[ClusterLabs] Adding a node to an active cluster

[Ken Gaillot]

On Wed, 2020-10-21 at 23:01 +0300, Andrei Borzenkov wrote:
> 21.10.2020 18:20, Jiaqi Tian1 пишет:
> > Hi,
> > Thanks for your reply. I'm using crmsh, are you refering to crm
> > cluster add node 
> > command? It gives error on my cluster, when I'm trying to add
> > ceha01 to a 
> > cluster by running crm cluster add ceha01.
> > "ERROR: cluster.join: Abort: Cluster is currently active. Run this
> > command on a 
> > node joining the cluster.
> > Connection to ceha01 closed."
> > Can you explain how this command works in this case, otherwise do
> > you have any 
> > other suggestions?
> 
> this command expects that node is in clean state and no cluster
> related
> services or configuration are present; it checks whether corosync
> service is active. To use this command you need to undo changes you
> did
> manually.
> 
> > BTW,
> > 1. Is there an issue updating cib.xml with cibadmin command when
> > Pacemaker is 
> > running?
> 
> Like - your changes may be overwritten by pacemaker?

cibadmin by default goes through the CIB manager rather than directly
modifying the file, so there's no problem there. (You could set
CIB_file in the environment to change that, but presumably that's not
happening here.)

> > 2. Do you have idea where(which config file) crm_node command
> > retrieves its data?
> 
> CIB
> 
> > Thanks,
> > Jiaqi Tian
> > 
> >     ----- Original message -----
> >     From: Andrei Borzenkov <arvidjaar at gmail.com>
> >     Sent by: "Users" <users-bounces at clusterlabs.org>
> >     To: Cluster Labs - All topics related to open-source clustering
> > welcomed
> >     <users at clusterlabs.org>
> >     Cc:
> >     Subject: [EXTERNAL] Re: [ClusterLabs] Adding a node to an
> > active cluster
> >     Date: Wed, Oct 21, 2020 10:46 AM
> >     On Wed, Oct 21, 2020 at 5:03 PM Jiaqi Tian1 <
> > Dylan.Jiaqi.Tian at ibm.com> wrote:
> >      >
> >      > Hi,
> >      > I'm trying to add a new node into an active pacemaker
> > cluster with
> >     resources up and running.
> >      > After steps:
> >      > 1. update corosync.conf files among all hosts in cluster
> > including the
> >     new node
> >      > 2. copy corosync auth file to the new node
> >      > 3. enable corosync and pacemaker on the new node
> >      > 4. adding the new node to the list of node in
> > /var/lib/pacemaker/cib/cib.xml
> >      >
> > 
> >     Do you mean you edited CIB while pacemaker was running?
> > 
> >      > Then I run crm status, the new node is displayed as offline.
> > It will not
> >     become online, unless we run restart corosync and pacemaker on
> > all nodes in
> >     cluster. But this is not what we want, since we want to keep
> > existing nodes
> >     and resources up and running. Also in this case crm_node -l
> > doesn't list the
> >     new node.
> >      >
> >      > So my question is:
> >      >
> >      > 1. Is there another approach to have the existing nodes
> > aware of the new
> >     node and have crm status indicates the node is online while
> > keeping other
> >     nodes and resources up and running?
> > 
> >     I think you are better off using high level tools like pcs or
> > crmsh.
> >     They automate all steps needed to make cluster recognize new
> > node
> >     online.
> > 
> >      > 2. which config file crm_node command reads?
> >      >
> > 
> >     CIB
> >     _______________________________________________
> >     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/
-- 
Ken Gaillot <kgaillot at redhat.com>