[ClusterLabs] [EXT] Re: Creating pacemaker cluster starting with one node

[Andreas Mock]

Hi Ulrich,

thank you for your recommendations. Currently I'm checking my possibilities. Fact is, I don't have spare machines.
I'll see which road to take.

Best regards
Andreas

-
-----Ursprüngliche Nachricht-----
Von: Users <users-bounces at clusterlabs.org> Im Auftrag von Windl, Ulrich
Gesendet: Montag, 4. August 2025 09:49
An: Cluster Labs - All topics related to open-source clustering welcomed <users at clusterlabs.org>
Betreff: Re: [ClusterLabs] [EXT] Re: Creating pacemaker cluster starting with one node

Hi!

While it's possible to start with one node and add more nodes later, I would recommend to configure all nodes that should run in the cluster at the same time. Lao it the nodes are physical I would recommend using mostly identical machines. We had some problems when the CPU model was different and VMs should be live-migrated from the newer CPU to the older one.
The point why I'm recommending to start with all nodes it testing: You should test fencing and resource migration before going productive. When adding nodes that would mean a downtime, specifically if you find configuration mistakes (not all nodes being equally configured).

Kind regards,
Ulrich Windl

> -----Original Message-----
> From: Users <users-bounces at clusterlabs.org> On Behalf Of Eugen Block
> Sent: Saturday, August 2, 2025 2:00 PM
> To: Cluster Labs - All topics related to open-source clustering 
> welcomed <users at clusterlabs.org>
> Subject: [EXT] Re: [ClusterLabs] Creating pacemaker cluster starting 
> with one node
> 
> Since we are familiar with SUSE products, we based our deployment on 
> the SUSE High Availability docs:
> 
> https://documentation.suse.com/en-us/sle-ha/15-SP7/single-html/SLE-HA-
> installation/index.html
> 
> So in short, we run „crm cluster init“ on the bootstrap node, and when 
> everything works as expected, „crm node join“ on the further nodes. Of 
> course, there’s some more details required, like the corosync config,
> csync2 etc.
> 
> Our own cluster was based on openSUSE, but when SUSE discontinued the 
> OpenStack product, we couldn’t upgrade anymore on openSUSE. So last 
> year, we reinstalled our cloud on Ubuntu by redeploying a pacemaker 
> cluster (single node), brought OpenStack back up, then joined the 
> second server. And it works great, we’re just waiting for the latest 
> updates to arrive in the distro for better systemd management.
> 
> Zitat von Andreas Mock <andreas.mock at drumedar.de>:
> 
> > Hi Thomas, hi Eugen,
> >
> > thank you for the reply.
> >
> > Is there anywhere a howto to bootstrap such an one node cluster?
> >
> > Best regards
> > Andreas
> >
> >
> > -
> > -----Ursprüngliche Nachricht-----
> > Von: Users <users-bounces at clusterlabs.org> Im Auftrag von Eugen 
> > Block
> > Gesendet: Samstag, 2. August 2025 11:18
> > An: Cluster Labs - All topics related to open-source clustering 
> > welcomed <users at clusterlabs.org>
> > Betreff: Re: [ClusterLabs] Creating pacemaker cluster starting with 
> > one
> node
> >
> > Hi,
> >
> > that's how we usually set up our Openstack cluster(s). We bootstrap 
> > our first node so we have a single-node pacemaker cluster running 
> > all Openstack control services, then "join" the other control nodes, 
> > expanding the cluster.
> > I'd recommend to practice in a test cluster, that has helped me a lot!
> >
> > Regards,
> > Eugen
> >
> > Zitat von Tomas Jelinek <tojeline at redhat.com>:
> >
> >> Hi Andreas,
> >>
> >> From the point of view of pcs, pacemaker and corosync, it is 
> >> possible to create a one-node cluster, run resources in it, and  
> >> add another node afterwards. Obviously, when the single node is 
> >> lost for whatever reason, there is no other node to run the resources.
> >>
> >> You should also consider whether the cluster's workload supports 
> >> this and what steps are necessary to make it work in this scenario.
> >>
> >> Regards,
> >> Tomas
> >>
> >>
> >> Dne 31. 07. 25 v 8:46 Andreas Mock napsal(a):
> >>> Hi all,
> >>>
> >>> I have to upgrade an old pacemaker cluster consisting of two nodes.
> >>> (There is no chance of rolling update).
> >>>
> >>> Is it possible to create a pacemaker “cluster” starting with only 
> >>> one node and afterwards attach the second node?
> >>>
> >>> Best regards
> >>>
> >>> Andreas
> >>>
> >>>
> >>> _______________________________________________
> >>> Manage your subscription:
> >>> https://lists.clusterlabs.org/mailman/listinfo/users
> >>>
> >>> ClusterLabs home: https://www.clusterlabs.org/
> >>
> >> _______________________________________________
> >> Manage your subscription:
> >> https://lists.clusterlabs.org/mailman/listinfo/users
> >>
> >> ClusterLabs home: https://www.clusterlabs.org/
> >
> >
> >
> > _______________________________________________
> > Manage your subscription:
> > https://lists.clusterlabs.org/mailman/listinfo/users
> >
> > ClusterLabs home: https://www.clusterlabs.org/ 
> > _______________________________________________
> > Manage your subscription:
> > https://lists.clusterlabs.org/mailman/listinfo/users
> >
> > ClusterLabs home: https://www.clusterlabs.org/
> 
> 
> 
> _______________________________________________
> Manage your subscription:
> https://lists.clusterlabs.org/mailman/listinfo/users
> 
> ClusterLabs home: https://www.clusterlabs.org/
_______________________________________________
Manage your subscription:
https://lists.clusterlabs.org/mailman/listinfo/users

ClusterLabs home: https://www.clusterlabs.org/