Fix bug with checking VRF's routes in route_check.py

[bratashX]

Signed-off-by: Petro Bratash petrox.bratash@intel.com

What I did

Add capability to check routes with VRF

How I did it

All routes in VRF have a specific prefix in name.

Route table when we have VRF:

admin@igk-5-dut:~$ redis-cli -n 0 keys "ROUTE_TABLE*" 
    1) "ROUTE_TABLE:Vrf2:20c0:cb50:0:80::/64"
    2) "ROUTE_TABLE:Vrf2:192.211.48.0/25"
    3) "ROUTE_TABLE:Vrf1:192.246.8.128/25"
    4) "ROUTE_TABLE:Vrf1:192.222.8.0/25"
    5) "ROUTE_TABLE:Vrf2:192.246.64.0/25"
    .......

Route table without VRF:

admin@cab18-2-dut:~$ redis-cli -n 0 keys "ROUTE_TABLE*"
1) "ROUTE_TABLE:192.168.0.0/21"
2) "ROUTE_TABLE:10.1.0.32"
3) "ROUTE_TABLE:fc02:1000::/64"
4) "ROUTE_TABLE:fe80::/64"
5) "ROUTE_TABLE:fc00:1::32"
.......

Remove Vrf prefix from route. Without Vrf, we can compare routes from APP and ASIC DB.

How to verify it

Run next command on SONiC:

1. sudo config vrf add Vrf1
2. sudo config interface vrf bind PortChannel101 Vrf1
3. sudo config interface ip add PortChannel101 10.0.0.56/31
4. /usr/local/bin/route_check.py

Also can run SONiC TC vrf/test_vrf.py::TestVrfDeletion::test_vrf1_neigh_after_restore and check syslog

Previous command output (if the output of a command-line utility has changed)

$ /usr/local/bin/route_check.py
Failure results: {{
    "Unaccounted_ROUTE_ENTRY_TABLE_entries": [
        "10.0.0.56/31"
    ]
}}
Failed. Look at reported mismatches above
add: []
del: []

New command output (if the output of a command-line utility has changed)

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: bratashX / name: Petro Bratash (388b545)

[prsunny]

What is the fix? We already have the check without VRF. Can you explain a bit more?

[bratashX]

What is the fix? We already have the check without VRF. Can you explain a bit more?

Previously, we skip routes with VRF prefix (ROUTE_TABLE from APP_DB):

sonic-utilities/scripts/route_check.py

Line 284 in be1866f

if not is_vrf(k) and not is_local(k):

But from ASIC_DB we read all routes (including routes from VRF) and have a problem with comparing.

[prsunny]

What is the fix? We already have the check without VRF. Can you explain a bit more?

Previously, we skip routes with VRF prefix (ROUTE_TABLE from APP_DB):

sonic-utilities/scripts/route_check.py

Line 284 in be1866f

if not is_vrf(k) and not is_local(k):

But from ASIC_DB we read all routes (including routes from VRF) and have a problem with comparing.

ok, in that case i suggest lets fix the proper way. For VRF routes, lets get the VRF ID from ASIC_DB and compare only routes from that VRF.

[bratashX]

What is the fix? We already have the check without VRF. Can you explain a bit more?

Previously, we skip routes with VRF prefix (ROUTE_TABLE from APP_DB):

sonic-utilities/scripts/route_check.py

Line 284 in be1866f

if not is_vrf(k) and not is_local(k):

But from ASIC_DB we read all routes (including routes from VRF) and have a problem with comparing.

ok, in that case i suggest lets fix the proper way. For VRF routes, lets get the VRF ID from ASIC_DB and compare only routes from that VRF.

We can't compare only routes from specific VRF separatelly because:

• on APP_DB used name of VRF (VRF1, VRF2..etc.)
  Example route entry from APP_DB:
  "ROUTE_TABLE:Vrf1:10.0.0.56/31"

• on ASIC_DB used virtual router oid
  Example route entry from ASIC_DB:
  ASIC_STATE:SAI_OBJECT_TYPE_ROUTE_ENTRY:{"dest":"10.0.0.56/31","switch_id":"oid:0x21000000000000","vr":"oid:0x30000000003e3"}

Mapping VRF names with OIDs is absent in REDIS (it's saved into swss code).

In my implementation, I compare all routes from APP_DB (without VRF + VRF routes) with all routes from ASIC_DB.

[bratashX]

@prsunny Could you please take a look?