[ClusterLabs Developers] Pacemaker & corosync - how does totem token works and formula

[Jan Friesse]

Hi,

On 08/05/2025 19:14, Vicki Chen wrote:
> Hi,
> 
> I'm a software developer currently researching how Corosync utilizes the Totem protocol token. Corosync employs the Totem protocol to monitor the health of cluster nodes by circulating a token among them. If a node doesn't receive the token within a specified timeout period (configured in corosync.conf), it's considered unresponsive. I'm interested in diving deeper into this mechanism. Is there a specific formula or method used to determine host failure detection time in Pacemaker? Could you provide details? Thanks!

Are you interested in Pacemaker or Corosync? For Corosync head to 
https://github.com/corosync/corosync/wiki/Developers#reference-documentation 
specially "Totem: A Reliable Ordered Delivery Protocol for 
Interconnected Local-Area Networks (PDF) - Deborah A. Agarwal - 1994".

Todays Corosync differs in many areas, especially using knet instead of 
IP multicast, but totem itself is more or less 1:1 implementation of 
protocol described in Deborah A. Agarwal thesis.

Regards
   Honza

> 
> best regards,
> 
> Vicki Chen
> 
> 
> 
> 
> 
> _______________________________________________
> Manage your subscription:
> https://lists.clusterlabs.org/mailman/listinfo/developers
> 
> ClusterLabs home: https://www.clusterlabs.org/
>