[rabbitmq-discuss] Task queues: use rabbitmq to build one, or use rabbitmq AS one

Mark Feeney mark.feeney at gmail.com
Sun Aug 12 20:58:34 BST 2012


Hi all.  I'm new to rabbit, but have read all the docs and tinkered
around a bit.  So far I am very impressed.

I have a system where we've built a number of custom, postgres-backed
task queues.  We keep adding similar functionality, so I'm evaluating
whether a messaging system might let us drop a bunch of code and get
us a more uniform system and some additional features.

Most of the tasks that get queued are of the "never lose this" variety
(durable everything, persistent messages).  The producers are very
bursty: there may be no work for days, then "suddenly" 10M tasks are
queued and worked through over a period of days.  In this sort of use
case apps treat the broker more like a database: a durable place to
put and get work items.  So... is it wise to plan to use rabbitmq as a
datastore (huge, long-lived, durable queues), or is it better to think
of it strictly as an IPC mechanism (try to keep its queues empty all
the time)?

Hope this question makes some sense.  Thanks,


Mark Feeney


More information about the rabbitmq-discuss mailing list