG in .env
# then please make sure that env_vars.get('IMAGE_DIR') directory
# exists otherwise this script would fail
if WITHOUT_CEPH and not os.path.isdir(IMAGE_DIR):
print("You have set WITHOUT_CEPH to True. So,"
"the {} must exists. But, it don't".format(IMAGE_DIR))
if env_vars.get('WITHOUT_CEPH') and not os.path.isdir(env_vars.get('IMAGE_DIR')):
print("You have set env_vars.get('WITHOUT_CEPH') to True. So,"
"the {} must exists. But, it don't".format(env_vars.get('IMAGE_DIR')))
sys.exit(1)
try:
@ -36,7 +35,7 @@ def main():
sys.exit(1)
# We want to get images entries that requests images to be created
images = client.get_prefix(IMAGE_PREFIX, value_in_json=True)
images = etcd_client.get_prefix(env_vars.get('IMAGE_PREFIX'), value_in_json=True)
images_to_be_created = list(filter(lambda im: im.value['status'] == 'TO_BE_CREATED', images))
for image in images_to_be_created:
@ -45,9 +44,9 @@ def main():
image_owner = image.value['owner']
image_filename = image.value['filename']
image_store_name = image.value['store_name']
image_full_path = os.path.join(BASE_PATH, image_owner, image_filename)
image_full_path = os.path.join(env_vars.get('BASE_DIR'), image_owner, image_filename)
image_stores = client.get_prefix(IMAGE_STORE_PREFIX, value_in_json=True)
image_stores = etcd_client.get_prefix(env_vars.get('IMAGE_STORE_PREFIX'), value_in_json=True)
user_image_store = next(filter(
lambda s, store_name=image_store_name: s.value["name"] == store_name,
image_stores
@ -56,27 +55,25 @@ def main():
image_store_pool = user_image_store.value['attributes']['pool']
except Exception as e:
logging.exception(e)
logger.exception(e)
else:
# At least our basic data is available
qemu_img_convert_command = ["qemu-img", "convert", "-f", "qcow2",
"-O", "raw", image_full_path, "image.raw"]
if WITHOUT_CEPH:
image_import_command = ["mv", "image.raw", os.path.join(IMAGE_DIR, image_uuid)]
if env_vars.get('WITHOUT_CEPH'):
image_import_command = ["mv", "image.raw", os.path.join(env_vars.get('IMAGE_DIR'), image_uuid)]
snapshot_creation_command = ["true"]
snapshot_protect_command = ["true"]
else:
image_import_command = ["rbd", "import", "image.raw",
"{}/{}".format(image_store_pool, image_uuid)]
snapshot_creation_command = ["rbd", "snap", "create",
"{}/{}@protected".format(image_store_pool, image_uuid)]
"{}/{}@protected".format(image_store_pool, image_uuid)]
snapshot_protect_command = ["rbd", "snap", "protect",
"{}/{}@protected".format(image_store_pool, image_uuid)]
# First check whether the image is qcow2
if qemu_img_type(image_full_path) == "qcow2":
@ -90,19 +87,18 @@ def main():
# Create and Protect Snapshot
subprocess.check_output(snapshot_creation_command)
subprocess.check_output(snapshot_protect_command)
except Exception as e:
logging.exception(e)
logger.exception(e)
else:
# Everything is successfully done
image.value["status"] = "CREATED"
client.put(image.key, json.dumps(image.value))
etcd_client.put(image.key, json.dumps(image.value))
else:
# The user provided image is either not found or of invalid format
image.value["status"] = "INVALID_IMAGE"
client.put(image.key, json.dumps(image.value))
etcd_client.put(image.key, json.dumps(image.value))
try:
os.remove("image.raw")
@ -111,4 +107,4 @@ def main():
if __name__ == "__main__":
main()
main()

