def __init__(self, clusterName, provisioner, zone=None):
self.clusterName = clusterName
self.zone = zone
self.provisionerType = provisioner
if provisioner == 'aws':
from toil.provisioners.aws.awsProvisioner import AWSProvisioner
self.provisioner = AWSProvisioner
elif provisioner == 'gce':
from toil.provisioners.gceProvisioner import GCEProvisioner
self.provisioner = GCEProvisioner()
else:
assert False, "Invalid provisioner '%s'" % provisioner
def cluster_factory(provisioner, clusterName=None, clusterType='mesos', zone=None, nodeStorage=50, nodeStorageOverrides=None, sseKey=None):
"""
Find and instantiate the appropriate provisioner instance to make clusters
in the given cloud.
Raises ClusterTypeNotSupportedException if the given provisioner does not
implement clusters of the given type.
:param provisioner: The cloud type of the cluster.
:param clusterName: The name of the cluster.
:param clusterType: The type of cluster: 'mesos' or 'kubernetes'.
:param zone: The cloud zone
:return: A cluster object for the the cloud type.
"""
if provisioner == 'aws':
try:
from toil.provisioners.aws.awsProvisioner import AWSProvisioner
except ImportError:
logger.error('The aws extra must be installed to use this provisioner')
raise
return AWSProvisioner(clusterName, clusterType, zone, nodeStorage, nodeStorageOverrides, sseKey)
elif provisioner == 'gce':
try:
from toil.provisioners.gceProvisioner import GCEProvisioner
except ImportError:
logger.error('The google extra must be installed to use this provisioner')
raise
return GCEProvisioner(clusterName, clusterType, zone, nodeStorage, nodeStorageOverrides, sseKey)
else:
raise RuntimeError("Invalid provisioner '%s'" % provisioner)
def cluster_factory(provisioner,
clusterName=None,
zone=None,
nodeStorage=50,
nodeStorageOverrides=None,
sseKey=None):
"""
:param clusterName: The name of the cluster.
:param provisioner: The cloud type of the cluster.
:param zone: The cloud zone
:return: A cluster object for the the cloud type.
"""
if provisioner == 'aws':
try:
from toil.provisioners.aws.awsProvisioner import AWSProvisioner
except ImportError:
logger.error(
'The aws extra must be installed to use this provisioner')
raise
return AWSProvisioner(clusterName, zone, nodeStorage,
nodeStorageOverrides, sseKey)
elif provisioner == 'gce':
try:
from toil.provisioners.gceProvisioner import GCEProvisioner
except ImportError:
logger.error(
'The google extra must be installed to use this provisioner')
raise
return GCEProvisioner(clusterName, zone, nodeStorage,
nodeStorageOverrides, sseKey)
else:
raise RuntimeError("Invalid provisioner '%s'" % provisioner)
class Cluster(object):
def __init__(self, clusterName, provisioner, zone=None):
self.clusterName = clusterName
self.zone = zone
self.provisionerType = provisioner
if provisioner == 'aws':
from toil.provisioners.aws.awsProvisioner import AWSProvisioner
self.provisioner = AWSProvisioner
elif provisioner == 'gce':
from toil.provisioners.gceProvisioner import GCEProvisioner
self.provisioner = GCEProvisioner()
else:
assert False, "Invalid provisioner '%s'" % provisioner
def sshCluster(self, args, **kwargs):
self.provisioner.sshLeader(self.clusterName, args, self.zone, **kwargs)
def rsyncCluster(self, args, **kwargs):
self.provisioner.rsyncLeader(self.clusterName, args, self.zone,
**kwargs)
# Why was this added? It is undocumented and calling private methods.
#ctx = self.provisioner._buildContext(self.clusterName, zone=self.zone)
#instances = self.provisioner._getNodesInCluster(ctx, self.clusterName, both=True)
#leader = self.provisioner._getLeader(self.clusterName, zone=self.zone)
#workers = [i for i in instances if i.public_dns_name != leader.public_dns_name]
#for instance in workers:
# self.provisioner._waitForNode(instance, 'toil_worker')
# self.provisioner._coreRsync(instance.public_dns_name, args, applianceName='toil_worker', **kwargs)
def destroyCluster(self):
self.provisioner.destroyCluster(self.clusterName, self.zone)
def clusterFactory(provisioner,
clusterName=None,
zone=None,
nodeStorage=50,
sseKey=None):
"""
:param clusterName: The name of the cluster.
