def handle_GET(self):
path = self.request['path']
_, stack_id = os.path.split(path)
stack_config = stacks.get_stack_config(self.splunk, stack_id)
kubernetes = clusters.create_client(self.splunk,
stack_config["cluster"])
core_api = kuberneteslib.CoreV1Api(kubernetes)
instances = {}
#FIXME: label_selector does exclude monitoring-console for now
# Can be removed as soon as the stack_id label has been removed
pods = core_api.list_namespaced_pod(
namespace=stack_config["namespace"],
label_selector=
"app=saas,stack_id=%s,app.kubernetes.io/name!=monitoring-console" %
stack_id,
).items
for pod in pods:
name = pod.metadata.name
match = re.match('.*%s-(.+)-([0-9]+)$' % stack_id, name)
if match:
number = int(match.group(2)) + 1
role = match.group(1)
else:
number = None
match = re.match('.*%s-(.+)$' % stack_id, name)
if match:
role = match.group(1)
else:
role = None
reasons = set()
if pod.status:
status = pod.status.phase
is_ready = None
if pod.status.conditions:
for condition in pod.status.conditions:
if condition.status == "False":
if condition.reason:
reasons.add(condition.reason)
if condition.type == "Ready":
is_ready = condition.status == "True"
if status == "Running":
if is_ready:
status = "ready"
else:
status = "running"
else:
status = "unknown"
instances[name] = {
"role": role.lower(),
"number": number,
"status": status.lower(),
"reasons": list(reasons),
}
self.send_entries(instances.values())
def update_deployment(splunk, kubernetes, stack_id):
stack_config = stacks.get_stack_config(splunk, stack_id)
cluster_name = stack_config["cluster"]
kubernetes = clusters.create_client(splunk, cluster_name)
cluster_config = clusters.get_cluster(splunk, cluster_name)
if stack_config["deployment_type"] == "distributed":
indexer_cluster.update(splunk, kubernetes, stack_id, stack_config)
search_head_cluster.update(splunk, kubernetes, stack_id, stack_config)
indexer_cluster.wait_until_ready(splunk, kubernetes, stack_id, stack_config)
search_head_cluster.wait_until_ready(splunk, kubernetes, stack_id, stack_config)
def update_deployment(splunk, kubernetes, stack_id):
core_api = kuberneteslib.CoreV1Api(kubernetes)
stack_config = stacks.get_stack_config(splunk, stack_id)
cluster_name = stack_config["cluster"]
kubernetes = clusters.create_client(splunk, cluster_name)
cluster_config = clusters.get_cluster(splunk, cluster_name)
if stack_config["deployment_type"] == "distributed":
indexer_cluster.update(splunk, kubernetes, stack_id, stack_config)
search_head_cluster.update(splunk, kubernetes, stack_id, stack_config)
indexer_cluster.wait_until_ready(splunk, kubernetes, stack_id,
stack_config)
search_head_cluster.wait_until_ready(splunk, kubernetes, stack_id,
stack_config)
create_load_balancers(core_api, stack_id, stack_config)
verify_load_balancers_completed(core_api, stack_id, stack_config)
def down(splunk, stack_id, force=False):
stacks.update_config(splunk, stack_id, {
"status": stacks.DELETING,
"deleted_time": time.time(),
})
stack_config = stacks.get_stack_config(splunk, stack_id)
cluster_name = stack_config["cluster"]
cluster_config = clusters.get_cluster(splunk, cluster_name)
api_client = clusters.create_client(splunk, cluster_name)
core_api = kuberneteslib.CoreV1Api(api_client)
custom_objects_api = kuberneteslib.CustomObjectsApi(api_client)
try:
services.delete_all_load_balancers(core_api, stack_id, stack_config["namespace"])
stack_deployment.delete_objects(api_client, stack_id, stack_config, cluster_config)
except:
if not force:
raise
stacks.update_config(splunk, stack_id, {
"status": stacks.DELETED,
})
def stop_case(splunk, test_id, case_id, case):
if "stack_id" not in case:
return
stack_id = case["stack_id"]
result = splunk.get("saas/stack/%s" % stack_id)
logging.debug("get stack result: %s" % result)
response = json.loads(result.body.read())["entry"][0]["content"]
