Extensive behavior customization and automation can be done by utilizing scripts for both the control plane with Python-based ControlScripts and the data plane with Lua-based DataScripts.

DataScript

DataScripts are a powerful mechanism for customizing the behavior of NSX Advanced Load Balancer on a per-virtual-service or even per-client basis. DataScripts are lightweight scripts coded in Lua. These scripts can be executed against each client making a TCP connection, an HTTP request or response, or other events within the data plane.

One or more DataScripts can be attached to the rules section of a virtual service.

Scripts may be uploaded or copied and pasted into either the Request Event Script or Response Event Script section. For instance, to restrict access to the secure directory, the following text would be pasted into the Request Event Script section:

if avi.http.uri == "/secure/" then
   avi.http.send(403)
end

See theVMware NSX Advanced Load Balancer DataScript Guide for complete documentation of commands and example DataScripts.

ControlScript

ControlScripts are Python-based scripts which execute on the NSX Advanced Load Balancer Controllers. They are initiated by Alert Actions, which themselves are triggered by events within the system. Rather than alerting an admin that a specific event has occurred, the ControlScript can take specific action, such as altering the NSX Advanced Load Balancerconfiguration or sending a custom message to an external system, such as telling VMware’s vCenter to scale out more servers if the current servers have reached resource capacity and are incurring lowered health scores.

To create a ControlScript,

  1. Navigate to Templates > Scripts > ControlScripts.

  2. Click Create. The New ControlScript screen appears.

  3. Enter a Name for the ControlScript.

    Note:

    The Enter Text option is selected by default.

  4. Enter the user defined alert action script in the text box provided or upload a .py file

  5. Click Save.

ControlScripts are executed with limited privileges within the Linux subsystem of the Controller. The file system access will be read-only, thus you cannot create new files or modify any files or directories.

Exception: ControlScripts are allowed both read and write access to the /tmp directory with a quota limit of 10 MB. Any file written in /tmp is temporary and will be lost as soon as the ControlScript ends.

There will not be any type of access to the Controller system files and the Controller system python libraries in the control-script. Any script files that need to be shared in the container for use in control script has to be placed in /opt/avi/python/lib/avi/scripts/csshared folder. This can be imported using import avi.scripts.csshared.<module> code.

See standard Python documentation for examples and definitions of Python commands. NSX Advanced Load Balancer configuration changes may be made by API calls from Linux to NSX Advanced Load Balancer through standard API mechanisms.

There are two sets of variables that are passed onto the ControlScripts. These include:

  • Environment Variables

  • Script Arguments

Environment Variables

The following environment variables can be used within a ControlScript:

  • USER/ API_TOKEN: Contains the user name and API token that can be used to authenticate with the NSX Advanced Load Balancer REST API.

  • TENANT/ TENANT_UUID: Contains the tenant context (name/ UUID) in which the ControlScript is executing.

  • DOCKER_GATEWAY: Contains the local IP address that the ControlScript can use to communicate with the NSX Advanced Load Balancer REST API.

  • VERSION: Contains the version of the Controller in which the ControlScript is executing.

  • EVENT_DESCRIPTION: Contains the description of the event that triggered the alert.

The below example shows the contents of these environment variables passed to a ControlScript on a Controller running release 21.1.4, retrieved in Python using os.environ().

{
    "USER": "admin",
    "API_TOKEN": "60bc0fc8ece748c4f20f0eabf7a25eb6584af0aa",
    "TENANT": "admin",
    "TENANT_UUID": "admin",
    "DOCKER_GATEWAY": "172.17.0.1",
    "VERSION": "21.1.4",
    "EVENT_DESCRIPTION": "Config vs1 update status is success (performed by user admin)",
}
Note:

The IP address passed in the DOCKER_GATEWAY environment variable must be used to access the NSX Advanced Load Balancer REST API.

