net-ops/lb-test-common/overlay/etc/sunet-machine-healthy/health-checks.d/lb_healthcheck.py.check

72 lines
2.1 KiB
Plaintext
Raw Normal View History

2024-09-30 12:58:22 +00:00
#!/usr/bin/env python3
import yaml
import subprocess
import time
import sys
groupyaml = '/etc/hiera/data/group.yaml'
def get_frontends(data):
try:
return list(data['sunet_frontend']['load_balancer']['websites'].keys())
except KeyError:
return []
def check_docker_instance_status(instance):
cmd = f"docker inspect -f {r'{{.State.Status}}'} {instance}"
result = subprocess.run(cmd, shell=True, capture_output=True, text=True)
return result.stdout.strip() == 'running'
def is_exabgp_running():
cmd = ["systemctl", "is-active", "exabgp.service"]
result = subprocess.run(cmd, capture_output=True, text=True)
return result.stdout.strip() == 'active'
def check_docker_instances(instances, max_retries=3, initial_wait=10):
for instance in instances:
retries = 0
while retries < max_retries:
if check_docker_instance_status(instance):
print(f"Instance: {instance} is running!")
break
else:
print(f"Instance: {instance} is not running! Will try again in {initial_wait * (2**retries)} seconds.")
time.sleep(initial_wait * (2**retries))
retries += 1
if retries == max_retries:
print(f"Max retries reached for instance: {instance}, exiting!")
sys.exit(1)
def check_exabgp_running(max_retries=3, initial_wait=10):
retries = 0
while retries < max_retries:
if is_exabgp_running():
print("ExaBGP service is running!")
break
else:
print(f"Exabgp is not running! Will try again in {initial_wait * (2**retries)} seconds.")
time.sleep(initial_wait * (2**retries))
retries += 1
if retries == max_retries:
print(f"Max retries reached for checking if exabgp is running, exiting!")
sys.exit(1)
with open(groupyaml, 'r') as f:
data = yaml.safe_load(f)
frontends = get_frontends(data)
instances = []
for frontend in frontends:
instances.append(frontend + '-haproxy-1')
instances.append(frontend + '-monitor-1')
instances.append(frontend + '-config-1')
instances.append('frontend-api-1')
instances.append('frontend-telegraf-1')
check_exabgp_running()
check_docker_instances(instances)
sys.exit(0)