logging.debug("get stack response: %s" % response)
stack_status = response["status"]
if stack_status == stacks.DELETING:
raise errors.RetryOperation("still in status %s" % stacks.DELETING)
elif stack_status == stacks.DELETED:
pass
elif stack_status != stacks.DELETED:
result = splunk.delete("saas/stack/%s" % stack_id)
response = json.loads(result.body.read())["entry"][0]["content"]
logging.debug("delete stack result: %s" % response)
raise errors.RetryOperation("issued deletion of stack %s" % (stack_id))
stack_config = stacks.get_stack_config(splunk, stack_id)
kube_client = clusters.create_client(splunk, stack_config["cluster"])
apps_api = kubernetes.AppsV1Api(kube_client)
datagen_deployments = apps_api.list_namespaced_deployment(
namespace=stack_config["namespace"],
label_selector="app=datagen,test=%s" % test_id,
).items
for deployment in datagen_deployments:
apps_api.delete_namespaced_deployment(
name=deployment.metadata.name,
namespace=stack_config["namespace"],
)
logging.debug("deleted deployment %s" % deployment.metadata.name)
searchgen_deployments = apps_api.list_namespaced_deployment(
namespace=stack_config["namespace"],
label_selector="app=searchgen,test=%s" % test_id,
).items
for deployment in searchgen_deployments:
apps_api.delete_namespaced_deployment(
name=deployment.metadata.name,
namespace=stack_config["namespace"],
)
logging.debug("deleted deployment %s" % deployment.metadata.name)
def up(splunk, stack_id):
stack_config = stacks.get_stack_config(splunk, stack_id)
cluster_name = stack_config["cluster"]
kubernetes = clusters.create_client(splunk, cluster_name)
cluster_config = clusters.get_cluster(splunk, cluster_name)
status = stack_config["status"]
if status == stacks.CREATING:
stack_deployment.create_deployment(splunk, kubernetes, stack_id, stack_config, cluster_config)
app_deployment.update_apps(splunk, kubernetes, stack_id)
logging.info("created")
stacks.update_config(splunk, stack_id, {
"status": stacks.CREATED,
})
elif status == stacks.UPDATING:
stack_deployment.update_deployment(splunk, kubernetes, stack_id)
app_deployment.update_apps(splunk, kubernetes, stack_id)
logging.info("updated")
stacks.update_config(splunk, stack_id, {
"status": stacks.CREATED,
})
elif status == stacks.CREATED:
logging.info("Everything is up-to-date")
else:
logging.warning("unexpected status: %s", status)
def handle_GET(self):
path = self.request['path']
_, stack_id = os.path.split(path)
stack_config = get_stack_config(self.splunk, stack_id)
result = {
"status": stack_config["status"],
"title": stack_config["title"] if "title" in stack_config else "",
"deployment_type": stack_config["deployment_type"],
"license_master_mode": stack_config["license_master_mode"],
"cluster": stack_config["cluster"],
"namespace": stack_config["namespace"],
}
if stack_config["deployment_type"] == "distributed":
result["indexer_count"] = stack_config["indexer_count"]
result["search_head_count"] = stack_config["search_head_count"]
api_client = clusters.create_client(
self.service, stack_config["cluster"])
from kubernetes import client as kubernetes
core_api = kubernetes.CoreV1Api(api_client)
hosts = services.get_load_balancer_hosts(
core_api, stack_id, services.search_head_role, stack_config["namespace"])
if hosts:
admin_password = instances.get_admin_password(core_api, stack_id, stack_config, services.search_head_role)
result.update({
"search_head_endpoint": ["http://%s" % hostname for hostname in hosts],
"search_head_password": admin_password,
})
if stack_config["license_master_mode"] == "local":
hosts = services.get_load_balancer_hosts(
core_api, stack_id, services.license_master_role, stack_config["namespace"])
if hosts:
admin_password = instances.get_admin_password(core_api, stack_id, stack_config, services.license_master_role)
