High Availability of NARAYANA coordinator

[Shimaa Kamal]

Hello,

I'm new to Narayana and this is the first time for me to deal with TM. 

I read many articles about Narayana TM, but I have a question about Narayana coordinator availability. (sorry if my question is trivial)

Can I have multiple instances of the coordinator in case if one of them crash, the others will continue coordinate the services? and how? (To avoid single point of failure)
Something like active-active or active-passive availability? 

Or the recovery manager of the coordinator is enough and will do that for me?

Thank you 

[Michael Musgrove]

Coordinators have a unique node id and we use that to guarantee that different instances of a coordinator do not try to manage the same LRA (since that would compromise the LRA state model). So, at the moment, if one of the coordinators fail you/your infrastructure must restart it (using the same node id and storage).

We do have plans for running multiple coordinators [1] but that work hasn't start yet.

Proper management of the coordinator is key for safe operations so I am planning on updating the quarkus guide explaining the relationship between coordinators and the quarkus extension and providing best practices.

[1] https://github.com/jbosstm/documentation/blob/main/discussions/lra/lra-plans-and-ideas.adoc

[Michael Musgrove]

I am marking this question as answered. But do track our updates via our blogs and documentation etc but I will mention any changes in this area in any future release blog.