def deploy_init_cluster(settings):
"""Provisions a GKE cluster, persistant disks, and any other prerequisites for deployment."""
print_separator("init-cluster")
# initialize the VM
node_name = get_node_name()
if not node_name:
raise Exception(
"Unable to retrieve node name. Was the cluster created successfully?"
)
if settings["DEPLOY_TO_PREFIX"] == "gcloud":
run("gcloud config set project %(GCLOUD_PROJECT)s" % settings)
# create private network so that dataproc jobs can connect to GKE cluster nodes
# based on: https://medium.com/@DazWilkin/gkes-cluster-ipv4-cidr-flag-69d25884a558
create_vpc(gcloud_project="%(GCLOUD_PROJECT)s" % settings,
network_name="%(GCLOUD_PROJECT)s-auto-vpc" % settings)
# create cluster
run(
" ".join([
"gcloud container clusters create %(CLUSTER_NAME)s",
"--project %(GCLOUD_PROJECT)s",
"--zone %(GCLOUD_ZONE)s",
"--machine-type %(CLUSTER_MACHINE_TYPE)s",
"--num-nodes 1",
#"--network %(GCLOUD_PROJECT)s-auto-vpc",
#"--local-ssd-count 1",
"--scopes",
"https://www.googleapis.com/auth/devstorage.read_write"
]) % settings,
verbose=False,
errors_to_ignore=["already exists"])
# create cluster nodes - breaking them up into node pools of several machines each.
# This way, the cluster can be scaled up and down when needed using the technique in
# https://github.com/mattsolo1/gnomadjs/blob/master/cluster/elasticsearch/Makefile#L23
#
i = 0
num_nodes_remaining_to_create = int(settings["CLUSTER_NUM_NODES"]) - 1
num_nodes_per_node_pool = int(settings["NUM_NODES_PER_NODE_POOL"])
while num_nodes_remaining_to_create > 0:
i += 1
run(
" ".join([
"gcloud beta container node-pools create %(CLUSTER_NAME)s-"
+ str(i),
"--cluster %(CLUSTER_NAME)s",
"--project %(GCLOUD_PROJECT)s",
"--zone %(GCLOUD_ZONE)s",
"--machine-type %(CLUSTER_MACHINE_TYPE)s",
"--num-nodes %s" % min(num_nodes_per_node_pool,
num_nodes_remaining_to_create),
#"--network %(GCLOUD_PROJECT)s-auto-vpc",
#"--local-ssd-count 1",
"--scopes",
"https://www.googleapis.com/auth/devstorage.read_write"
]) % settings,
verbose=False,
errors_to_ignore=["already exists"])
num_nodes_remaining_to_create -= num_nodes_per_node_pool
run(" ".join([
"gcloud container clusters get-credentials %(CLUSTER_NAME)s",
"--project %(GCLOUD_PROJECT)s",
"--zone %(GCLOUD_ZONE)s",
]) % settings)
# create elasticsearch disks
run(" ".join([
"kubectl apply -f %(DEPLOYMENT_TEMP_DIR)s/deploy/kubernetes/elasticsearch/ssd-storage-class.yaml"
% settings,
]))
#run(" ".join([
# "gcloud compute disks create %(CLUSTER_NAME)s-elasticsearch-disk-0 --type=pd-ssd --zone=us-central1-b --size=%(ELASTICSEARCH_DISK_SIZE)sGi" % settings,
#]), errors_to_ignore=["already exists"])
#run(" ".join([
# "kubectl apply -f %(DEPLOYMENT_TEMP_DIR)s/deploy/kubernetes/elasticsearch/es-persistent-volume.yaml" % settings,
#]))
# if cluster was already created previously, update it's size to match CLUSTER_NUM_NODES
#run(" ".join([
# "gcloud container clusters resize %(CLUSTER_NAME)s --size %(CLUSTER_NUM_NODES)s" % settings,
#]), is_interactive=True)
# create persistent disks
for label in (
"postgres",
): # "mongo"): # , "elasticsearch-sharded"): # "elasticsearch"
run(" ".join([
"gcloud compute disks create",
"--zone %(GCLOUD_ZONE)s",
"--size %(" + label.upper().replace("-", "_") + "_DISK_SIZE)s",
"%(CLUSTER_NAME)s-" + label + "-disk",
]) % settings,
verbose=True,
errors_to_ignore=["already exists"])
elif settings["DEPLOY_TO"] == "kube-solo":
run("mkdir -p %(POSTGRES_DBPATH)s" % settings)
run("mkdir -p %(MONGO_DBPATH)s" % settings)
run("mkdir -p %(ELASTICSEARCH_DBPATH)s" % settings)
elif settings["DEPLOY_TO"] == "minikube":
# fix time sync issues on MacOSX which could interfere with token auth (https://github.com/kubernetes/minikube/issues/1378)
run("minikube ssh -- docker run -i --rm --privileged --pid=host debian nsenter -t 1 -m -u -n -i date -u $(date -u +%m%d%H%M%Y)"
)
else:
raise ValueError("Unexpected DEPLOY_TO_PREFIX: %(DEPLOY_TO_PREFIX)s" %
settings)
# set VM settings required for elasticsearch
if settings["DEPLOY_TO"] == "minikube":
run("minikube ssh 'sudo /sbin/sysctl -w vm.max_map_count=262144'" %
locals())
elif settings["DEPLOY_TO"] == "kube-solo":
run("corectl ssh %(node_name)s \"sudo /sbin/sysctl -w vm.max_map_count=262144\""
% locals())
#else:
# run(" ".join([
# "gcloud compute ssh "+node_name,
# "--zone %(GCLOUD_ZONE)s",
# "--command \"sudo /sbin/sysctl -w vm.max_map_count=262144\""
# ]) % settings)
# print cluster info
run("kubectl cluster-info", verbose=True)
def deploy_init_cluster(settings):
"""Provisions a GKE cluster, persistant disks, and any other prerequisites for deployment."""