result.update({
"license_master_endpoint": ["http://%s" % hostname for hostname in hosts],
"license_master_password": admin_password,
})
hosts = services.get_load_balancer_hosts(
core_api, stack_id, services.cluster_master_role, stack_config["namespace"])
if hosts:
admin_password = instances.get_admin_password(core_api, stack_id, stack_config, services.cluster_master_role)
result.update({
"cluster_master_endpoint": ["http://%s" % hostname for hostname in hosts],
"cluster_master_password": admin_password,
})
hosts = services.get_load_balancer_hosts(
core_api, stack_id, services.deployer_role, stack_config["namespace"])
if hosts:
admin_password = instances.get_admin_password(core_api, stack_id, stack_config, services.deployer_role)
result.update({
"deployer_endpoint": ["http://%s" % hostname for hostname in hosts],
"deployer_password": admin_password,
})
hosts = services.get_load_balancer_hosts(
core_api, stack_id, services.standalone_role, stack_config["namespace"])
if hosts:
admin_password = instances.get_admin_password(core_api, stack_id, stack_config, services.standalone_role)
result.update({
"standalone_endpoint": ["http://%s" % hostname for hostname in hosts],
"standalone_password": admin_password,
})
hosts = services.get_load_balancer_hosts(
core_api, stack_id, services.indexer_role, stack_config["namespace"])
if hosts:
admin_password = instances.get_admin_password(core_api, stack_id, stack_config, services.indexer_role)
result.update({
"indexer_endpoint": ["%s:9997" % hostname for hostname in hosts],
"indexer_password": admin_password,
})
self.send_result(result)
def run_cases(splunk, test_id, test):
cases_collection = get_performance_test_cases_collection(splunk)
cases = cases_collection.query(
query=json.dumps({
"test_id": test_id,
}),
sort="index:1",
)
for case in cases:
case_id = case["_key"]
status = case["status"]
if status == CASE_FINISHED:
continue
if status == CASE_WAITING:
result = splunk.post(
"saas/stacks", **{
"deployment_type": case["deployment_type"],
"indexer_count": case["indexer_count"],
"search_head_count": case["search_head_count"],
"cpu_per_instance": case["cpu_per_instance"],
"etc_storage_in_gb": case["etc_storage_in_gb"],
"other_var_storage_in_gb": case["other_var_storage_in_gb"],
"indexer_var_storage_in_gb":
case["indexer_var_storage_in_gb"],
"memory_per_instance": case["memory_per_instance"],
"title":
"Performance Test %s and Case %s" % (test_id, case_id),
"cluster": test["cluster"],
})
response = json.loads(result.body.read())["entry"][0]["content"]
stack_id = response["stack_id"]
logging.info("created stack %s for test case %s" %
(stack_id, case_id))
case.update({
"status": CASE_STARTING,
"stack_id": stack_id,
})
cases_collection.update(case_id, json.dumps(case))
raise errors.RetryOperation(
"waiting for stack %s in test case %s starting up ..." %
(stack_id, case_id))
elif status == CASE_STARTING:
stack_id = case["stack_id"]
stack = splunk.get("saas/stack/%s" % stack_id)
stack_status = json.loads(
stack.body.read())["entry"][0]["content"]["status"]
if stack_status == stacks.CREATING:
raise errors.RetryOperation()
if stack_status != stacks.CREATED:
raise Exception("unexpected stack status: %s" % stack_status)
logging.info("successfully created stack %s for case %s" %
(stack_id, case_id))
stack_config = stacks.get_stack_config(splunk, stack_id)
kube_client = clusters.create_client(splunk,
stack_config["cluster"])
cluster_config = clusters.get_cluster(splunk, test["cluster"])
node_selector_labels = cluster_config["node_selector"].split(",")
node_selector_for_generators = {}
for label in node_selector_labels:
if label:
kv = label.split("=")
if len(kv) != 2:
raise errors.ApplicationError(
"invalid node selector format (%s)" %
cluster_config.node_selector)
node_selector_for_generators[kv[0]] = kv[1]
