Mailing-List: contact users-help@qpid.apache.org; run by ezmlm
Precedence: bulk
Reply-To: users@qpid.apache.org
Message-ID: <1479766853.14463.97.camel@redhat.com>
Subject: Re: [Qpid C++ Broker][HA] How to configure the HA-cluster without
 rgmanager?
From: Alan Conway <aconway@redhat.com>
To: users@qpid.apache.org
Date: Mon, 21 Nov 2016 17:20:53 -0500
In-Reply-To: <20161120175328198353118@airlink-software.com>
References: <20161120175328198353118@airlink-software.com>
Organization: Red Hat
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Content-Transfer-Encoding: 8bit
archived-at: Mon, 21 Nov 2016 22:21:32 -0000

On Sun, 2016-11-20 at 17:53 +0800, lei.dai@airlink-software.com wrote:
> Hi,
> 
> I'm trying to configure the qpid HA-cluster without rgmanager but it
> doesn't work.
> 
> The version Infos:
> ---------------------------
> CentOS 7.2
> qpid-cpp-1.35.0
> qpid-proton-0.14.0
> ---------------------------
> 
> I installed qpid-cpp and qpid-proton in according to the INSTALL.txt
> with default configurations on 3 CentOS VMs.
> The IP addresses are: 172.16.28.231, 172.16.28.232, 172.16.23.233.
> 
> Started qpid on every VM with command:
> $qpidd --auth no --ha-cluster yes --ha-brokers-url
> 172.16.28.231,172.16.28.232,172.16.28.233
> 
> But when I use qpid-ha to show the status of the cluster, all nodes
> are in joining status.
> It means the nodes didn't connect to each other?

rgmanager is the thing that promotes the initial primary, and promotes
a new primary if that fails. The brokers are waiting for something to
promote one of them.

> Did I have some mistakes in the configurations or the usage?
> I read the documents on the qpid site. It says that rgmanager can
> manage the cluster and set the primary node by calling qpidd-primary
> script.
> But I didn't find the qpidd-primary script in my system.

Looks like it has been converted to a systemd script:

   /usr/lib/systemd/system/qpidd-primary.service

It simply calls `qpid-ha promote`

> Is the rgmanager necessary to use HA-cluster?

You need something to do the basic cluster-management tasks:
- detect primary failure, choose a backup to promote
- ensure exactly one primary is running at all times
- enforce quorum

Qpid HA is designed so that you can use something other than rgmanager
(e.g. pacemaker) but only scripts for rgmanager are provided. If you
want to use something else you'd need to write your own scripts.

Cheers,
Alan.

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@qpid.apache.org
For additional commands, e-mail: users-help@qpid.apache.org