:param provisioner: The cloud type of the cluster.
:param zone: The cloud zone
:return: A cluster object for the the cloud type.
"""
if provisioner == 'aws':
try:
from toil.provisioners.aws.awsProvisioner import AWSProvisioner
from toil.lib.ec2Credentials import enable_metadata_credential_caching
except ImportError:
logger.error(
'The aws extra must be installed to use this provisioner')
raise
enable_metadata_credential_caching() # monkey patch for AWS
return AWSProvisioner(clusterName, zone, nodeStorage, sseKey)
elif provisioner == 'gce':
try:
from toil.provisioners.gceProvisioner import GCEProvisioner
except ImportError:
logger.error(
'The google extra must be installed to use this provisioner')
raise
return GCEProvisioner(clusterName, zone, nodeStorage, sseKey)
elif provisioner == 'azure':
try:
from toil.provisioners.azure.azureProvisioner import AzureProvisioner
except ImportError:
logger.error(
'The azure extra must be installed to use this provisioner')
raise
return AzureProvisioner(clusterName, zone, nodeStorage)
else:
raise RuntimeError("Invalid provisioner '%s'" % provisioner)
def main():
parser = getBasicOptionParser()
parser = addBasicProvisionerOptions(parser)
parser.add_argument("--leaderNodeType", dest="leaderNodeType", required=True,
help="Non-preemptable node type to use for the cluster leader.")
parser.add_argument("--keyPairName", dest='keyPairName', required=True,
help="The name of the AWS or ssh key pair to include on the instance")
parser.add_argument("--boto", dest='botoPath',
help="The path to the boto credentials directory. This is transferred to all "
" nodes in order to access the AWS jobStore from non-AWS instances.")
parser.add_argument("-t", "--tag", metavar='NAME=VALUE', dest='tags', default=[], action='append',
help="Tags are added to the AWS cluster for this node and all of its "
"children. Tags are of the form:\n"
" -t key1=value1 --tag key2=value2\n"
"Multiple tags are allowed and each tag needs its own flag. By "
"default the cluster is tagged with "
" {\n"
" \"Name\": clusterName,\n"
" \"Owner\": IAM username\n"
" }. ")
parser.add_argument("--vpcSubnet",
help="VPC subnet ID to launch cluster in. Uses default subnet if not specified. "
"This subnet needs to have auto assign IPs turned on.")
parser.add_argument("--nodeTypes", dest='nodeTypes', default=None, type=str,
help="Comma-separated list of node types to create while launching the leader. The "
"syntax for each node type depends on the "
"provisioner used. For the aws provisioner this is the name of an "
"EC2 instance type followed by a colon and the price in dollar to "
"bid for a spot instance, for example 'c3.8xlarge:0.42'. Must also provide "
"the --workers argument to specify how many workers of each node type to create")
parser.add_argument("-w", "--workers", dest='workers', default=None, type=str,
help="Comma-separated list of the number of workers of each node type to launch "
"alongside the leader when the "
"cluster is created. This can be useful if running toil without "
"auto-scaling but with need of more hardware support")
parser.add_argument("--leaderStorage", dest='leaderStorage', type=int, default=50,
help="Specify the size (in gigabytes) of the root volume for the leader instance. "
"This is an EBS volume.")
parser.add_argument("--nodeStorage", dest='nodeStorage', type=int, default=50,
help="Specify the size (in gigabytes) of the root volume for any worker instances "
"created when using the -w flag. This is an EBS volume.")
config = parseBasicOptions(parser)
tagsDict = None if config.tags is None else createTagsDict(config.tags)
spotBids = []
nodeTypes = []
preemptableNodeTypes = []
numNodes = []
numPreemptableNodes = []
leaderSpotBid = None
if config.provisioner == 'aws':
logger.info('Using aws provisioner.')
try:
from toil.provisioners.aws.awsProvisioner import AWSProvisioner
except ImportError:
logger.error('The aws extra must be installed to use this provisioner')
raise
provisioner = AWSProvisioner()
elif config.provisioner == 'gce':
logger.info('Using a gce provisioner.')
