[rabbitmq-discuss] "badarg ram_copies route binding" error since 1.5.0

[Ben Hood]

Aymerick,

On Fri, Dec 19, 2008 at 4:13 PM, Aymerick Jéhanne
<aymerick at fotonauts.com> wrote:
> I am on Ubuntu 8.04.1, and I have installed rabbitmq through the apt
> repository as explained here: http://www.rabbitmq.com/debian.html#apt
> [{release,"OTP  APN 181 01",
>           "R11B",
>           "5.5.5",
>           [{kernel,"2.11.5","/usr/lib/erlang/lib/kernel-2.11.5"},
>            {stdlib,"1.14.5","/usr/lib/erlang/lib/stdlib-1.14.5"},
>            {sasl,"2.1.5.1","/usr/lib/erlang/lib/sasl-2.1.5.1"}],
>           permanent}].
>
> I don't think the code will help.
>
> Actually, things work well when testing on my dev box (same OS /
> Erlang version) with a restricted data set. But the bug appears while
> testing with a larger data set on a staging server: I create hundreds
> of persistent queues with 2 or 4 bindings on each queue (but only a
> few exchanges), in a row, then I publish to an exchange ... and
> strangely the crash trace shows a queue that must not receive the
> published message, so it seems to be an issue while 'iterating' on
> bindings.
>
> It's time for me to learn a bit more Erlang, I really need the
> queue.unbind feature of rabbitmq 1.5.0 :-)
>
> I will too try to upgrade Erlang, as R11B seems to be a rather old
> version.

Interesting.

If you are running R11B-5 you may be running into the OTP-7025 issue
which was fixed in R12B-1.

However, I don't see how you could have this code because the mnesia
select statement that provokes this was not present in the 1.4.0 tree
and we patched this with a workaround before 1.5.0 was released. Hence
I think you may have a version of Rabbit that is post 1.4 and pre 1.5.

Is there any way this is at all possible (i.e. have you checked
anything out from hg)?

HTH,

Ben