0
metadata/__init__.py Normal file
View file

View file

@ -1,22 +0,0 @@
import logging
from etcd3_wrapper import Etcd3Wrapper
from decouple import config
from ucloud_common.vm import VmPool
logging.basicConfig(
level=logging.DEBUG,
filename="log.txt",
filemode="a",
format="%(asctime)s: %(levelname)s - %(message)s",
datefmt="%d-%b-%y %H:%M:%S",
)
VM_PREFIX = config("VM_PREFIX")
USER_PREFIX = config("USER_PREFIX")
etcd_client = Etcd3Wrapper(host=config("ETCD_URL"))
VM_POOL = VmPool(etcd_client, VM_PREFIX)

View file

@ -2,14 +2,15 @@ import os
from flask import Flask, request
from flask_restful import Resource, Api
from config import etcd_client, VM_POOL, USER_PREFIX
from config import etcd_client, env_vars, vm_pool
app = Flask(__name__)
api = Api(app)
def get_vm_entry(mac_addr):
return next(filter(lambda vm: mac_addr in list(zip(*vm.network))[1], VM_POOL.vms), None)
return next(filter(lambda vm: mac_addr in list(zip(*vm.network))[1], vm_pool.vms), None)
# https://stackoverflow.com/questions/37140846/how-to-convert-ipv6-link-local-address-to-mac-address-in-python
@ -43,8 +44,8 @@ class Root(Resource):
return {'message': 'Metadata for such VM does not exists.'}, 404
else:
# {user_prefix}/{realm}/{name}/key
etcd_key = os.path.join(USER_PREFIX, data.value['owner_realm'],
# {env_vars.get('USER_PREFIX')}/{realm}/{name}/key
etcd_key = os.path.join(env_vars.get('USER_PREFIX'), data.value['owner_realm'],
data.value['owner'], 'key')
etcd_entry = etcd_client.get_prefix(etcd_key, value_in_json=True)
user_personal_ssh_keys = [key.value for key in etcd_entry]
@ -81,5 +82,10 @@ class Root(Resource):
api.add_resource(Root, '/')
if __name__ == '__main__':
def main():
app.run(debug=True, host="::", port="80")
if __name__ == '__main__':
main()

0
network/__init__.py Normal file
View file

3
scheduler/__init__.py Normal file
View file

@ -0,0 +1,3 @@
import logging
logger = logging.getLogger(__name__)

View file

@ -1,25 +0,0 @@
import logging
from decouple import config
from etcd3_wrapper import Etcd3Wrapper
from ucloud_common.vm import VmPool
from ucloud_common.host import HostPool
from ucloud_common.request import RequestPool
logging.basicConfig(
level=logging.DEBUG,
filename="log.txt",
filemode="a",
format="%(asctime)s: %(levelname)s - %(message)s",
datefmt="%d-%b-%y %H:%M:%S",
)
vm_prefix = config("VM_PREFIX")
host_prefix = config("HOST_PREFIX")
request_prefix = config("REQUEST_PREFIX")
etcd_client = Etcd3Wrapper(host=config("ETCD_URL"))
vm_pool = VmPool(etcd_client, vm_prefix)
host_pool = HostPool(etcd_client, host_prefix)
request_pool = RequestPool(etcd_client, request_prefix)