try:
from toil.provisioners.gceProvisioner import GCEProvisioner
except ImportError:
logger.error('The google extra must be installed to use this provisioner')
raise
provisioner = GCEProvisioner()
else:
assert False
#Parse leader node type and spot bid
parsedBid = config.leaderNodeType.split(':', 1)
if len(config.leaderNodeType) != len(parsedBid[0]):
leaderSpotBid = float(parsedBid[1])
config.leaderNodeType = parsedBid[0]
if (config.nodeTypes or config.workers) and not (config.nodeTypes and config.workers):
raise RuntimeError("The --nodeTypes and --workers options must be specified together,")
if config.nodeTypes:
nodeTypesList = config.nodeTypes.split(",")
numWorkersList = config.workers.split(",")
if not len(nodeTypesList) == len(numWorkersList):
raise RuntimeError("List of node types must be same length as list of numbers of workers.")
for nodeTypeStr, num in zip(nodeTypesList, numWorkersList):
parsedBid = nodeTypeStr.split(':', 1)
if len(nodeTypeStr) != len(parsedBid[0]):
#Is a preemptable node
preemptableNodeTypes.append(parsedBid[0])
spotBids.append(float(parsedBid[1]))
numPreemptableNodes.append(int(num))
else:
nodeTypes.append(nodeTypeStr)
numNodes.append(int(num))
provisioner.launchCluster(leaderNodeType=config.leaderNodeType,
leaderSpotBid=leaderSpotBid,
nodeTypes=nodeTypes,
preemptableNodeTypes=preemptableNodeTypes,
numWorkers=numNodes,
numPreemptableWorkers = numPreemptableNodes,
keyName=config.keyPairName,
botoPath=config.botoPath,
clusterName=config.clusterName,
spotBids=spotBids,
userTags=tagsDict,
zone=config.zone,
leaderStorage=config.leaderStorage,
nodeStorage=config.nodeStorage,
vpcSubnet=config.vpcSubnet)
def main():
parser = getBasicOptionParser()
parser = addBasicProvisionerOptions(parser)
parser.add_argument(
"--leaderNodeType",
dest="leaderNodeType",
required=True,
help="Non-preemptable node type to use for the cluster leader.")
parser.add_argument(
"--keyPairName",
dest='keyPairName',
required=True,
help="On AWS, the name of the AWS key pair to include on the instance."
" On Google/GCE, this is the ssh key pair."
" On Azure, this will be used as the owner tag.")
parser.add_argument(
"--publicKeyFile",
dest='publicKeyFile',
default="~/.ssh/id_rsa.pub",
help="On Azure, the file"
" containing the key pairs (the first key pair will be used).")
parser.add_argument(
"--boto",
dest='botoPath',
help="The path to the boto credentials directory. This is transferred "
"to all nodes in order to access the AWS jobStore from non-AWS instances."
)
parser.add_argument(
"-t",
"--tag",
metavar='NAME=VALUE',
dest='tags',
default=[],
action='append',
help="Tags are added to the AWS cluster for this node and all of its "
"children. Tags are of the form:\n"
" -t key1=value1 --tag key2=value2\n"
"Multiple tags are allowed and each tag needs its own flag. By "
"default the cluster is tagged with "
" {\n"
" \"Name\": clusterName,\n"
" \"Owner\": IAM username\n"
" }. ")
parser.add_argument(
"--vpcSubnet",
help="VPC subnet ID to launch cluster in. Uses default subnet if not "
"specified. This subnet needs to have auto assign IPs turned on.")
parser.add_argument(
"--nodeTypes",
dest='nodeTypes',
default=None,
type=str,
help="Comma-separated list of node types to create while launching the "
"leader. The syntax for each node type depends on the provisioner "
"used. For the aws provisioner this is the name of an EC2 instance "
"type followed by a colon and the price in dollar to bid for a spot "
"instance, for example 'c3.8xlarge:0.42'. Must also provide the "
"--workers argument to specify how many workers of each node type "
"to create.")