print_separator("init-cluster")
if settings["DEPLOY_TO_PREFIX"] == "gcloud":
run("gcloud config set project %(GCLOUD_PROJECT)s" % settings)
# create private network so that dataproc jobs can connect to GKE cluster nodes
# based on: https://medium.com/@DazWilkin/gkes-cluster-ipv4-cidr-flag-69d25884a558
create_vpc(gcloud_project="%(GCLOUD_PROJECT)s" % settings, network_name="%(GCLOUD_PROJECT)s-auto-vpc" % settings)
# create cluster
run(" ".join([
"gcloud container clusters create %(CLUSTER_NAME)s",
"--project %(GCLOUD_PROJECT)s",
"--zone %(GCLOUD_ZONE)s",
"--machine-type %(CLUSTER_MACHINE_TYPE)s",
"--num-nodes %(CLUSTER_NUM_NODES)s",
#"--network %(GCLOUD_PROJECT)s-auto-vpc",
#"--local-ssd-count 1",
"--scopes", "https://www.googleapis.com/auth/devstorage.read_write"
]) % settings, verbose=False, errors_to_ignore=["already exists"])
run(" ".join([
"gcloud container clusters get-credentials %(CLUSTER_NAME)s",
"--project %(GCLOUD_PROJECT)s",
"--zone %(GCLOUD_ZONE)s",
]) % settings)
# create disks
run(" ".join([
"kubectl apply -f %(DEPLOYMENT_TEMP_DIR)s/deploy/kubernetes/elasticsearch/ssd-storage-class.yaml" % settings,
]))
#run(" ".join([
# "gcloud compute disks create %(CLUSTER_NAME)s-elasticsearch-disk-0 --type=pd-ssd --zone=us-central1-b --size=%(ELASTICSEARCH_DISK_SIZE)sGi" % settings,
#]), errors_to_ignore=["already exists"])
#run(" ".join([
# "kubectl apply -f %(DEPLOYMENT_TEMP_DIR)s/deploy/kubernetes/elasticsearch/es-persistent-volume.yaml" % settings,
#]))
# if cluster was already created previously, update it's size to match CLUSTER_NUM_NODES
#run(" ".join([
# "gcloud container clusters resize %(CLUSTER_NAME)s --size %(CLUSTER_NUM_NODES)s" % settings,
#]), is_interactive=True)
# create persistent disks
for label in ("postgres",): # "mongo"): # , "elasticsearch-sharded"): # "elasticsearch"
run(" ".join([
"gcloud compute disks create",
"--zone %(GCLOUD_ZONE)s",
"--size %("+label.upper().replace("-", "_")+"_DISK_SIZE)s",
"%(CLUSTER_NAME)s-"+label+"-disk",
]) % settings, verbose=True, errors_to_ignore=["already exists"])
else:
run("mkdir -p %(POSTGRES_DBPATH)s" % settings)
run("mkdir -p %(MONGO_DBPATH)s" % settings)
run("mkdir -p %(ELASTICSEARCH_DBPATH)s" % settings)
# initialize the VM
node_name = get_node_name()
if not node_name:
raise Exception("Unable to retrieve node name. Was the cluster created successfully?")
# set VM settings required for elasticsearch
if settings["DEPLOY_TO"] == "local":
run("corectl ssh %(node_name)s \"sudo /sbin/sysctl -w vm.max_map_count=262144\"" % locals())
#else:
# run(" ".join([
# "gcloud compute ssh "+node_name,
# "--zone %(GCLOUD_ZONE)s",
# "--command \"sudo /sbin/sysctl -w vm.max_map_count=262144\""
# ]) % settings)
# deploy ConfigMap file so that settings key/values can be added as environment variables in each of the pods
#with open(os.path.join(output_dir, "deploy/kubernetes/all-settings.properties"), "w") as f:
# for key, value in settings.items():
# f.write("%s=%s\n" % (key, value))
#run("kubectl delete configmap all-settings")
#run("kubectl create configmap all-settings --from-file=deploy/kubernetes/all-settings.properties")
#run("kubectl get configmaps all-settings -o yaml")
run("kubectl cluster-info", verbose=True)