[LRUG] Resque reliability (was re: Experienced RoR dev needed)

[Damian Janowski]

On Sun, Jun 5, 2011 at 7:02 PM, Sam Livingston-Gray <geeksam at gmail.com> wrote:
>> Date: Sun, 5 Jun 2011 02:06:00 -0700 (PDT)
>> From: Robert Redmond <robert_redmond at yahoo.com>
>> Subject: [LRUG] Experienced RoR dev needed
>
> [snip]
>
>> Make the push notification system more robust, we are using apn_sender to send
>> the push notifications, Reddis to store in memory which IPhones need to be sent
>> a APN and Resque to schedule the push notifications. This seems not to work 100%
>> reliable at the moment. It may be the server config or we may need a different
>> approach
>
> Chiming in from across the pond here.  Resque is not 100% reliable, by
> design.  To quote Chris Wanstrath, "Resque does not advertise itself
> as a system that will never lose your jobs. Part of the design is we
> don't care if jobs are lost."[1]  This came up at a pdxruby.org
> presentation[2] in January, and I brought it back up again in a recent
> thread[3] on our mailing list.

If you want something 100% reliable, you need a library that uses
Redis' (B)RPOPLPUSH semantics. For example:
https://github.com/cyx/peons.

http://redis.io/commands/rpoplpush