[openamq-dev] amq_server message limit

[Pieter Hintjens]

On Wed, Apr 9, 2008 at 10:48 PM, Dmitri Tsyganov
<dmitri.tsyganov at dowjones.com> wrote:

>  I also noticed that limits only apply to a connection group
> AMQ_CONNECTION_GROUP_NORMAL - regular clients.
>  It does not apply to a AMQ_CONNECTION_GROUP_CLUSTER, which I think is used
> in MTA setup.
>  Does this mean that slow MTA node can cause "peering" queue to grow
> indefinitely? Maybe in some MTA configurations, like 1 - forward all?

You have put your finger on it.  We have three different scenarios in
which limits are not applied:

 - null channel/connection, which happens in specific cases that I
forget now but will look into again
 - normal connections logged in as "super users"
 - peering connections, which are also logged in as "super users"

This is the cause IMO.  Peering connections are considered "super
users" and get to exceed queue limits.

So the following might do it - add this to your amq_server.cfg:

    <security name = "plain">
        <user name = "peering" password = "peering" group = "normal"  />
    </security>

I'll check more throroughly because this will have side-effects such
as not allowing MTAs to connect to a "locked" broker, which normally
they can.

Probably we will want to define a separate set of queue limits for MTA queues.

-Pieter