[rabbitmq-discuss] Fwd: Failed to configure Rabbitmq HA(pacemaker)

[Steve Powell]

ftr
Steve Powell  (a happy new bunny)
----------some more definitions from the SPD----------
avoirdupois (phr.) 'Would you like peas with that?'
distribute (v.) To denigrate an award ceremony.
definite (phr.) 'It's hard of hearing, I think.'
modest (n.) The most mod.

Begin forwarded message:

> From: Jae Sang Lee <hyangii at gmail.com>
> Date: 2 January 2012 01:21:40 GMT
> To: Steve Powell <steve at rabbitmq.com>
> Subject: Re: [rabbitmq-discuss] Failed to configure Rabbitmq HA(pacemaker)
> 
> Yes, a reason of problem(the slave doesn't 'promote' properly) is not equal of UID/GID between master/slave.
> 
> I also try to make mirrored queue for HA. 
> 
> Thank you very much!
> 
> 
> 
> 2011/12/31 Steve Powell <steve at rabbitmq.com>
> I'm just guessing, but there is this on the pacemaker page:
> 
> > The main trick to HA Rabbit is to ensure that when the passive node becomes
> > the active node, it must have the same node-name as the failed node. It must
> > also have read and write access to the files in the shared storage, and if
> > it's going to also be part of a cluster then it must also have the same
> > Erlang cookie.
> >
> > The server runs as the user rabbitmq which is a member of the group
> > rabbitmq. You must ensure that this user and group have the same UIDs and
> > GIDs on both nodes. You can probably save yourself some time by explicitly
> > creating the rabbitmq user and group with the same UID and GID on all nodes
> > before installing the RabbitMQ server at all. If necessary, edit/etc/passwd
> > and /etc/group on both nodes, and then reinstall the RabbitMQ server to
> > ensure all necessary files are owned by the correct user and group.
> 
> That could explain why the slave doesn't 'promote' properly.
> 
> However, that said, there is also this:
> 
> > This page documents a legacy technique for achieving active-passive high
> > availability with RabbitMQ. Active-active mirrored queues are easier to use
> > and do not impose a delay at failover.
> 
> So, yes, pacemaker HA is a 'legacy' technique though it should still work
> and it is possible that in certain circumstances it may offer advantages.
> However, closely coupled rabbit nodes in a cluster with mirrored queues
> offers the best HA solution so far.
> 
> Steve Powell  (a happy bunny)
> ----------some more definitions from the SPD----------
> avoirdupois (phr.) 'Would you like peas with that?'
> distribute (v.) To denigrate an award ceremony.
> definite (phr.) 'It's hard of hearing, I think.'
> modest (n.) The most mod.
> 
> On 30 Dec 2011, at 00:51, Jae Sang Lee wrote:
> > active-active HA looks good. but that manual isn't kindly now so I tried to
> > HA using pacemaker first.
> > HA using pacemaker is old way and don't recommend now?
> >
> > 2011/12/29 Joseph Marlin <joseph.a.marlin at gmail.com>
> > Is there a specific reason that you do not want to use Rabbit's new built in
> > active-active HA?
> >
> > http://www.rabbitmq.com/ha.html
> >
>