ACL test plan

[DRAFT, UNDER DEVELOPMENT]

• Overview

Related documents Teamd documentation N/A

##Overview The purpose to test functionality of ACL rules on the SONIC switch DUT with and without LAGs configured, closely resembling production environment. The test assumes all necessary configuration, including ACL and LAG configuration, BGP routes, are already pre-configured on the SONIC switch before test runs.

The test will validate that traffic is properly routed, according to the ACL rules and LAG setup, i.e. packets are forwarded through expected LAG-member-ports.

###Scope

The test is targeting a running SONIC system with fully functioning configuration. The purpose of the test is not to test specific SAI API, but functional testing of ACL on SONIC system, making sure that traffic flows correctly, according to BGP routes advertised by BGP peers of SONIC switch, and the LAG configuration.

NOTE: ACL LAG test will be able to run only in the testbed specifically created for LAG.

###Scale / Performance

###Related DUT CLI commands

Some DUT configuration is done via minigraph. See more information below

###Related SAI APIs

DUT Requirement

---

#Test structure

##Setup configuration

• LAG configuration as in LAG Test.
• BGP sessions:
  • 16 front panel ports north bound towards spine devices
  • 16 front panel ports combine each two to have 8 LAGs south bound towards spines.
• ACL

Ansible infrastructure changes

Scripts for generating ACL configuration on SONIC

There will be acl_test_table.j2 and acl_test_rules.j2 scripts which will... j2 files will be invoked by Ansible playbook which will setup and start ACL PTF test.

Ansible scripts to setup and run ACL test

acl_testbed.yml

In high level description, the script will perform following steps:

1. Run lognanalyzer 'init' phase
2. Run ACL Test (tags: testbed)
3. Run loganalyzer 'analyze' phase

acl_testbed.yml

acl_testbed.yml will:

• Generate acl json files out of the corresponding j2 files.
• Perform test-case-specific setup (if needed)
  • and invoke lag.common.yml

[LG] Not sure I totally understand how the common part will be used. Will you forward it the testcase itself and there will be such a switch case on this input value? [answer] you're right, inside will have a switch.

Setup of DUT switch

Setup of SONIC DUT will be done by Ansible script. During setup Ansible will push json file containing configuration for ACL. Data will be consumed by orchagent.

Sample: table.json [ { "ACL_TABLE:Drop_IP": { "policy_desc" : "Drop_IP_Traffic", "type" : "L3", "ports" : "Ethernet0" }, "OP": "SET" } ] rule.json [ { "ACL_RULE_TABLE:Drop_IP:TheDrop": { "priority" : "55", "ETHER_TYPE" : "2048", "PACKET_ACTION" : "DROP" }, "OP": "SET" } ]

NOTE

• Due to SONiC specifics, tables and rules configuration will reside in two different jsons and table configuration will be applied before rules to ensure correct configuration application order.

J2 templates to generate ACL configuration on SONIC

A separate scenario in the ansible sonic_test playbook will push these files to the DUT during testbed setup.

Setup of fanout switch

important No changes need to be done on the fanout switch. Fanout switch is setup once and its setup does not change across testbed setups.

Setup of VMs

---

vEOS VMs will be setup during testbed setup with proper LAG layout.

#PTF Test

Input files for PTF test

---

PTF test will be provided with a text input files describing the LAG layout and BGP routes on the DUT switch, route_info.txt and route_info.txt. Please see Ansible infrastructure changes section for description of both files.

Data in the files will be used to

• generate traffic (using route_info.txt)
• properly validate traffic is passing through valid LAGs physical ports.(using route_info.txt)

##Validation of traffic Each LAG will be mentioned in route_info.txt, as member of ECMP group. For each port we'll have port-to-lag mapping, see route_info.txt example

Traffic validation

For each route (src_ip):

• validate that physical port through which packet was received belongs to one of the LAG ports, mentioned as ECMP members in the route_info.txt for given src__ip.
  1. using data from route_info.txt find mapping from physical port index to LAG index
  2. using route_info.txt check LAG index is in the ECMP group for the src__ip
• traffic distributed evenly between LAG member ports - by keeping packet counters in PTF test for each port
• traffic is distributed evenly between LAGs - by keeping counters in PTF for each LAG

The PTF test will keep per-port counter variables for counting packets arriving on different ports. The counters for LAG member ports will be used to compare for event traffic distribution. SONIC doesn't have LAG counters.

Address types to validate

PTF test will send traffic for both IPV4 and IPV6 routes.

##Test cases

The test assumes there is a mechanism to validate logs on the DUT where we should be able to analyze /var/log/syslog for error and warning messages related to the current test. In case such messages are detected, the test is considered failed. See lognalyzer related comments in lag.common.yml section

Test case #1 - Verify TCP traffic

Test objective

Verify traffic between legs evenly distributed. Traffic is forwarded by SONIC DUT.

• PTF host will send packets according to the route_info.txt - will create packets with dst_ip according to route prefixes.
• When packet reaches to SONIC DUT, it will route packet according to BGP routes, and send it to one of vEOS BGP peers.
• PTF test will receive a copy of the packet and perform validations described in Validation of Traffic

NOTE: We are not targeting testing traffic coming into the DUT from BGP peers.

