[ClusterLabs] pcs node removal still crm_node it is removed node is listing as lost node

[Ken Gaillot]

On Thu, 2023-07-13 at 11:03 -0500, Ken Gaillot wrote:
> On Thu, 2023-07-13 at 09:58 +0000, S Sathish S via Users wrote:
> > Hi Team,
> >  
> > Problem Statement : we are trying to remove node on pcs cluster,
> > post
> > execution also still crm_node 
> > it is removed node is listing as lost node.
> >  
> > we have checked corosync.conf file it is removed but still it is
> > displaying on
> > crm_node -l.
> >  
> > [root at node1 ~]# pcs cluster node remove node2 --force
> > Destroying cluster on hosts: 'node2'...
> > node2: Successfully destroyed cluster
> > Sending updated corosync.conf to nodes...
> > node1: Succeeded
> > node1: Corosync configuration reloaded
> >  
> > [root at node1 ~]# crm_node -l
> > 1 node1 member
> > 2 node2 lost
> 
> This looks like a possible regression. The "node remove" command
> should
> erase all knowledge of the node, but I can reproduce this, and I
> don't
> see log messages I would expect. I'll have to investigate further.

This turned out to be a regression introduced in Pacemaker 2.0.5. It is
now fixed in the main branch by commit 3e31da00, expected to land in
2.1.7 toward the end of this year.

It only affected "crm_node -l".

> 
> >  
> > In RHEL 7.x we are using below rpm version not seeing this issue
> > while removing the node.
> > pacemaker-2.0.2-2.el7
> > corosync-2.4.4-2.el7
> > pcs-0.9.170-1.el7
> >  
> > In RHEL 8.x we are using below rpm version but seeing above issue
> > over here.
> > pacemaker-2.1.6-1.el8
> > corosync-3.1.7-1.el8
> > pcs-0.10.16-1.el8
> >  
> > Thanks and Regards,
> > S Sathish S
-- 
Ken Gaillot <kgaillot at redhat.com>