I am writing an application that has to fetch data from - say a million
URLs. I currently have an implementation which looks like the code below
//Make sure that we just have 500 or less goroutines fetching from URLs
sem := make(chan struct{}, min(500, len(urls)))
//Check if all URLs in the request are valid and if so spawn a goroutine to
fetch data.
for _, u := range urls {
_, err := url.Parse(u)
if err != nil {
log.Printf("%s returned an error- %v", u, err)
continue
}
go fetch(ctx, sem, u)
}
func fetch(ctx context.Context, sem chan struct{}, u string) {
sem <- struct{}{}
defer func() { <-sem }()
req, err := http.NewRequest(http.MethodGet, u, nil)
if err != nil {
log.Printf("%s returned an error while creating a request- %v", u, err)
return
}
req = req.WithContext(ctx)
res, err := http.DefaultClient.Do(req)
if err != nil {
log.Printf("%s returned an error while performing a request - %v", u, err)
return
}
//Close response body as soon as function returns to prevent resource
lekage.
//https://golang.org/pkg/net/http/#Response
defer res.Body.Close()
}
Would this application choke when a million goroutines are spawned and are
waiting for a place on the sem channel? I have profiled my code using
pprof and see no problems when I tested it with 50k URLs.
What is the cost of a goroutine waiting on the semaphore channel? Would it
be ~2KB?
Is using a worker pool like the one mentioned here
<http://marcio.io/2015/07/handling-1-million-requests-per-minute-with-golang/>
better?
What would be the advantages? I am of the opinion that the runtime
scheduler is a better judge when it comes to managing goroutines.
Another question is - Would it be better that acquire the semaphore within
the loop such that I limit the number of goroutines spawned? Mr Dave Cheney
suggested otherwise in his talk here
<https://www.youtube.com/watch?time_continue=1336&v=yKQOunhhf4A>.
Any other suggestions are also welcome.
TIA!
--
You received this message because you are subscribed to the Google Groups
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