apps_api = kubernetes.AppsV1Api(kube_client)
core_api = kubernetes.CoreV1Api(kube_client)
if stack_config["deployment_type"] == "standalone":
indexer_hosts = services.get_load_balancer_hosts(
core_api, stack_id, services.standalone_role,
stack_config["namespace"])
elif stack_config["deployment_type"] == "distributed":
indexer_hosts = services.get_load_balancer_hosts(
core_api, stack_id, services.indexer_role,
stack_config["namespace"])
else:
raise Exception("unexpected deployment type: %s" %
stack_config["deployment_type"])
data_volume_in_gb_per_day = int(case["data_volume_in_gb_per_day"])
logging.debug("data_volume_in_gb_per_day=%s" %
(data_volume_in_gb_per_day))
data_volume_in_gb_per_second = data_volume_in_gb_per_day / 24 / 60 / 60
logging.debug("data_volume_in_gb_per_second=%s" %
(data_volume_in_gb_per_second))
data_volume_in_kb_per_second = data_volume_in_gb_per_second * 1024 * 1024
logging.debug("data_volume_in_kb_per_second=%s" %
(data_volume_in_kb_per_second))
max_kb_per_second_per_data_generator = 100
logging.debug("max_kb_per_second_per_data_generator=%s" %
(max_kb_per_second_per_data_generator))
number_of_data_generators = max(
int(data_volume_in_kb_per_second /
max_kb_per_second_per_data_generator) + 1, 1)
logging.debug("number_of_data_generators=%s" %
(number_of_data_generators))
data_volume_in_kb_per_second_per_data_generator = data_volume_in_kb_per_second / \
number_of_data_generators
logging.debug(
"data_volume_in_kb_per_second_per_data_generator=%s" %
(data_volume_in_kb_per_second_per_data_generator))
deployment_name = "datagen-%s" % (stack_id)
try:
apps_api.read_namespaced_deployment(
deployment_name, namespace=stack_config["namespace"])
data_gen_deployment_already_exists = True
except kubernetes.rest.ApiException as e:
if e.status != 404:
raise
data_gen_deployment_already_exists = False
if not data_gen_deployment_already_exists:
apps_api.create_namespaced_deployment(
namespace=stack_config["namespace"],
body=kubernetes.V1Deployment(
metadata=kubernetes.V1ObjectMeta(
name=deployment_name,
namespace=stack_config["namespace"],
labels={
"app": "datagen",
"test": test_id,
"case": case_id,
},
),
spec=kubernetes.V1DeploymentSpec(
replicas=number_of_data_generators,
selector=kubernetes.V1LabelSelector(
match_labels={
"name": "datagen-%s" % (stack_id),
}),
template=kubernetes.V1PodTemplateSpec(
metadata=kubernetes.V1ObjectMeta(labels={
"name":
"datagen-%s" % (stack_id),
"app":
"datagen",
"test":
test_id,
"case":
case_id,
"stack":
stack_id,
}, ),
spec=kubernetes.V1PodSpec(
containers=[
kubernetes.V1Container(
name="datagen",
image=
"blackhypothesis/splunkeventgenerator:latest",
resources=kubernetes.
V1ResourceRequirements(
requests={
"memory": "10Mi",
"cpu": "500m",
},
limits={
"memory": "50Mi",
"cpu": "1",
},
),
env=[
kubernetes.V1EnvVar(
name="DSTHOST",
value=";".join(
map(
lambda host: host +
":9996",
indexer_hosts)),
),
kubernetes.V1EnvVar(
name="KB_S",
value="%s" %
data_volume_in_kb_per_second_per_data_generator,
),
],
),
],
node_selector=node_selector_for_generators,
),
),
),
),
)
logging.info("created %s data generators for case %s" %
(number_of_data_generators, case_id))
if stack_config["deployment_type"] == "standalone":
search_head_host = services.get_load_balancer_hosts(
core_api, stack_id, services.standalone_role,
stack_config["namespace"])[0]
elif stack_config["deployment_type"] == "distributed":
search_head_host = services.get_load_balancer_hosts(
core_api, stack_id, services.search_head_role,
stack_config["namespace"])[0]
else:
raise Exception("unexpected deployment type: %s" %
stack_config["deployment_type"])
searches_per_day = int(case["searches_per_day"])
logging.debug("searches_per_day=%s" % (searches_per_day))
searches_per_second = searches_per_day / 24 / 60 / 60
logging.debug("searches_per_second=%s" % (searches_per_second))
