for i := 0; i < mr.nMap; i++ {
DPrintf("worker number is %d
", mr.workerNumber)
worker_str = <- mr.registerChannel
DPrintf("Worker_str is %s
",worker_str)
args := &DoJobArgs{mr.file,"Map",i,mr.nReduce}
var reply DoJobReply
var ret bool
ret = call(worker_str, "Worker.DoJob", args, &reply)
if ret {
fmt.Println("wk worker done.
")
fmt.Println(worker_str)
mr.registerChannel <- worker_str // <=======stuck here
} else
{
fmt.Println("wk worker fail.
")
}
DPrintf("map finished.")
}
btw, mr is instance of this:
type MapReduce struct {
nMap int // Number of Map jobs
nReduce int // Number of Reduce jobs
file string // Name of input file
MasterAddress string
registerChannel chan string
DoneChannel chan bool
alive bool
l net.Listener
stats *list.List
// Map of registered workers that you need to keep up to date
Workers map[string]*WorkerInfo
// add any additional state here
workerNumber int
}
My code hang when I do this "mr.registerChannel <- worker_str "
I really don't understand why. worker_str is available, and I want to put this resource back after using this worker. Put it back to channel, let next job use available workers.
Why it hang? thanks
In go, channels can be used for synchronization if they are not buffered. Here, the process that is responsible for consuming the mr.registerChannel
is trying to write to it. When you read from, or write to and unbuffered channel, you will wait until there is another process on the other end to write to, or read from the channel, respectively.
So, when this block attempts to write to the channel, it blocks waiting for someone to read what it wrote. Since this block is also responsible for reading, it will wait forever for itself in a deadlock. You need to either redesign this so that you hand the string back off to something else to read, or you need to use a buffered channel and don't expect to trap on the read line worker_str = <- mr.registerChannel
. That would have to be re-written as a for/select or something.