def test_get_asg_tag_fail(self):
tags = [
{
'ResourceId': 'string',
'ResourceType': 'string',
'Key': 'eks-rolling-update:desired_capacity',
'Value': '6',
'PropagateAtLaunch': True
},
]
response = get_asg_tag(tags, "foo")
self.assertEqual(response, {})
def scale_up_asg(cluster_name, asg, count):
asg_old_max_size = asg['MaxSize']
asg_old_desired_capacity = asg['DesiredCapacity']
desired_capacity = asg_old_desired_capacity + count
asg_tags = asg['Tags']
asg_name = asg['AutoScalingGroupName']
# remove any stale suspensions from asg that may be present
modify_aws_autoscaling(asg_name, "resume")
asg_tag_desired_capacity = get_asg_tag(asg_tags,
app_config["ASG_DESIRED_STATE_TAG"])
asg_tag_orig_capacity = get_asg_tag(asg_tags,
app_config["ASG_ORIG_CAPACITY_TAG"])
asg_tag_orig_max_capacity = get_asg_tag(
asg_tags, app_config["ASG_ORIG_MAX_CAPACITY_TAG"])
if desired_capacity == asg_old_desired_capacity:
logger.info(
f'Desired and current capacity for {asg_name} are equal. Skipping ASG.'
)
if asg_tag_desired_capacity.get('Value'):
logger.info(
'Found capacity tags on ASG from previous run. Leaving alone.')
return int(asg_tag_desired_capacity.get('Value')), int(
asg_tag_orig_capacity.get('Value')), int(
asg_tag_orig_max_capacity.get('Value'))
else:
save_asg_tags(asg_name, app_config["ASG_ORIG_CAPACITY_TAG"],
asg_old_desired_capacity)
save_asg_tags(asg_name, app_config["ASG_DESIRED_STATE_TAG"],
asg_old_desired_capacity)
save_asg_tags(asg_name, app_config["ASG_ORIG_MAX_CAPACITY_TAG"],
asg_old_max_size)
return asg_old_desired_capacity, asg_old_desired_capacity, asg_old_max_size
# only scale up if no previous desired capacity tag set
if asg_tag_desired_capacity.get('Value'):
logger.info(
'Found previous desired capacity value tag set on asg from a previous run.'
)
logger.info(
f'Maintaining previous capacity of {asg_old_desired_capacity} to not overscale.'
)
asg_instance_count = count_all_cluster_instances(cluster_name)
# check cluster health before doing anything
if not validate_cluster_health(
asg_name, int(asg_tag_desired_capacity.get('Value')),
asg_instance_count):
logger.info('Exiting since ASG healthcheck failed')
raise Exception('ASG healthcheck failed')
return int(asg_tag_desired_capacity.get('Value')), int(
asg_tag_orig_capacity.get('Value')), int(
asg_tag_orig_max_capacity.get('Value'))
else:
logger.info(
'No previous capacity value tags set on ASG; setting tags.')
save_asg_tags(asg_name, app_config["ASG_ORIG_CAPACITY_TAG"],
asg_old_desired_capacity)
save_asg_tags(asg_name, app_config["ASG_DESIRED_STATE_TAG"],
desired_capacity)
save_asg_tags(asg_name, app_config["ASG_ORIG_MAX_CAPACITY_TAG"],
asg_old_max_size)
# only change the max size if the new capacity is bigger than current max
if desired_capacity > asg_old_max_size:
scale_asg(asg_name, asg_old_desired_capacity, desired_capacity,
desired_capacity)
else:
scale_asg(asg_name, asg_old_desired_capacity, desired_capacity,
asg_old_max_size)
cluster_health_wait = app_config['CLUSTER_HEALTH_WAIT']
logger.info(
f'Waiting for {cluster_health_wait} seconds for ASG to scale before validating cluster health...'
