def verify_load_balancer_completed(core_api, stack_id, stack_config, role):
# getaddrinfo(host, port, 0, SOCK_STREAM)
hosts = services.get_load_balancer_hosts(core_api, stack_id, role,
stack_config["namespace"])
if not len(hosts):
raise errors.RetryOperation("waiting for %s load balancer hostname" %
role)
for host in hosts:
try:
import socket
socket.gethostbyname(host)
except socket.error:
raise errors.RetryOperation(
"Waiting for %s ingress connection (cannot resolve hostname)" %
role)
pass
if role == services.standalone_role or role == services.deployer_role or role == services.search_head_role or role == services.cluster_master_role:
try:
instances.create_client(core_api, stack_id, stack_config, role)
except ssl.SSLEOFError:
raise errors.RetryOperation(
"Waiting for %s ingress connection (SSL protocol error)" %
role)
except TimeoutError:
raise errors.RetryOperation(
"Waiting for %s ingress connection (timeout)" % role)
def push_deployer_bundle(core_api, stack_id, stack_config):
search_head_hostnames = services.get_load_balancer_hosts(
core_api, stack_id, services.search_head_role,
stack_config["namespace"])
if len(search_head_hostnames) == 0:
raise errors.RetryOperation(
"Waiting for hostname for search heads ...")
search_head_hostname = search_head_hostnames[0]
if is_sh_cluster_restart_in_progress(core_api, stack_id, stack_config):
raise errors.RetryOperation(
"wait for SH cluster restart process completed ...")
service = instances.create_client(core_api, stack_id, stack_config,
services.deployer_role)
service.post(
"apps/deploy",
target="https://%s:8089" % (search_head_hostname),
action="all",
advertising="true",
force="true",
)
logging.info("pushed SH deployer bundle")
if is_sh_cluster_restart_in_progress(core_api, stack_id, stack_config):
raise errors.RetryOperation(
"wait for SH cluster restart process completed ...")
def wait_until_ready(splunk, kubernetes, stack_id, stack_config):
indexer_cluster = get(splunk, kubernetes, stack_id, stack_config)
if not indexer_cluster:
raise Exception("could not find indexer cluster")
if not "status" in indexer_cluster:
raise errors.RetryOperation("waiting for cluster master status")
status = indexer_cluster["status"]
target_indexer_count = int(stack_config["indexer_count"])
actualy_ready_replica = status["readyReplicas"]
if target_indexer_count != actualy_ready_replica:
raise errors.RetryOperation(
"waiting for target number of indexers (expected %s, got %s)" % (
target_indexer_count,
actualy_ready_replica,
))
cluster_master_phase = status["clusterMasterPhase"]
if cluster_master_phase != "Ready":
raise errors.RetryOperation(
"waiting for cluster master to become ready (currently it's in %s phase)"
% (cluster_master_phase))
phase = status["phase"]
if phase != "Ready":
raise errors.RetryOperation(
"waiting for indexer cluster to become ready (currently it's in %s phase)"
% (phase))
def wait_until_ready(splunk, kubernetes, stack_id, stack_config):
license_master = get(splunk, kubernetes, stack_id, stack_config)
if not license_master:
raise Exception("could not find license master")
if not "status" in license_master:
raise errors.RetryOperation("waiting for license master status")
status = license_master["status"]
phase = status["phase"]
if phase != "Ready":
raise errors.RetryOperation("waiting for license master to become ready (currently it's in %s phase)" % (
phase
))
def wait_until_ready(splunk, kubernetes, stack_id, stack_config):
standalone = get(splunk, kubernetes, stack_id, stack_config)
if not standalone:
raise Exception("could not find standalone")
if not "status" in standalone:
raise errors.RetryOperation("waiting for standalone status")
status = standalone["status"]
phase = status["phase"]
if phase != "Ready":
raise errors.RetryOperation("waiting for standalone to become ready (currently it's in %s phase)" % (
phase
))
def stop_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 "stopped" in case and case["stopped"] == True:
continue
if status == CASE_WAITING:
pass
elif status == CASE_STARTING:
stop_case(splunk, test_id, case_id, case)
logging.info("stopped test case %s" % case_id)
elif status == CASE_RUNNING:
stop_case(splunk, test_id, case_id, case)
logging.info("stopped test case %s" % case_id)
elif status == CASE_STOPPING:
stop_case(splunk, test_id, case_id, case)
logging.info("stopped test case %s" % case_id)
elif status == CASE_FINISHED:
pass
else:
logging.error(
"stop_cases: unexpected status for test case %s: %s" %
(case_id, status))
raise errors.RetryOperation()
case.update({"stopped": True})
cases_collection.update(case_id, json.dumps(case))
def wait_until_ready(splunk, kubernetes, stack_id, stack_config):
search_head_cluster = get(splunk, kubernetes, stack_id, stack_config)
if not search_head_cluster:
raise Exception("could not find search head cluster")
status = search_head_cluster["status"]
# captain: ...
# captainReady: true
# deployerPhase: Ready
# initialized: true
# maintenanceMode: false
# members:
# - active_historical_search_count: 0
# active_realtime_search_count: 0
# adhoc_searchhead: false
# is_registered: true
# name: ...
# status: Up
# minPeersJoined: true
# phase: Ready
# readyReplicas: 3
# replicas: 3
# selector: ...
target_search_head_count = int(stack_config["search_head_count"])
actualy_ready_replica = status["readyReplicas"]
if target_search_head_count != actualy_ready_replica:
raise errors.RetryOperation(
"waiting for target number of search heads (expected %s, got %s)" %
(
target_search_head_count,
actualy_ready_replica,
))
deployer_phase = status["deployerPhase"]
if deployer_phase != "Ready":
raise errors.RetryOperation(
"waiting for deployer to become ready (currently it's in %s phase)"
% (deployer_phase))
captain_ready = status["captainReady"]
if not captain_ready:
raise errors.RetryOperation(
"search head cluster captain not yet ready")
phase = status["phase"]
if phase != "Ready":
raise errors.RetryOperation(
"waiting for search head cluster to become ready (currently it's in %s phase)"
% (phase))
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 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()