Test case #2 - min-link verification

Test objective

Verify min-link functionality.

Test configurations

[TODO][WORKITEM][clarify] - vEOS VM port shutdown/UP. Should be done using Arista command, from Ansible.

• Arista command details, for both shutdown/up.
• ssh possible from PTF host to the VMs?

####Test description

Admin down one interface at a time on DUT switch and verify that on both end the lag interface is down.

For each LAG interface on DUT:

1. Bring down LAG port on DUT

• lag.yml uses ifconfig command to put 1 member port of the LAG to 'down' state on DUT.
  • ansible will invoke interface_facts.py to validate corresponding LAG on DUT went down.
• lag.yml passes --lag-down-index to PTF lag_test.py indicating which LAG has member port shutdown on DUT.
• lag_test.py will send packets to all other LAGs (except lag-down-index) and validate
  • packets arrive on those LAGs
  • packets are spread evenly between LAGs
  • packets are spread evenly between port-members of the LAGs.

2. Bring back 'up' the LAG port on DUT and verify that the LAG interface is back to normal.

• ansible will call ifconfig to bring port 'up' on DUT.
• ansible will invoke interface_facts.py to validate corresponding LAG on DUT changed to UP.
• PTF test sends packets for routes which have given LAG as their ECMP member and validate packets are arriving.

Shutdown each LAG interface on fanout and VM.

To simulate the real production environment we'll shutdown the EOS interface which simulates the halt of sending LACP packets on one of the LAG members; then shutdown the fanout switch interface simulates the carrier down events. Combine them together we can better simulate the scenario in the production that the neighbor device ‘shutdown’ command.

Following steps will be performed for each LAG port:

######Shutdown LAG from fantout swtich and VMs and validate no traffic flows.

1. Shutdown the EOS interface
   • [clarify][how] lag.yml runs commands on VM to shutdown port.
     • [clarify] What is mapping betweetn LAG-port on DUT and EOS (VM)? Which port(s) to shutdown on VM for given LAG port on DUT?
2. Shutdown the fanout switch interface
   • [clarify][how] lag.yml runs commands on fanout switch to shutdown port.
     • [clarify] What is mapping betweetn LAG-port on DUT and port(s) on fanout switch?
3. Verify the LAG port is down on DUT
   • lag.yml invokes interface_facts.py to check LAG went down on DUT.
4. Verify that the traffic will be distributed to the rest of the 'up' LAGs
   • PTF test will run traffic, count received packets and perform comparisons on counters to check traffic evenly distributed among LAGs which are 'up'

######Bring the LAG 'up' again and validate the traffic flows through it.

1. Bring 'up' the fanout switch interface
   • [clarify][how] lag.yml runs commands on fanout switch to bring 'up' the port.
2. Bring 'up' the VM interface
   • [clarify][how] lag.yml runs commands on VM to bring the port up.
3. Verify LAG interface is 'up' on DUT
   • lag.yml invokes interface_facts.py to check LAG went down on DUT.
4. Verify traffic flows evenly through all LAGs, including the LAG which was made 'up'
   • PTF test will run traffic, count received packets and perform comparisons on counters to check traffic evenly distributed among LAGs which are 'up'

Test case #3 - LACP verification

Test objective

Verify LACP packet slow mode.

In Slow mode, the rate is - 1 packet/30 seconds.

####Test description

The purpose of this test is to make sure that the LACP rate is correctly negotiated and set on both ends of the LAG.

Following are pre-conditions for the test case:

• The DUT switch is always started with the LACP rate set to 'slow'.
• The VMs are always set with rate 'fast' on startup.
• After startup all VMs will negotiate LACP rate with DUT and set their own rate to 'slow'.

The validation will be implemented as Ansible instructions in lag.yml, without invoking PTF:

1. Ansible playbook connects to each VM.
2. Ansible playbook validates that each VM has LACP rate set to 'slow'.
3. Ansible connects to DUT and validates LACP rate is set to 'slow'.

[workitem] Clarify what are the commands to access the VMs and check for LACP rate setting.

Open Questions

Tests affected by LAG testbed setup

FIB Test

FIB test needs modification to be able to run in LAG test bed setup. [LG] the test should be agnostic to the setup. the test exptect route info so not clear to me why the test need to be changed. Please elaborate. [answer] In LAG testbed the route information will contain LAG-port indices inside ECMP group. However PTF is only aware of physical ports. we need to introduce mapping of physical-port-to-LAG port to validate packets properly. Fib test needs to change to use this mapping. Another option is to have 2 different versions of fib.j2 - one for fib testbed, another for LAG testbed.

It's required to run in this setup, so we could validate the BGP routes setup.

Other tests

TBD

FAQ

Investigating issues in VM

To ssh into VM peers:

1. To find out peers, run on DUT

• docker exec -it bgp bash
• vtysh -c 'sh ip bgp summ' - will print IPs of peers

2. ssh into the needed peer IP with credentials.