subChannel triggerDelete lead to dead loop about grpc-go HOT 3 OPEN

Do you have a reliable way to reproduce this?

I have #6969 out to address some underlying issues in channelz, and want to do more work there to simplify and remove any possibility of this kind of thing.

from grpc-go.

Do you have a reliable way to reproduce this?

I have #6969 out to address some underlying issues in channelz, and want to do more work there to simplify and remove any possibility of this kind of thing.

This exception just occurred at service start，but service restart several times it disappeared.
I just found log INFO: [channelz] attempt to delete child with id 3 from a parent (id=5) that doesn't currently exist

from grpc-go.

Do you have a reliable way to reproduce this?

I have #6969 out to address some underlying issues in channelz, and want to do more work there to simplify and remove any possibility of this kind of thing.

we add a log in this method

func (sc *subChannel) deleteSelfFromTree() (deleted bool) {
	if !sc.closeCalled || len(sc.sockets) != 0 {
		return false
	}
	logger.Infof("deleteSelfFromTree | pid:%d id:%d", sc.pid, sc.id) // log
	sc.cm.findEntry(sc.pid).deleteChild(sc.id)
	return true
}

print this log when service start
INFO: [core] [Channel #7 SubChannel #7] Subchannel created

keep printing this log when service dead loop
INFO: [channelz] deleteSelfFromTree | pid:7 id:7

from grpc-go.