[ClusterLabs] last_man_standing (LMS) and wait_for_all (WFA)

[Jan Friesse]

Hi,
WFA is essential to prevent existence of 2 quorate partitions.

Let's use example in votequorum(5). Stop at point 5 and imagine rest 5 
nodes boots up without WFA and creates new partition - there is 5 nodes 
out of 8 so such partition is quorate -> two quorate partitions exists.

On 02/05/2025 19:47, Faaland, Olaf P. via Users wrote:
> I should say we deal with the common startup race described in votequorum(5), whereby nodes that don’t come up are fenced once quorum is attained, by other means.

I'm really not sure what you mean or where is the problem.

Regards
   Honza

> 
> -Olaf
> 
> From: Faaland, Olaf P. <faaland1 at llnl.gov>
> Date: Friday, May 2, 2025 at 10:25 AM
> To: users at clusterlabs.org <users at clusterlabs.org>
> Subject: last_man_standing (LMS) and wait_for_all (WFA)
> Hi,
> 
> In votequorum(5), discussing last_man_standing (LMS) and wait_for_all (WFA) with corosync-3.1.8-1.el8.x86_64, it says:
> 
> 
> Enabling LMS allows the cluster to dynamically recalculate expected_votes and quorum under specific circumstances. It is essential to enable WFA when using LMS in High Availability clusters.
> 
> Why is that essential?
> 
> I have many, many clusters, and there will be times when a node is down because we need replacement parts or something similar, and the cluster must be rebooted (e.g. an OS update).
> 
> Thanks!
> Olaf
> 
> 
> _______________________________________________
> Manage your subscription:
> https://lists.clusterlabs.org/mailman/listinfo/users
> 
> ClusterLabs home: https://www.clusterlabs.org/
>