[crm]: Fix "nexthop group" test failure due to incorrect nexthop#633
[crm]: Fix "nexthop group" test failure due to incorrect nexthop#633volodymyrsamotiy wants to merge 1 commit intosonic-net:masterfrom
Conversation
Signed-off-by: Volodymyr Samotiy <volodymyrs@mellanox.com>
|
@volodymyrsamotiy , I observed this issue and made a fix. However, I was planning to change it in such a way as to " |
|
@prsunny , I believe we should skip INCOMPLETE entries as well (it also could mean that entry was already deleted). If we need only REACHABLE and STALE neighbors maybe we can just use: Since you already made a fix should I close this PR? |
|
@volodymyrsamotiy , thanks for the feedback. I updated the fix and raised #635 |
Temporary solution for tests failing from recent regression as a result of upgrading to a newer SAI. Original PR: sonic-net#20049
…net#3424) Submodule src/sonic-sairedis 4ee82cb..1cf2eea: > Add acl counter match logic based on acl entry field (sonic-net#511) > Add specific comparison logic for ACL counter (sonic-net#484) Submodule src/sonic-swss 46bc1f4..660530e: > Fix VLAN error introduced with new 4.9 kernel behavior (sonic-net#1001) > Warmboot Vlan neigh restore fix (sonic-net#1040) Submodule src/sonic-utilities 11b4cf1..f76fc2c: > [warm reboot] Skip ASIC config pre-check if current image does not support it (sonic-net#637) > [FastReboot]: Send SIGINT to all teamd before stop (sonic-net#633) > [warm/fast reboot] provide strict option to prevent warm reboot under certain conditions (sonic-net#631) Signed-off-by: Ying Xie <ying.xie@microsoft.com>
Submodule src/sonic-utilities fe2c656..afaedb7: > Revert "[FastReboot]: Send SIGINT to all teamd before stop (sonic-net#633)" (sonic-net#650) > Fixed config Asym PFC CLI. (sonic-net#632) > [config] Do no stop or restart dependent services (sonic-net#582) Signed-off-by: Ying Xie <ying.xie@microsoft.com>
Signed-off-by: Volodymyr Samotiy volodymyrs@mellanox.com
Description of PR
Sometimes CRM "nexthop group" test fails due to incorrect nexthop set for the route.
Currently nexthop is retrieved using the command:
Problem is that mentioned command returns all neighbors for the specified interface, including already deleted.
It happens because in Linux neighbor entries are not deleted immediately, they are moved to another state and after that once some timeout exceeded they are deleted.
Before "nexthop group" test there is "ipv4 nexhop" test which adds static neighbor and then deletes it.
According to described above when "nexthop group" test is running that deleted neighbor entry could still be present but with the state "incomplete/failed".
As a result "ip neigh show" returns both valid and invalid (deleted) entries, but test uses only first from the list and if first is that deleted neighbor then test fails.
To fix the issue need to modify "ip neigh show" command in order to get only valid and reachable neighbor entries:
Type of change
Approach
How did you do it?
Modified "ip neigh show" command in order to get only valid and reachable neighbor entries:
How did you verify/test it?
Execute CRM test and verified that it passed.
Any platform specific information?
N/A
Supported testbed topology if it's a new test case?
N/A
Documentation
N/A