Script Arguments

The arguments that are provided to the script are provided as an array in this format:

['/home/admin/{{ ALERT NAME }}' , '{{ ALERT DETAILS }}']

Alert Details

Alert details are provided as JSON data and can be parsed using JSON as json.loads(sys.argv[1]). Here is the pretty print json of the data provided to the ControlScript:

{
  "name": "System-CC-Alert-cluster-878226b4-ff2c-4e6b-a9a7-66aa58ad23f1-1580412633.0-1580412633-38875205",
  "throttle_count": 0,
  "level": "ALERT_LOW",
  "reason": "threshold_exceeded",
  "obj_name": "AWS",
  "threshold": 1,
  "events": [
    {
      "event_id": "AWS_ACCESS_FAILURE",
      "event_details": {
        "aws_infra_details": {
          "vpc_id": "vpc-0617ccd15673817c0",
          "region": "eu-central-1",
          "error_string": "AuthFailure: AWS was not able to validate the provided access credentials\\n\\tstatus code: 401, request id: f22641cb-140b-4ff2-ab8d-0008a73c6fbf",
          "cc_id": "cloud-8a3e601b-d06d-4998-bc41-ac3004330066"
        }
      },
      "obj_uuid": "cluster-878226b4-ff2c-4e6b-a9a7-66aa58ad23f1",
      "obj_name": "AWS",
      "report_timestamp": 1580412633
    }
  ]
}
Note:

If the size of the JSON data is more than 128kb, then it will be discarded.

Examples

To understand scripting, use the examples on how to use the values, given below:

Getting Data Passed to the ControlScript

#!/usr/bin/python
#
# NSX Advanced Load Balancer ControlScript
#
# This sample ControlScript will output the environment values, and alert
# arguments that are passed from the alert that triggered the alert script.
# You can use these values to help construct your python script actions to
# handle the alert.
#
#
import os
import sys
 
if __name__ == "__main__":
    print("Environment Vars: %s \n" % os.environ)
    print("Alert Arguments: %s \n" % sys.argv)

Sticky Pool Group

#!/usr/bin/python3
import os
import sys
import json
from avi.sdk.avi_api import ApiSession
import urllib3
import requests
  
if hasattr(requests.packages.urllib3, 'disable_warnings'):
    requests.packages.urllib3.disable_warnings()
  
if hasattr(urllib3, 'disable_warnings'):
    urllib3.disable_warnings()
  
def ParseAviParams(argv):
    if len(argv) != 2:
        return
    alert_params = json.loads(argv[1])
    print(str(alert_params))
    return alert_params
  
def get_api_token():
    return os.environ.get('API_TOKEN')
  
def get_api_user():
    return os.environ.get('USER')
  
def get_api_endpoint():
    return os.environ.get('DOCKER_GATEWAY') or 'localhost'
  
def get_tenant():
    return os.environ.get('TENANT')
  
def failover_pools(session, pool_uuid, pool_name, retries=5):
    if retries <= 0:
        return 'Too many retry attempts - aborting!'
    query = 'refers_to=pool:%s' % pool_uuid
    pg_result = session.get('poolgroup', params=query)
    if pg_result.count() == 0:
        return 'No pool group found referencing pool %s' % pool_name
  
    pg_obj = pg_result.json()['results'][0]
  
    highest_up_pool = None
    highest_down_pool = None
  
    for member in pg_obj['members']:
        priority_label = member['priority_label']
        member_ref = member['pool_ref']
        pool_runtime_url = ('%s/runtime/detail' %
                            member_ref.split('/api/')[1])
        pool_obj = session.get(pool_runtime_url).json()[0]
        if pool_obj['oper_status']['state'] == 'OPER_UP':
            if (not highest_up_pool or
                int(highest_up_pool[1]) < int(priority_label)):
                highest_up_pool = (member, priority_label,
                                   pool_obj['name'])
        elif (not highest_down_pool or
              int(highest_down_pool[1]) < int(priority_label)):
            highest_down_pool = (member, priority_label,
                                 pool_obj['name'])
  
    if not highest_up_pool:
        return ('No action required as all pools in the '
                'pool group are now down.')
    elif not highest_down_pool:
        return ('No action required as all pools in the '
                'pool group are now up.')
  