parser.add_argument(
"-w",
"--workers",
dest='workers',
default=None,
type=str,
help=
"Comma-separated list of the number of workers of each node type to "
"launch alongside the leader when the cluster is created. This can be "
"useful if running toil without auto-scaling but with need of more "
"hardware support")
parser.add_argument(
"--leaderStorage",
dest='leaderStorage',
type=int,
default=50,
help="Specify the size (in gigabytes) of the root volume for the leader "
"instance. This is an EBS volume.")
parser.add_argument(
"--nodeStorage",
dest='nodeStorage',
type=int,
default=50,
help="Specify the size (in gigabytes) of the root volume for any worker "
"instances created when using the -w flag. This is an EBS volume.")
parser.add_argument(
'--forceDockerAppliance',
dest='forceDockerAppliance',
action='store_true',
default=False,
help=
"Disables sanity checking the existence of the docker image specified "
"by TOIL_APPLIANCE_SELF, which Toil uses to provision mesos for "
"autoscaling.")
parser.add_argument(
"--azureStorageCredentials",
dest='azureStorageCredentials',
type=str,
default=credential_file_path,
help=
"The location of the file containing the Azure storage credentials. If not specified,"
" the default file is used with Azure provisioning. Use 'None' to disable"
" the transfer of credentials.")
config = parseBasicOptions(parser)
tagsDict = None if config.tags is None else createTagsDict(config.tags)
# checks the validity of TOIL_APPLIANCE_SELF before proceeding
checkToilApplianceSelf = applianceSelf(
forceDockerAppliance=config.forceDockerAppliance)
spotBids = []
nodeTypes = []
preemptableNodeTypes = []
numNodes = []
numPreemptableNodes = []
leaderSpotBid = None
if config.provisioner == 'aws':
logger.info('Using aws provisioner.')
try:
from toil.provisioners.aws.awsProvisioner import AWSProvisioner
except ImportError:
logger.error(
'The aws extra must be installed to use this provisioner')
raise
provisioner = AWSProvisioner()
elif config.provisioner == 'azure':
try:
from toil.provisioners.azure.azureProvisioner import AzureProvisioner
except ImportError:
raise RuntimeError(
'The aws extra must be installed to use this provisioner')
provisioner = AzureProvisioner()
elif config.provisioner == 'gce':
logger.info('Using a gce provisioner.')
try:
from toil.provisioners.gceProvisioner import GCEProvisioner
except ImportError:
logger.error(
'The google extra must be installed to use this provisioner')
raise
provisioner = GCEProvisioner()
else:
assert False
#Parse leader node type and spot bid
parsedBid = config.leaderNodeType.split(':', 1)
if len(config.leaderNodeType) != len(parsedBid[0]):
leaderSpotBid = float(parsedBid[1])
config.leaderNodeType = parsedBid[0]
if (config.nodeTypes
or config.workers) and not (config.nodeTypes and config.workers):
raise RuntimeError(
"The --nodeTypes and --workers options must be specified together,"
)
if config.nodeTypes:
nodeTypesList = config.nodeTypes.split(",")
numWorkersList = config.workers.split(",")
if not len(nodeTypesList) == len(numWorkersList):
raise RuntimeError(
"List of node types must be the same length as the list of workers."
)
for nodeTypeStr, num in zip(nodeTypesList, numWorkersList):
parsedBid = nodeTypeStr.split(':', 1)
if len(nodeTypeStr) != len(parsedBid[0]):
#Is a preemptable node
preemptableNodeTypes.append(parsedBid[0])
spotBids.append(float(parsedBid[1]))
numPreemptableNodes.append(int(num))
else:
nodeTypes.append(nodeTypeStr)
numNodes.append(int(num))
provisioner.launchCluster(
leaderNodeType=config.leaderNodeType,
leaderSpotBid=leaderSpotBid,
nodeTypes=nodeTypes,
preemptableNodeTypes=preemptableNodeTypes,
numWorkers=numNodes,
numPreemptableWorkers=numPreemptableNodes,
keyName=config.keyPairName,
botoPath=config.botoPath,
clusterName=config.clusterName,
spotBids=spotBids,
userTags=tagsDict,
zone=config.zone,
leaderStorage=config.leaderStorage,
nodeStorage=config.nodeStorage,
vpcSubnet=config.vpcSubnet,
publicKeyFile=config.publicKeyFile,
azureStorageCredentials=config.azureStorageCredentials)