[submodule] Update sonic-snmpagent submodule

[raphaelt-nvidia]

Why I did it

Update sonic-snmpagent submodule to pick up new commits:

21d7d97 2021-07-12 Fix: SonicV2Connector behavior change: get_all will return empty dict if (#226)
0813b42 2021-07-12 Entries under .1.3.6.1.2.1.31.1.1.1.18 OID should return the "description" field of PORT_TABLE entries in APPL_DB or CONFIG_DB. (#224)
7a78703 2021-07-08 Install dotnet core to fix python gcov warning for code covery color bar showing (#215)
e0f36a5 2021-06-30 [multi-asic]: Udpate to use SonicDBConfig from swsscommon (#219)
266bd15 2021-06-10 Restored snmp vlan support per RFC1213 and added the missing support for RFC2863 (#218)

How I did it

Update the submodule pointer to including the new commits

How to verify it

Unit tests are run via make target/python-wheels/asyncsnmp-2.1.0-py3-none-any.whl.

Which release branch to backport (provide reason below if selected)

Description for the changelog

Update submodule pointer for sonic-snmpagent #218.

A picture of a cute animal (not mandatory but encouraged)

[raphaelt-nvidia]

I believe that a failure in the automatic tests of this PR is due to interaction between the code I inserted (recall that PR #218 in sonic-snmpagent reverts #191, which was itself a revert of #169, and added code) and testing code in sonic-mgmt/tests/snmp/test_snmp_interfaces.py. The places where the test fails were added by SuvarnaMeenakshi in sonic-net/sonic-mgmt@35e1969. The sonic-mgmt code I tested with was before this update. I tried running the test with debugger, both with my code from this PR, which fails, and with a recent image from the master branch, which passes. In both instances of this code sequence:

result = verify_port_ifindex(snmp_facts, speed_snmp)
pytest_assert(not result, "Unexpected comparsion of SNMP: {}".format(result))

it appears that when testing with the master version, verify_port_ifindex is passed two objects with a large amount of data with no intersection, yielding empty result, whereas with the code from this PR, the objects are smaller but there is an intersection, yielding non-empty result and failing the test. In one of the places, the test has this comment:

TODO: Remove this check after operational status of mgmt interface # is implemented for multi-asic platform

Could the test author or someone familiar with what it does check if the test makes some assumption that the code in this PR violates?

[liat-grozovik]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[liat-grozovik]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[liat-grozovik]

@SuvarnaMeenakshi can you please review and provide your feedback to the issue raised on the test?

[liat-grozovik]

@qiluo-msft and @SuvarnaMeenakshi can you please refer to the question above?

[SuvarnaMeenakshi]

I believe that a failure in the automatic tests of this PR is due to interaction between the code I inserted (recall that PR #218 in sonic-snmpagent reverts #191, which was itself a revert of #169, and added code) and testing code in sonic-mgmt/tests/snmp/test_snmp_interfaces.py. The places where the test fails were added by SuvarnaMeenakshi in Azure/sonic-mgmt@35e1969. The sonic-mgmt code I tested with was before this update. I tried running the test with debugger, both with my code from this PR, which fails, and with a recent image from the master branch, which passes. In both instances of this code sequence:

result = verify_port_ifindex(snmp_facts, speed_snmp)
pytest_assert(not result, "Unexpected comparsion of SNMP: {}".format(result))

it appears that when testing with the master version, verify_port_ifindex is passed two objects with a large amount of data with no intersection, yielding empty result, whereas with the code from this PR, the objects are smaller but there is an intersection, yielding non-empty result and failing the test. In one of the places, the test has this comment:

TODO: Remove this check after operational status of mgmt interface # is implemented for multi-asic platform

Could the test author or someone familiar with what it does check if the test makes some assumption that the code in this PR violates?

This looks like a valid failure and is not due to the test case, but is due to the mismatch of "interface description" between config_db and SNMP facts.

Restored snmp vlan support per RFC1213 and added the missing support for RFC2863 by raphaelt-nvidia · Pull Request #218 · Azure/sonic-snmpagent (github.com) – This PR modified the ifAlias information in RFC2863, this changes the output of 1.3.6.1.2.1.31.1.1.1.18 OID.
sonic-mgmt/snmp_facts.py at master · Azure/sonic-mgmt (github.com) – This OID is used in snmp_facts to get interface description of Port channel and management interface.

The sonic-mgmt test_snmp_interfaces sonic-mgmt/test_snmp_interfaces.py at master · Azure/sonic-mgmt (github.com) compares the interface description from config_db to result from SNMP facts.

Prior to PR 218, the description of eth0 and port channels are empty string from SNMP walk and config_db, so test comparison passed.

But after PR 218, SNMP displays “alias” as description of Portchannel and management interface, config_db does not ‘description’ for port channel and management interface, this mismatch results in the failure.

Ideally, interface description output in snmp_facts should be :
(Pdb) snmp_facts[u'snmp_interfaces'][u'117'][u'description'] – Ethernet interface
u'ARISTA02T1:Ethernet1'

(Pdb) snmp_facts[u'snmp_interfaces'][u'10000'][ u'description'] - Management Interface
u''

We will not have description for port channel and management interface in config_db, so SNMP facts should be fixed to match this.

[SuvarnaMeenakshi]

@raphaelt-nvidia Can you modify SNMP PR 218, so that interface description of Port Channel and management interface matches the value retrieved from config db, which will be empty string.
The below added code in "def interface_alias" in rfc2863.py; should be removed.

    if not result:
        #RFC2863 tables don't have descriptions for LAG, vlan & mgmt; take from RFC1213
        oid = self.get_oid(sub_id)
        if oid in self.oid_lag_name_map:
            result = self.oid_lag_name_map[oid]
        elif oid in self.mgmt_oid_name_map:
            result = self.mgmt_alias_map[self.mgmt_oid_name_map[oid]]
        elif oid in self.vlan_oid_name_map:
            result = self.vlan_oid_name_map[oid]

[raphaelt-nvidia]

What is the rationale for providing values for 1.3.6.1.2.1.31.1.1.1.18 OID port interfaces, but not for portchannels, vlans or management interfaces? It was my understanding that one of the purposes of this work was to rectify the deficiency and provide these values from whatever source was available.

[SuvarnaMeenakshi]

@raphaelt-nvidia Can you take latest SNMP changes to this PR.

[qiluo-msft]

Could you update your PR description with the commits information? Sample #8161