)
time.sleep(cluster_health_wait)
asg_instance_count = count_all_cluster_instances(cluster_name)
# check cluster health before doing anything
if not validate_cluster_health(asg_name, desired_capacity,
asg_instance_count):
logger.info('Exiting since ASG healthcheck failed')
raise Exception('ASG healthcheck failed')
return desired_capacity, asg_old_desired_capacity, asg_old_max_size
def update_asgs(asgs, cluster_name):
for asg in asgs:
logger.info('\n')
logger.info(
'**** Starting rolling update for autoscaling group {} ****'.
format(asg['AutoScalingGroupName']))
asg_name = asg['AutoScalingGroupName']
asg_lc_name = asg['LaunchConfigurationName']
asg_old_max_size = asg['MaxSize']
instances = asg['Instances']
asg_old_desired_capacity = asg['DesiredCapacity']
asg_tags = asg['Tags']
# return a list of outdated instances
outdated_instances = []
for instance in instances:
if instance_outdated(instance, asg_lc_name):
outdated_instances.append(instance)
logger.info('Found {} outdated instances'.format(
len(outdated_instances)))
# skip to next asg if there are no outdated instances
if len(outdated_instances) == 0:
continue
# remove any stale suspentions from asg that may be present
modify_aws_autoscaling(asg_name, "resume")
# check for previous run tag on asg
asg_tag_desired_capacity = get_asg_tag(
asg_tags, app_config["ASG_DESIRED_STATE_TAG"])
if asg_tag_desired_capacity.get('Value'):
logger.info(
'Found previous desired capacity value tag set on asg from a previous run. Value: {}'
.format(asg_tag_desired_capacity.get('Value')))
logger.info('Maintaining previous capacity to not overscale')
asg_new_desired_capacity = int(
asg_tag_desired_capacity.get('Value'))
asg_tag_original_capacity = get_asg_tag(
asg_tags, app_config["ASG_ORIG_CAPACITY_TAG"])
logger.info(
'Maintaining original old capacity from a previous run so we can scale back down to original size of: {}'
.format(asg_tag_original_capacity.get('Value')))
asg_old_desired_capacity = int(
asg_tag_original_capacity.get('Value'))
else:
logger.info('No previous capacity value tag set on asg')
# save original capacity to asg tags
logger.info('Setting original capacity on asg')
save_asg_tags(asg_name, app_config["ASG_ORIG_CAPACITY_TAG"],
asg_old_desired_capacity)
asg_new_desired_capacity = asg_old_desired_capacity + len(
outdated_instances)
# save new capacity to asg tags
save_asg_tags(asg_name, app_config["ASG_DESIRED_STATE_TAG"],
asg_new_desired_capacity)
# only change the max size if the new capacity is bigger than current max
if asg_new_desired_capacity > asg_old_max_size:
asg_new_max_size = asg_new_desired_capacity
else:
# dont change the size
asg_new_max_size = asg_old_max_size
# get number of k8s nodes before we scale used later
# to determine how many new nodes have been created
k8s_nodes = get_k8s_nodes()
# now scale up
scale_asg(asg_name, asg_old_desired_capacity, asg_new_desired_capacity,
asg_new_max_size)
logger.info(
'Waiting for {} seconds for asg {} to scale before validating cluster health...'
.format(app_config['CLUSTER_HEALTH_WAIT'], asg_name))
time.sleep(app_config['CLUSTER_HEALTH_WAIT'])
# check how many instances are running
asg_instance_count = count_all_cluster_instances(cluster_name)
# check cluster health before doing anything
if validate_cluster_health(asg_name, asg_new_desired_capacity,
asg_instance_count):
# pause aws autoscaling so new instances dont try
# to spawn while instances are being terminated
modify_aws_autoscaling(asg_name, "suspend")
# start draining and terminating
for outdated in outdated_instances:
# catch any failures so we can resume aws autoscaling
try:
# get the k8s node name instead of instance id
node_name = get_node_by_instance_id(
k8s_nodes, outdated['InstanceId'])
drain_node(node_name)
delete_node(node_name)
terminate_instance(outdated['InstanceId'])
if not instance_terminated(outdated['InstanceId']):
raise Exception(
'Instance is failing to terminate. Cancelling out.'
)
detach_instance(outdated['InstanceId'], asg_name)
if not instance_detached(outdated['InstanceId']):
raise Exception(
'Instance is failing to detach from ASG. Cancelling out.'
)
except Exception as e:
logger.info(e)
raise RollingUpdateException(
"Rolling update on asg failed", asg_name)
# resume aws autoscaling
modify_aws_autoscaling(asg_name, "resume")
# remove aws tag
delete_asg_tags(asg_name, app_config["ASG_DESIRED_STATE_TAG"])
delete_asg_tags(asg_name, app_config["ASG_ORIG_CAPACITY_TAG"])
logger.info('*** Rolling update of asg {} is complete! ***'.format(
asg_name))
else:
logger.info('Exiting since asg healthcheck failed')
raise Exception('Asg healthcheck failed')
logger.info('All asgs processed')