View file

@ -1,21 +1,15 @@
import bitmath
from collections import Counter
from functools import reduce
from ucloud_common.vm import VmPool, VMStatus
from ucloud_common.host import HostPool, HostStatus
from ucloud_common.request import RequestEntry, RequestPool, RequestType
import bitmath
from decouple import config
from config import etcd_client as client
vm_pool = VmPool(client, config("VM_PREFIX"))
host_pool = HostPool(client, config("HOST_PREFIX"))
request_pool = RequestPool(client, config("REQUEST_PREFIX"))
from common.host import HostStatus
from common.request import RequestEntry, RequestType
from common.vm import VMStatus
from config import vm_pool, host_pool, request_pool, env_vars
def accumulated_specs(vms_specs):
def accumulated_specs(vms_specs):
if not vms_specs:
return {}
return reduce((lambda x, y: Counter(x) + Counter(y)), vms_specs)
@ -23,7 +17,7 @@ def accumulated_specs(vms_specs):
def remaining_resources(host_specs, vms_specs):
# Return remaining resources host_specs - vms
_vms_specs = Counter(vms_specs)
_remaining = Counter(host_specs)
@ -69,7 +63,7 @@ def get_suitable_host(vm_specs, hosts=None):
# Find out remaining resources after
# host_specs - already running vm_specs
remaining = remaining_resources(host.specs, running_vms_accumulated_specs)
# Find out remaining - new_vm_specs
remaining = remaining_resources(remaining, vm_specs)
@ -111,7 +105,8 @@ def assign_host(vm):
r = RequestEntry.from_scratch(type=RequestType.StartVM,
uuid=vm.uuid,
hostname=vm.hostname)
hostname=vm.hostname,
request_prefix=env_vars.get("REQUEST_PREFIX"))
request_pool.put(r)
vm.log.append("VM scheduled for starting")

View file

@ -4,28 +4,26 @@
# 2. Introduce a status endpoint of the scheduler -
# maybe expose a prometheus compatible output
import logging
from ucloud_common.request import RequestEntry, RequestType
from config import etcd_client as client
from config import (host_pool, request_pool, vm_pool, request_prefix)
from helper import (get_suitable_host, dead_host_mitigation, dead_host_detection,
assign_host, NoSuitableHostFound)
from common.request import RequestEntry, RequestType
from config import etcd_client
from config import host_pool, request_pool, vm_pool, env_vars
from .helper import (get_suitable_host, dead_host_mitigation, dead_host_detection,
assign_host, NoSuitableHostFound)
from scheduler import logger
def main():
logging.info("%s SESSION STARTED %s", '*' * 5, '*' * 5)
logger.info("%s SESSION STARTED %s", '*' * 5, '*' * 5)
pending_vms = []
for request_iterator in [
client.get_prefix(request_prefix, value_in_json=True),
client.watch_prefix(request_prefix, timeout=5, value_in_json=True),
etcd_client.get_prefix(env_vars.get('REQUEST_PREFIX'), value_in_json=True),
etcd_client.watch_prefix(env_vars.get('REQUEST_PREFIX'), timeout=5, value_in_json=True),
]:
for request_event in request_iterator:
request_entry = RequestEntry(request_event)
logging.debug("%s, %s", request_entry.key, request_entry.value)
logger.debug("%s, %s", request_entry.key, request_entry.value)
# Never Run time critical mechanism inside timeout
# mechanism because timeout mechanism only comes
@ -35,9 +33,9 @@ def main():
# Detect hosts that are dead and set their status
# to "DEAD", and their VMs' status to "KILLED"
logging.debug("TIMEOUT event occured")
logger.debug("TIMEOUT event occured")
dead_hosts = dead_host_detection()
logging.debug("Dead hosts: %s", dead_hosts)
logger.debug("Dead hosts: %s", dead_hosts)
dead_host_mitigation(dead_hosts)
# If there are VMs that weren't assigned a host
@ -49,15 +47,16 @@ def main():
pending_vm_entry = pending_vms.pop()
r = RequestEntry.from_scratch(type="ScheduleVM",
uuid=pending_vm_entry.uuid,
hostname=pending_vm_entry.hostname)
hostname=pending_vm_entry.hostname,
request_prefix=env_vars.get("REQUEST_PREFIX"))
request_pool.put(r)
elif request_entry.type == RequestType.ScheduleVM:
vm_entry = vm_pool.get(request_entry.uuid)
if vm_entry is None:
logging.info("Trying to act on {} but it is deleted".format(request_entry.uuid))
logger.info("Trying to act on {} but it is deleted".format(request_entry.uuid))
continue
client.client.delete(request_entry.key) # consume Request
etcd_client.client.delete(request_entry.key) # consume Request
# If the Request is about a VM which is labelled as "migration"
# and has a destination
@ -67,12 +66,13 @@ def main():
get_suitable_host(vm_specs=vm_entry.specs,
hosts=[host_pool.get(request_entry.destination)])
except NoSuitableHostFound:
logging.info("Requested destination host doesn't have enough capacity"
logger.info("Requested destination host doesn't have enough capacity"
"to hold %s", vm_entry.uuid)
else:
r = RequestEntry.from_scratch(type=RequestType.InitVMMigration,
uuid=request_entry.uuid,
destination=request_entry.destination)
destination=request_entry.destination,
request_prefix=env_vars.get("REQUEST_PREFIX"))
request_pool.put(r)
# If the Request is about a VM that just want to get started/created
@ -86,7 +86,7 @@ def main():
vm_pool.put(vm_entry)
pending_vms.append(vm_entry)
logging.info("No Resource Left. Emailing admin....")
logger.info("No Resource Left. Emailing admin....")
if __name__ == "__main__":