    if int(highest_down_pool[1]) <= int(highest_up_pool[1]):
        return ('No action required. The highest-priority available '
                'pool (%s) already has a higher priority than the '
                'highest-priority non-available pool (%s)' %
                (highest_up_pool[2], highest_down_pool[2]))
  
    highest_up_pool[0]['priority_label'] = highest_down_pool[1]
    highest_down_pool[0]['priority_label'] = highest_up_pool[1]
  
    p_result = session.put('poolgroup/%s' % pg_obj['uuid'], pg_obj)
    if p_result.status_code < 300:
        return ', '.join(['Pool %s priority changed to %s' % (p[0], p[1])
                           for p in ((highest_up_pool[2], highest_down_pool[1]),
                                     (highest_down_pool[2], highest_up_pool[1]))])
    if p_result.status_code == 412:
        return failover_pools(session, pool_uuid, pool_name, retries - 1)
  
    return 'Error setting pool priority: %s' % p_result.text
  
if __name__ == "__main__":
    alert_params = ParseAviParams(sys.argv)
    events = alert_params.get('events', [])
    if len(events) > 0:
        token = get_api_token()
        user = get_api_user()
        api_endpoint = get_api_endpoint()
        tenant = get_tenant()
  
        pool_uuid = events[0]['obj_uuid']
        pool_name = events[0]['obj_name']
        event_id = events[0]['event_id']
        try:
            with ApiSession(api_endpoint, user,
                            token=token,
                            tenant=tenant) as session:
                result = failover_pools(session, pool_uuid, pool_name)
        except Exception as e:
            result = str(e)
    else:
        result = 'No event data for ControlScript'
  
    print(result)
 
# Use with a ControlScript and Alert(s) to perform 'sticky' failover of pool groups.
#
# Alert should trigger on 'Pool Up' and 'Pool Down' events.
#

Add Route to GCP SE

#!/usr/bin/python
 
import sys, os, json, traceback, re, time
from avi.sdk.avi_api import ApiSession
from oauth2client.client import GoogleCredentials
from googleapiclient import discovery
 
'''
This ControlScript is executed on the Controller every time there is a
CC_IP_ATTACHED or a CC_IP_DETACHED event.
 
CC_IP_ATTACHED: Event is triggered when a VIP is attached to a SE
CC_IP_DETACHED: Event is triggered when a VIP is detached from a SE, usually
when a SE goes down or a scale in occurs
 
The goal of this script is to add a route to GCP with the destination as the
VIP and nextHopIp as the GCP instance IP on which the SE is running after a
CC_IP_ATTACHED event. After a CC_IP_DETACHED event, the goal of the script is
to remove the corresponding route.
 
Script assumptions:
 
1) The Controller GCP instance has scope=compute-rw to be able to modify
routes in GCP
2) 'description' field in the Service Engine Group is configured as a
JSON encoded string containing GCP project, zone and network
 
Event details contain the SE UUID and the VIP.
 
1) GET SE object from UUID and extract SE IP address (which is
the same as the GCP instance IP address) and Service Engine Group link
 
2) GET Service Engine Group object. The 'description' field in the
Service Engine Group is a JSON encoded string containing GCP project and
network URL. Extract project and network from the 'description' field
 
3) Extract all routes matching destRange as VIP from GCP
 
4) If event is CC_IP_DETACHED, remove matching route with
destRange as vip and nextHopIp as instance IP in the appr network
If event is CC_IP_ATTACHED and no matching route exists already, add a new
route with destRange as vip and nextHopIp as instance IP in appr network
'''
 
def parse_avi_params(argv):
    if len(argv) != 2:
        return {}
    script_parms = json.loads(argv[1])
    return script_parms
 
def create_avi_endpoint():
    token=os.environ.get('API_TOKEN')
    user=os.environ.get('USER')
    # tenant=os.environ.get('TENANT')
    return ApiSession.get_session(os.environ.get('DOCKER_GATEWAY'), user, token=token,
                                  tenant='admin')
 
def google_compute():
    credentials = GoogleCredentials.get_application_default()
    return discovery.build('compute', 'v1', credentials=credentials)
 
def gcp_program_route(gcp, event_id, project, network, inst_ip, vip):
    # List all routes for vip
    result = gcp.routes().list(project=project,
                                   filter='destRange eq %s' % vip).execute()
    if (('items' not in result or len(result['items']) == 0)
        and event_id == 'CC_IP_DETACHED'):
        print(('Project %s destRange %s route not found' %
              (project, vip)))
        return
 