max_searches_per_second_per_generator = 5
logging.debug("max_searches_per_second_per_generator=%s" %
(max_searches_per_second_per_generator))
number_of_search_generators = max(
int(searches_per_second /
max_searches_per_second_per_generator) + 1, 1)
logging.debug("number_of_search_generators=%s" %
(number_of_search_generators))
searches_per_second_per_generator = searches_per_second / \
number_of_search_generators
logging.debug("searches_per_second_per_generator=%s" %
(searches_per_second_per_generator))
search_template = case["search_template"]
if searches_per_day > 0 and search_template:
deployment_name = "searchgen-%s" % (stack_id)
try:
apps_api.read_namespaced_deployment(
deployment_name, namespace=stack_config["namespace"])
search_gen_deployment_already_exists = True
except kubernetes.rest.ApiException as e:
if e.status != 404:
raise
search_gen_deployment_already_exists = False
if not search_gen_deployment_already_exists:
admin_password = instances.get_admin_password(
core_api, stack_id, stack_config,
services.search_head_role)
apps_api.create_namespaced_deployment(
namespace=stack_config["namespace"],
body=kubernetes.V1Deployment(
metadata=kubernetes.V1ObjectMeta(
name=deployment_name,
namespace=stack_config["namespace"],
labels={
"app": "searchgen",
"test": test_id,
"case": case_id,
},
),
spec=kubernetes.V1DeploymentSpec(
replicas=number_of_search_generators,
selector=kubernetes.V1LabelSelector(
match_labels={
"name": "searchgen-%s" % (stack_id),
}),
template=kubernetes.V1PodTemplateSpec(
metadata=kubernetes.V1ObjectMeta(labels={
"name":
"searchgen-%s" % (stack_id),
"app":
"searchgen",
"test":
test_id,
"case":
case_id,
"stack":
stack_id,
}, ),
spec=kubernetes.V1PodSpec(
containers=[
kubernetes.V1Container(
name="searchgen",
image=
"hovu96/splunk-searchgen:latest",
resources=kubernetes.
V1ResourceRequirements(
requests={
"memory": "10Mi",
"cpu": "500m",
},
limits={
"memory": "50Mi",
"cpu": "1",
},
),
env=[
kubernetes.V1EnvVar(
name="SEARCH_GEN_SPL",
value=search_template,
),
kubernetes.V1EnvVar(
name="SEARCH_GEN_HOST",
value=search_head_host,
),
kubernetes.V1EnvVar(
name="SEARCH_GEN_USER",
value="admin",
),
kubernetes.V1EnvVar(
name=
"SEARCH_GEN_PASSWORD",
value=admin_password,
),
kubernetes.V1EnvVar(
name="SEARCH_GEN_SPS",
value="%s" %
searches_per_second_per_generator,
),
],
),
],
node_selector=
node_selector_for_generators,
),
),
),
),
)
logging.info("created %s search generators for case %s" %
(number_of_search_generators, case_id))
else:
logging.info("no search generators started")
case.update({
"status": CASE_RUNNING,
"time_started_running": time.time(),
})
cases_collection.update(case_id, json.dumps(case))
raise errors.RetryOperation("running test case %s ..." % case_id)
elif status == CASE_RUNNING:
time_started_running = case["time_started_running"]
time_now = time.time()
seconds_running_to_far = time_now - time_started_running
target_run_duration = test["run_duration"]
logging.debug(
"time_started_running=%s time_now=%s seconds_running_to_far=%s"
% (time_started_running, time_now, seconds_running_to_far))
if seconds_running_to_far < (target_run_duration * 60):
logging.debug("still waiting")
raise errors.RetryOperation()
logging.info("time elapsed for case %s" % (case_id))
case.update({
"status": CASE_STOPPING,
"time_finished_running": time.time(),
})
cases_collection.update(case_id, json.dumps(case))
raise errors.RetryOperation("stopping test case %s" % case_id)
elif status == CASE_STOPPING:
stop_case(splunk, test_id, case_id, case)
case.update({
"status": CASE_FINISHED,
})
cases_collection.update(case_id, json.dumps(case))
logging.info("finished test case %s" % case_id)
else:
logging.error("run_cases: unexpected status for test case %s: %s" %
(case_id, status))
raise errors.RetryOperation()