So the general picture is as follows: qpid-ha manages data replication for
hot-standby in a single-primary, many-backup cluster. It does *not* manage
cluster membership: starting brokers, selecting a primary, promoting a
backup when the primary fails, and ensuring there is only one primary, even
in the event of a network partition. For all that it relies on a cluster
manager.
The interaction with the cluster manager involves:
1. Configuration to tell the cluster manager to run a cluster with 1
primary, N backups and failover
2. Provide scripts or triggers so the cluster manager can do the following
things:
- start a broker (always starts in backup mode)
- promote a backup broker to primary (run qpid-ha promote command)
- stop a broker
The rgmanager files are:
1. etc/cluster.conf-example.xml.in
2. etc/qpidd.in qpidd-primary.in
rgmanager uses a virtual IP for failover, and uses the qpidd.in and
qpidd-primary.in scripts to start/stop/promote brokers.
The rgmanager configuration is a little complicated because rgmanager does
not really support the notion of a "warm backup" cluster. To make it work
we pretend that primary and backup are different service types, so
rgmanager thinks it is keepi ng N qpidd services alive at all times and
exactly one qpidd-primary service. In fact the qpidd-primary service isn't
really a separate service, and it's scripts promote the existing backup
broker rather than starting a new broker. The fact that rgmanager doesn't
understand the relationship means the scripts and configuration need to do
a bunch of work to keep things straight.
I believe pacemaker does have native support for warm backups so the
configuation and scripts hopefully will be much simpler.
If you're interested in contributing this to the project, start a JIRA and
we can track it more easily. I can't dedicate a lot of time but hopefully I
can give you pointers to keep you on track.
On Fri, Nov 10, 2017 at 4:53 AM, andi welchlin <andi.welch...@gmail.com>
wrote:
> Hello Alan,
>
> thank you for your answer. I could install pacemaker using "apt-get install
> pacemaker".
>
> If you could give me some hints what to do to get it run it would be
> welcome.
>
> Thanks,
> Andreas
>
> On Wed, Nov 8, 2017 at 8:01 PM, Alan Conway <acon...@redhat.com> wrote:
>
> > rgmanager has been replaced by pacemaker on RHEL7, I'm not sure what is
> > available on Ubuntu. There is no out-of-the-box integration with
> pacemaker
> > or other cluster managers, but all that is needed is some configuration
> and
> > a couple of scripts to enable the cluster manager to start, stop and
> > promote Qpid brokers at the appropriate time - the cluster manager
> already
> > knows what the "appropriate time" is based on its liveness and membership
> > rules. If you're interested in trying to make it work, I can give you
> more
> > pointers on what is needed.
> >
> > On Wed, Nov 8, 2017 at 1:18 PM, Steve Huston <shus...@riverace.com>
> wrote:
> >
> > > I have used the qpid C++ broker in clusters. On RHEL 6. Works very
> well.
> > >
> > > > -----Original Message-----
> > > > From: andi welchlin [mailto:andi.welch...@gmail.com]
> > > > Sent: Wednesday, November 08, 2017 12:00 PM
> > > > To: users@qpid.apache.org
> > > > Subject: HA Cluster using Qpid C++ Broker 1.36.0
> > > >
> > > > Hello,
> > > >
> > > > I would like to configure a HA cluster using the Qpid C++ message
> > broker
> > > > 1.36.0. In the handbook I read that rgmanager is neccessary.
> > > >
> > > > When I try to install it on Ubuntu 16.04 it is not found. On a Fedora
> > > installation
> > > > it is also not found.
> > > >
> > > > Is anyone here using the Qpid message broker in a cluster?
> > > >
> > > > Are you using rgmanager or is there some other cluster manager usable
> > on
> > > > Linux?
> > > >
> > > > Any hints are welcome ... even if you say "noone uses the Qpid broker
> > in
> > > a
> > > > cluster".
> > > >
> > > > Kind Regards,
> > > > Andreas
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: users-unsubscr...@qpid.apache.org
> > > For additional commands, e-mail: users-h...@qpid.apache.org
> > >
> > >
> >
>
