[ClusterLabs] Podman active-active bundle

Ken Gaillot kgaillot at redhat.com
Mon Jul 22 14:30:48 UTC 2024


On Fri, 2024-07-19 at 09:43 +0000, MAIER Jakob wrote:
> Hi,
> 
> I am aware of the documentation that you have linked here, and I have
> managed to setup a bundle running on two nodes. What I am having
> problems with is running them as master-slave. Would this be handled
> by the primitive running inside the bundle, since the podman
> resource-agent does not have the promote/demote actions? If possible,
> I would like the promotion and demotion only depend on the state of
> the container, not what is running inside (so no primitive). 

Yes, promotion is handled by the primitive inside the bundle. Only OCF
resource agents support promote and demote actions, so there has to be
a primitive inside the bundle. Set the promoted-max option on the
bundle to activate support for promotion (how many instances can be
promoted at any given time, usually 1).

> 
> Also, how would you setup networking to have a single virtual ip for
> all clones in the bundle? I have not found any documentation on how
> to do something like that.

I'm not aware of that being possible. I think what you want is a
separate virtual IP resource colocated with the bundle's promoted
instance, and a port mapping to forward connections to the container.

> 
> Concerning the RHEL UBI image, I have created my own base image that
> has pacemaker-remoted installed, so that should not be a problem. 
> 
> Best,
> Jakob
> 
> 
> -----Original Message-----
> From: Ken Gaillot <kgaillot at redhat.com> 
> Sent: Mittwoch, 17. Juli 2024 16:09
> To: Cluster Labs - All topics related to open-source clustering
> welcomed <users at clusterlabs.org>
> Cc: MAIER Jakob <Jakob.MAIER at frequentis.com>
> Subject: Re: [ClusterLabs] Podman active-active bundle
> 
> *EXTERNAL source*
> 
> 
> Hi,
> 
> The upstream bundle documentation is at:
> 
> https://clusterlabs.org/pacemaker/doc/2.1/Pacemaker_Explained/html/collective.html#bundles-containerized-resources
> 
> There is also a walk-through:
> 
> https://projects.clusterlabs.org/w/projects/pacemaker/bundle_walk-through/
> 
> Unfortunately, since that was written, the HA packages were taken off
> the RHEL UBI container image, so you'll have to adjust the
> instructions for some other distro base image that does have the HA
> packages. It should be nearly identical though.
> 
> On Wed, 2024-07-17 at 11:56 +0000, MAIER Jakob via Users wrote:
> > Hi!
> > 
> > I would like to setup a number of podman containers in two nodes
> > such 
> > that every container is running on both nodes (active-active). One
> > of 
> > the containers should be master while the other is slave, and when 
> > there is a failure with the containers healthcheck, there should be
> > a 
> > failover. These two master-slave containers should also have a
> > virtual 
> > ip address assigned to them, that points to the current master.
> > There 
> > is sadly not a lot of documentation on bundles. I have found some 
> > documentation from redhat on such “complex bundles”, but
> > unfortunately 
> > that is also not very detailed, and I don’t really know how to 
> > proceed.
> > 
> > Does anybody know how I could go about setting something like this
> > up?
> > 
> > Kind regards,
> > Jakob
> > 
> > _______________________________________________
> > Manage your subscription:
> > https://lists.clusterlabs.org/mailman/listinfo/users
> > 
> > ClusterLabs home: https://www.clusterlabs.org/
> --
> Ken Gaillot <kgaillot at redhat.com>
> 
-- 
Ken Gaillot <kgaillot at redhat.com>



More information about the Users mailing list