    if event_id == 'CC_IP_DETACHED':
        # Remove route for vip nextHop instance
        for r in result['items']:
            if (r['network'] == network and r['destRange'] == vip and
                r['nextHopIp'] == inst_ip):
                result = gcp.routes().delete(project=project,
                                             route=r['name']).execute()
                print(('Route %s delete result %s' % (r['name'], str(result))))
                # Wait until done or retries exhausted
                if 'name' in result:
                    start = int(time.time())
                    for i in range(0, 20):
                        op_result = gcp.globalOperations().get(project=project,
                                operation=result['name']).execute()
                        print(('op_result %s' % str(op_result)))
                        if op_result['status'] == 'DONE':
                            if 'error' in result:
                                print(('WARNING: Route delete had errors '
                                      'result %s' % str(op_result)))
                            else:
                                print(('Route delete done result %s' %
                                      str(op_result)))
                            break
                        if int(time.time()) - start > 20:
                            print(('WARNING: Wait exhausted last op_result %s' %
                                  str(op_result)))
                            break
                        else:
                            time.sleep(1)
                else:
                    print('WARNING: Unable to obtain name of route delete '
                          'operation')
    elif event_id == 'CC_IP_ATTACHED':
        # Add routes to instance
        # Route names can just have - and alphanumeric chars
        rt_name = re.sub('[./]+', '-', 'route-%s-%s' % (inst_ip, vip))
        route = {'name': rt_name,
            'destRange': vip, 'network': network,
            'nextHopIp': inst_ip}
        result = gcp.routes().insert(project=project,
                                         body=route).execute()
        print(('Route VIP %s insert result %s' %
                (vip, str(result))))
         
def handle_cc_alert(session, gcp, script_parms):
    se_name = script_parms['obj_name']
    print(('Event Se %s %s' % (se_name, str(script_parms))))
    if len(script_parms['events']) == 0:
        print ('WARNING: No events in alert')
        return
 
    # GET SE object from Avi for instance IP address and SE Group link
    rsp = session.get('serviceengine?uuid=%s' %
                      script_parms['events'][0]['event_details']['cc_ip_details']['se_vm_uuid'])
    if rsp.status_code in range(200, 299):
        se = json.loads(rsp.text)
        if se['count'] == 0 or len(se['results']) == 0:
            print(('WARNING: SE %s no results' %
                script_parms['events'][0]['event_details']['cc_ip_details']['se_vm_uuid']))
            return
        inst_ip = next((v['ip']['ip_addr']['addr'] for v in
                se['results'][0]['mgmt_vnic']['vnic_networks']
                if v['ip']['mask'] == 32 and v['mode'] != 'VIP'), '')
        if not inst_ip:
            print(('WARNING: Unable to find IP with mask 32 SE %s' % str(se['results'][0])))
            return
 
        # GET SE Group object for GCP project, zones and network
        # https://localhost/api/serviceenginegroup/serviceenginegroup-99f78850-4d1f-4b7b-9027-311ad1f8c60e
        seg_ref_list = se['results'][0]['se_group_ref'].split('/api/')
        seg_rsp = session.get(seg_ref_list[1])
        if seg_rsp.status_code in range(200, 299):
            vip = '%s/32' % script_parms['events'][0]['event_details']['cc_ip_details']['ip']['addr']
            seg = json.loads(seg_rsp.text)
            descr = json.loads(seg.get('description', '{}'))
            project = descr.get('project', '')
            network = descr.get('network', '')
            if not project or not network:
                print(('WARNING: Project, Network is required descr %s' %
                      str(descr)))
                return
            gcp_program_route(gcp, script_parms['events'][0]['event_id'],
                              project, network, inst_ip, vip)
        else:
            print(('WARNING: Unable to retrieve SE Group %s status %d' %
                  (se['results'][0]['se_group_ref'], seg_rsp.status_code)))
            return
    else:
        print(('WARNING: Unable to retrieve SE %s' %
               script_parms['events'][0]['obj_uuid']))
 
 
# Script entry
 
if __name__ == "__main__":
    script_parms = parse_avi_params(sys.argv)
    try:
        admin_session = create_avi_endpoint()
        gcp = google_compute()
        handle_cc_alert(admin_session, gcp, script_parms)
    except Exception:
        print(('WARNING: Exception with Avi/Gcp route %s' %
               traceback.format_exc()))

Denial of Service Attack Handling

#!/usr/bin/python
import sys, os, json
from avi.sdk.avi_api import ApiSession
 
 
'''
This control script will be executed in the Avi Controller when an
alert due to a DOS_ATTACK event is generated.
 
