qpid-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Rothkin, Steve (NY81)" <Steve.Roth...@Honeywell.com>
Subject RE: Queue mirroring with message grouping and clustering on Windows
Date Tue, 16 Jul 2013 15:45:15 GMT
For what I'm trying to do, is using HA Queue replication between standalone (NOT clustered)
nodes a viable option (one that will give me the high availability/reliability, load balancing,
message grouping, and failover)? Or is the clustering required for the message grouping feature
to work?

At http://qpid.apache.org/releases/qpid-0.22/cpp-broker/book/ha-queue-replication.html it
states "When the replica queue is on the active node of a cluster, there is no automatic failover.
Automatic failover will be supported in the future, in the meantime you can use the workaround
detailed in Section 1.11.12, "Failover of federation links and queue replication."."
But it doesn't say anything about failover for queues on unclustered nodes -- is that automatic
or is the same (or some other) workaround required?

> > > From: Steve Huston [mailto:shuston@riverace.com]
> > > > I'm considering a Qpid.22 implementation under MS Windows for
> > > > message queueing.  In the future we might go to a mixed
> > > > environment with both Windows and Linux computers.
> > > > For fault tolerance, I want the queues to be mirrored across 2 to
> > > > 3 computers which are connected by high speed LAN. Each queue will
> > > > have multiple consumers on different computers (including some NOT
> > > > hosting the queue), so I also need to use the message grouping
> > > > feature to ensure that messages from a single source are not
> > > > processed
> > out of order.
> > >
> > > Ok.
> > >
> > > > 1)   Is clustering required to do this (on RabbitMQ it is, but RabbitMQ
> > appears
> > > > to not support message grouping yet)?
> > >
> > > For mirroring across a set of nodes for FT, yes.
> > >
> > > > 2)   What is required to use clustering on Windows (and is it even
> > available)?
> > > > So far I've read that Corosync is required for clustering and in
> > > > another place (a few years old) I read that Corosync isn't ported
> > > > to/doesn't build on Windows.
> > >
> > > The new HA module in Qpid requires integration with a resource
> > > manager; on Linux this is rgmanager (corosync is involved to manage
> > > the cluster itself, but it's not directly involved with Qpid).
> > > Currently there is no integration with a resource manager on Windows
> > > clusters. It's probably not a gigantic amount of work to get it
> > > there, but it's
> > work that is needed.
> >
> > If there is a mix of linux and windows computers hosting the queues,
> > do they need to use the same (or a compatible) resource manager?
> 
> Yes, the brokers participating in the cluster do need to use the same resource manager.
> 
> > As a way of getting around the need for a resource manager on Windows,
> > if I made a cluster of linux machines (or linux VMs running under
> > Windows) to host the mirrored queues would it then be ok/supported to
> > have windows clients (NOT part of the cluster) to write to and read from the queues?
> 
> Absolutely.
> 
> > If I wanted to try to do the windows resource manager implemtentation
> > (assuming my boss would allow it, and given that I have zero
> > experience developing within Qpid), is there existing documentation
> > that makes it clear (from the Qpid point of view) what I need to add
> > to the code. If so, pointers to where to look would be helpful.
> 
> You wouldn't necessarily need to add anything to Qpid - my understanding is that it's
> more of a Windows scripting/integration issue and not so much a Qpid coding issue. You
> can check this book:
> http://qpid.apache.org/releases/qpid-0.22/cpp-broker/book/chapter-ha.html
> Section 1.11.10 has the info to get you going.
> 
> > BTW, while the queues have to be HA/fault tolerant, I'm expecting that
> > most of them DON'T need the messages to be persistent because the
> > messages also have a timeliness aspect to them. So if all the hosting
> > nodes go down at the same time, it may not make sense to be able to
> > recover the messages if it takes more than a few minutes for one of them to come
> back up.
> 
> That's going to make your life much easier.
> 
> -Steve


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


Mime
View raw message