Bug 2227234

Summary:	Improve error message when adding a new node to a cluster with sbd is not possible
Product:	Red Hat Enterprise Linux 8	Reporter:	Tomas Jelinek <tojeline>
Component:	pcs	Assignee:	Tomas Jelinek <tojeline>
Status:	POST ---	QA Contact:	cluster-qe <cluster-qe>
Severity:	low	Docs Contact:
Priority:	medium
Version:	8.0	CC:	cluster-maint, idevat, mlisik, mmazoure, mpospisi, omular, tojeline
Target Milestone:	rc	Keywords:	Triaged
Target Release:	8.10
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	Enhancement
Doc Text:	Feature: Provide a guidance in error messages when adding or removing a node in a cluster with odd number of nodes, SBD enabled without disks, and auto_tie_breaker disabled. Reason: Originally, pcs in this situation just informed, that it was going to enable auto_tie_breaker, and then exited with an error saying corosync was running. This was not explanatory and it didn't provide enough information for users to solve this issue. Result: Error messages have been updated. They now explain that auto_tie_breaker must be enabled due to SBD and provide instructions to stop the cluster, enable auto_tie_breaker, start the cluster and run command for adding or removing a node again.	Story Points:	---
Clone Of:	2175797	Environment:
Last Closed:		Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:	2175797
Bug Blocks:

Description Tomas Jelinek 2023-07-28 13:01:34 UTC

+++ This bug was initially created as a clone of Bug #2175797 +++

Description of problem:
pcs can get to a situation, when adding a new node to a cluster with sbd is impossible, because there are 2 conditions - one prompting the user that cluster has to be offline due to auto_tie_breaker and the other one that it is not possible to get CIB (because the cluster is offline). Pcs should be aware of this situation and provide a more intuitive output. 


Version-Release number of selected component (if applicable):
found in pcs-0.11.4-6.el9


How reproducible:
every time the number of nodes would be even after adding a node (thus the cluster would have auto_tie_breaker) and sbd without disks is used.


Steps to Reproduce:

## enable sbd on 3 node cluster

[root@virt-553 ~]# pcs stonith sbd enable 
Running SBD pre-enabling checks...
virt-484: SBD pre-enabling checks done
virt-493: SBD pre-enabling checks done
virt-553: SBD pre-enabling checks done
Distributing SBD config...
virt-553: SBD config saved
virt-493: SBD config saved
virt-484: SBD config saved
Enabling sbd...
virt-493: sbd enabled
virt-484: sbd enabled
virt-553: sbd enabled
Warning: Cluster restart is required in order to apply these changes.

[root@virt-553 ~]# pcs cluster stop --all && pcs cluster start --all
virt-553: Stopping Cluster (pacemaker)...
virt-484: Stopping Cluster (pacemaker)...
virt-493: Stopping Cluster (pacemaker)...
virt-493: Stopping Cluster (corosync)...
virt-553: Stopping Cluster (corosync)...
virt-484: Stopping Cluster (corosync)...
virt-553: Starting Cluster...
virt-493: Starting Cluster...
virt-484: Starting Cluster...


## Try to add new node to the cluster

1. in a running cluster

[root@virt-553 ~]# pcs cluster node add virt-551
No addresses specified for host 'virt-551', using 'virt-551'
No watchdog has been specified for node 'virt-551'. Using default watchdog '/dev/watchdog'
Warning: auto_tie_breaker quorum option will be enabled to make SBD fencing effective. Cluster has to be offline to be able to make this change.
Checking corosync is not running on nodes...
Error: virt-493: corosync is running
Error: virt-484: corosync is running
Error: virt-553: corosync is running
Running SBD pre-enabling checks...
virt-551: SBD pre-enabling checks done
Error: Errors have occurred, therefore pcs is unable to continue
[root@virt-553 ~]# echo $?
1

> Node can't be added in a running cluster due to auto_tie_breaker sbd check. 


2. in a stopped cluster

[root@virt-553 ~]# pcs cluster stop --all
virt-553: Stopping Cluster (pacemaker)...
virt-493: Stopping Cluster (pacemaker)...
virt-484: Stopping Cluster (pacemaker)...
virt-484: Stopping Cluster (corosync)...
virt-553: Stopping Cluster (corosync)...
virt-493: Stopping Cluster (corosync)...

[root@virt-553 ~]# pcs cluster node add virt-551
No addresses specified for host 'virt-551', using 'virt-551'
No watchdog has been specified for node 'virt-551'. Using default watchdog '/dev/watchdog'
Error: Unable to load CIB to get guest and remote nodes from it, those nodes cannot be considered in configuration validation, use --force to override
Warning: auto_tie_breaker quorum option will be enabled to make SBD fencing effective. Cluster has to be offline to be able to make this change.
Checking corosync is not running on nodes...
virt-484: corosync is not running
virt-553: corosync is not running
virt-493: corosync is not running
Running SBD pre-enabling checks...
virt-551: SBD pre-enabling checks done
Error: Errors have occurred, therefore pcs is unable to continue
[root@virt-553 ~]# echo $?
1

> Node can't be added in a stopped cluster because CIB is unavailable.


Actual results:
In this state, it's not possible to add a node (without using --force), because the 2 checks are mutually exclusive - cluster needs to be stopped for auto_tie_breaker and cluster needs to be started to get CIB. 


Expected results:
More intuitive error message, which explains the situation why the node in this state can never be added and what to do to solve it (for example disable sbd first). Alternative solutions can be discussed as well.

--- Additional comment from Tomas Jelinek on 2023-03-06 16:20:14 CET ---

possible solutions:
* enable auto_tie_breaker
* disable sbd temporarily
* use --force in pcs cluster node add

Comment 1 Tomas Jelinek 2023-08-10 10:45:19 UTC

Upstream patch: https://github.com/ClusterLabs/pcs/commit/8a19eb351a98019993ae4a17ccf735a51cc67dc9

See bz2175797 comment 2 for tests.