As Nathan said, it is up to the spout. Most of the spouts I’ve worked on/with do not track the number of times a specific tuple fails. With failures due to timeouts, you probably don’t want to stop replaying them (it would lead to data loss by circumventing Storm’s guaranteed delivery capability).
The other case is when the tuple data is corrupt (e.g. a bolt that parses JSON receives malformed JSON). In this case it is up to your processing code to catch the error, and react accordingly. One pattern for this is to log the error, send the bad data somewhere like a dead letter queue, and ack the tuple. Hope this helps. -Taylor On Oct 27, 2014, at 4:29 PM, Nathan Marz <[email protected]> wrote: > It's up to the spout. > > On Mon, Oct 27, 2014 at 10:09 AM, Vladi Feigin <[email protected]> wrote: > Hi, > > I know this scenario is a bit extreme but possible > If for some reason a tuple fails constantly , will Storm send it over again > and again ? > Is there some threshold to stop it ? > Thank you, > Vladi > > > > -- > Twitter: @nathanmarz > http://nathanmarz.com
signature.asc
Description: Message signed with OpenPGP using GPGMail
