Closed jiahua-wang closed 6 months ago
Is the root case already known? Was it double join?
Double join could be an explanation. More likely scenario is that m_taskThread is not joinable because after MdioIpcServer constructor startMdioThread() was not called or because startMdioThread() failed. @kenneth-arista can provide more test details.
Double join could be an explanation. More likely scenario is that m_taskThread is not joinable because after MdioIpcServer constructor startMdioThread() was not called or because startMdioThread() failed. @kenneth-arista can provide more test details.
Can you investigate which exactly scenario is that ?
Double join could be an explanation. More likely scenario is that m_taskThread is not joinable because after MdioIpcServer constructor startMdioThread() was not called or because startMdioThread() failed. @kenneth-arista can provide more test details.
Can you investigate which exactly scenario is that ?
Unfortunately, I don't have all the logs when the problem occurred before integrating Jiahua's patch. Prior to the patch, I would often seen several syncd core files after each full sonic-mgmt test runs. I suspect these crashes are triggered in tests involving reboot.
I recall that a syncd crash was observed once during sonic-mgmt override_config_table
tests.
This patch will address both cases of double join and thread not starting properly.
@jiahua-wang - please update the branch? thanks
/azp run
/azp run
/AzurePipelines run Azure.sonic-sairedis
@kcudnik Can we merge this PR
/AzurePipelines run Azure.sonic-sairedis
@kcudnik @rlhui @judyjoseph All checks were passed when I created the PR for the first time. After I merged other commits, 2 checks always fail at the place not related to my code change. Should I close this one and create another one with the same code change?
All passed do you want me to merge?
@kcudnik Please merge.
This PR is to address the issue syncd crash observed during sonic-mgmt reboot tests #17346