An example params passed to the control script dos-attack.py is as follows
 
params = [u'/home/admin/Dos_Attack-l4vs',
           '{"name": "Dos_Attack-virtualservice-d1093604-e1f0-476a-ad91-01c5224c5641-1461261720.83-1461261716-77911185",
             "throttle_count": 0,
             "level": "ALERT_HIGH",
             "reason": "threshold_exceeded",
             "obj_name": "l4vs",
             "threshold": 1,
             "events":
             [
                 {
                      "event_id": "DOS_ATTACK",
                      "event_details":
                      {
                          "dos_attack_event_details":
                          {
                              "attack_count": 2150.0,
                              "attack": "SYN_FLOOD",
                              "ipgroup_uuids": [
                                  "ipaddrgroup-f6883289-39fa-418f-94c2-3b8f8093cd7a"
                               ],
                               "src_ips": ["10.10.90.67"]
                          }
                      },
                      "obj_uuid": "virtualservice-d1093604-e1f0-476a-ad91-01c5224c5641",
                      "obj_name": "l4vs",
                      "report_timestamp": 1461261716
                 }
             ]
            }'
         ]
The DOS_ATTACK event was generated due to a SYN_FLOOD from client 10.10.90.67. It was
traffic to the Virtual Service : "l4vs".
 
The offending client ip is added as NETWORK_SECURITY_POLICY_ACTION_TYPE_DENY in the
network seurity policy for the virtual service
 
'''
 
def ParseAviParams(argv):
    if len(argv) != 2:
        return
    alert_dict = json.loads(argv[1])
    return alert_dict
 
def create_avi_endpoint():
    token=os.environ.get('API_TOKEN')
    user=os.environ.get('USER')
    # tenant=os.environ.get('TENANT')
    return ApiSession.get_session(os.environ.get('DOCKER_GATEWAY'), user, token=token,
                                  tenant='admin')
 
def add_ns_rules_dos(session, dos_params):
    vs_name = dos_params['obj_name']
    vs_uuid = ''
    client_ips = []
    vs_name = dos_params['obj_name']
    for event in dos_params['events']:
        vs_uuid = event['obj_uuid']
        dos_attack_event_details = event['event_details']['dos_attack_event_details']
        if dos_attack_event_details['attack'] != 'SYN_FLOOD':
            continue
        for ip in dos_attack_event_details['src_ips']:
            client_ips.append(ip)
    if len(client_ips) == 0:
        print ('DOS ATTACK is not SYN_FLOOD. Ignoring')
        return
 
    print('VS name : ' + vs_name + ' VS UUID : ' + vs_uuid + ' Client IPs : ' + str(client_ips))
    ip_list = []
    for ip in client_ips:
        ip_addr_obj = {
            'addr' : ip,
            'type' : 'V4'
        }
        ip_list.append(ip_addr_obj)
 
    match_obj = {
        'match_criteria' : 'IS_IN',
        'addrs' : ip_list
    }
    ns_match_target_obj = {
        'client_ip' : match_obj
    }
    ns_rule_dos_obj = {
        'enable' : True,
        'log'    : True,
        'match'  : ns_match_target_obj,
        'action' : 'NETWORK_SECURITY_POLICY_ACTION_TYPE_DENY'
    }
    ns_policy_dos_obj = {
        'vs_name' : vs_name,
        'vs_uuid' : vs_uuid,
        'rules'   : [
            ns_rule_dos_obj,
        ]
    }
    print('ns_policy_dos_obj : ' + str(ns_policy_dos_obj))
    try :
        session.post(path='networksecuritypolicydos?action=block',
                 data=ns_policy_dos_obj)
    except Exception as e:
        print(str(e))
    print(('Added Client IPs ' + str(client_ips) + \
           ' in the blocked list for VS : ' + vs_name))
 
if __name__ == "__main__":
    alert_dict = ParseAviParams(sys.argv)
    try :
        admin_session = create_avi_endpoint()
    except Exception as e:
        print('login failed to Avi Controller!' + str(e))
        sys.exit(0)
    add_ns_rules_dos(admin_session, alert_dict)