View file

View file

@ -1,13 +1,10 @@
import unittest
import sys
import json
import multiprocessing
import time
import sys
import unittest
from datetime import datetime
from os.path import dirname
BASE_DIR = dirname(dirname(__file__))
sys.path.insert(0, BASE_DIR)
@ -15,13 +12,12 @@ from main import (
accumulated_specs,
remaining_resources,
VmPool,
dead_host_detection,
dead_host_mitigation,
main,
)
from config import etcd_client
class TestFunctions(unittest.TestCase):
@classmethod
def setUpClass(cls):

View file

@ -1,24 +1,18 @@
import unittest
import sys
import json
import multiprocessing
import time
import unittest
from datetime import datetime
from os.path import dirname
BASE_DIR = dirname(dirname(__file__))
sys.path.insert(0, BASE_DIR)
from main import (
accumulated_specs,
remaining_resources,
VmPool,
dead_host_detection,
dead_host_mitigation,
main,
config
)
class TestDeadHostMechanism(unittest.TestCase):
def setUp(self):
self.client = config.etcd_client

View file

@ -1,16 +1,50 @@
import argparse
import subprocess as sp
arg_parser = argparse.ArgumentParser(prog='ucloud',
description='Open Source Cloud Management Software')
arg_parser.add_argument('component',
choices=['api', 'scheduler', 'host',
'filescanner','imagescanner',
'metadata'])
arg_parser.add_argument('component_args', nargs='*')
args = arg_parser.parse_args()
import multiprocessing as mp
import logging
try:
command = ['pipenv', 'run', 'python', 'main.py', *args.component_args]
sp.run(command, cwd=args.component)
except Exception as error:
print(error)
from os.path import join as join_path
if __name__ == "__main__":
arg_parser = argparse.ArgumentParser(prog='ucloud',
description='Open Source Cloud Management Software')
arg_parser.add_argument('component',
choices=['api', 'scheduler', 'host',
'filescanner', 'imagescanner',
'metadata'])
arg_parser.add_argument('component_args', nargs='*')
args = arg_parser.parse_args()
logging.basicConfig(
level=logging.DEBUG,
filename=join_path("logs.txt"),
filemode="a",
format="%(name)s %(asctime)s: %(levelname)s - %(message)s",
datefmt="%d-%b-%y %H:%M:%S",
)
if args.component == 'api':
from api.main import main
main()
elif args.component == 'host':
from host.main import main
hostname = args.component_args
mp.set_start_method('spawn')
main(*hostname)
elif args.component == 'scheduler':
from scheduler.main import main
main()
elif args.component == 'filescanner':
from filescanner.main import main
main()
elif args.component == 'imagescanner':
from imagescanner.main import main
main()
elif args.component == 'metadata':
from metadata.main import main
main()