[fast-reboot] Fix arp entries will be cleared after swssconfig restores. #5842
Conversation
|
|
| stdout_logfile=syslog | ||
| stderr_logfile=syslog | ||
| dependent_startup=true | ||
| dependent_startup_wait_for=swssconfig:exited | ||
| dependent_startup_wait_for=neighsyncd:running |
There was a problem hiding this comment.
Not sure that neighsyncd:running state does mean that neighsyncd already called AppRestartAssist::registerAppTable().
Potentially there is still a risk for this race condition to happen.
Probably AppRestartAssist::registerAppTable should be fixed - instead of clearing pending data let the orch consume it?
The AppRestartAssist is related to warm logic so such change may affect the warm reboot/restart.
@yxieca @prsunny @tahmed-dev @lguohan
There was a problem hiding this comment.
This issue is caused by the race condition between the init table and start writing table on two threads. I propose to put the init operation into swssconfig. It does not need to care what the detailed implementation of the underlying 'running' state.
There was a problem hiding this comment.
I agreed with @stepanblyschak. Could we add a check for warm-reboot at the beginning of AppRestartAssist if it's only needed for warm-reboot?
|
I attempted to fix it in sonic-swss sonic-net/sonic-swss#1498. Could you help to have a look at? |
- Why I did it
Fix the issue that arp entries restored by swssconfig will be cleared by neighsyncd.
- How I did it
Let swssconfig be executed after than neighsyncd
- How to verify it
After executing fast-reboot command, the arp entries in arp.json can be restored to redis.
- Which release branch to backport (provide reason below if selected)
- Description for the changelog
- A picture of a cute animal (not mandatory but encouraged)