There is a problem that hostcfgd blocks forever and does not react to SIGTERM causing a delay in warm boot. about sonic-swss-common HOT 8 CLOSED

ACK, will start investigation this issue.

from sonic-swss-common.

I found there are 2 place have infinite loop issue:

https://github.com/Azure/sonic-swss-common/blob/15c0f729648ee9f26ac9aea767d5712db34ca298/common/pubsub.cpp#L125

https://github.com/Azure/sonic-swss-common/blob/36e1f61691df7aa44782a377e4082cc4380f1018/common/configdb.h#L97

We need exit these infinite loop when receive SIGTERM by add a SIGTERM handler and check SIGTERM status in loop. also we may need check if other signal also need handle, for example SIGKILL

Also need check if some other place in swss common have same issue.

from sonic-swss-common.

@qiluo-msft any reason the issue is not opened on buildimage for tracking? i believe we only track it and not submodules.

from sonic-swss-common.

To minimize code change and make code scalable, my proposal is:
1.Add a CancellationToken class.
2.And add new parameter to these existed API.
3.Code change in hostcfgd to use the new API with a signal handler.

from sonic-swss-common.

Here is a draft PR for this issue:

#606

from swsscommon import swsscommon

global_cancellation_token = swsscommon.SignalHandlerHelper.GetCancellationToken(2)
swsscommon.SignalHandlerHelper.RegisterSignalHandler(2)

c=swsscommon.ConfigDBConnector()
c.subscribe('A', lambda a: None)
c.connect()
c.listen(global_cancellation_token, None)
^C>>>

from sonic-swss-common.

Update: here is latest example:

from swsscommon import swsscommon

global_cancellation_token = swsscommon.CancellationToken()

swsscommon.SignalHandlerHelper.RegisterCancellationToken(2, global_cancellation_token)
swsscommon.SignalHandlerHelper.RegisterSignalHandler(2)

c=swsscommon.ConfigDBConnector()
c.subscribe('A', lambda a: None)
c.connect()
c.listen(global_cancellation_token, None)

^C>>>

from sonic-swss-common.

@qiluo-msft any reason the issue is not opened on buildimage for tracking? i believe we only track it and not submodules.

If it is obvious issue in submodule, we will move it to submodule's repo.

from sonic-swss-common.

@liuh-80, @qiluo-msft as this one is blocking sonic-net/sonic-buildimage#10510. can you please see how we can expedite the solution and the review?

from sonic-swss-common.