https://github.com/surfly/gevent/blob/master/gevent/greenlet.py#L457
On Fri, Oct 3, 2014 at 11:02 AM, Dikang Gu <[email protected]> wrote: > hmm, it seems greenlet.killall will call the joinall first. > > On Fri, Oct 3, 2014 at 12:27 AM, Dikang Gu <[email protected]> wrote: > >> Cool, are you going to change that? Or I can try to patch it as well. I >> already committed some changes to the instagram wal-e fork. >> >> Thanks >> Dikang. >> >> On Thu, Oct 2, 2014 at 6:09 PM, Daniel Farina <[email protected]> wrote: >> >>> On Thu, Oct 2, 2014 at 5:43 PM, Dikang Gu <[email protected]> wrote: >>> > Hello there, >>> > >>> > We are using Wal-e to backup to S3, and I find that if one of the >>> backup >>> > segment failed because of timeout, all of the backup will fail. I dig >>> into >>> > the code and find this line >>> > >>> https://github.com/wal-e/wal-e/blob/master/wal_e/worker/pg/wal_transfer.py#L147 >>> . >>> > >>> > So I'm wondering why do you killall the greenlets instead of letting >>> other >>> > good ones finish? Any reason behind this? >>> >>> Thinking back, it's probably a general preference to 'join' to exit >>> promptly when there is trouble, and this code path has not been a >>> material problem for anyone who has reported bugs before. >>> >>> I think it'd be fine to be more lenient about waiting for other >>> greenlets to finish. Try patching it. >>> >> >> >> >> -- >> Dikang >> >> > > > -- > Dikang > > -- Dikang -- You received this message because you are subscribed to the Google Groups "wal-e" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
