Incomprehensible behavior RAFT

Community Support
1

Hello.
I use Opennebula 5.12.0.4 and noticed incomprehensible behavior RAFT.
I have 3 servers in a cluster.

HA & FEDERATION SYNC STATUS                                                    
ID NAME            STATE      TERM       INDEX      COMMIT     VOTE  FED_INDEX 
 0 oned-preprod-1. follower   41502      6241529    6241529    3     -1
 1 oned-preprod-2. follower   41502      6241529    6241529    -1    -1
 2 oned-preprod-3. leader     41502      6241529    6241529    3     -1

When i restarted opennebula service on a leader my 2 other servers can’t choose new leader and waiting when server oned-preprod-3 will online.

In the logs in this moment:

Failed to get heartbeat from leader. Starting election proccess
Vote not granted from follower 0: [one.zone.voterequest] Candidate's term is outdated
Follower 0 is in term 41501 current term is 41500. Turning into follower
Could not find information driver 'monitord'
oned is set to follower mode
Failed to get heartbeat from leader. Starting election proccess
Got vote from follower 0. Total votes: 1
Got majority of votes
Becoming leader of the zone. Last log record: 6241525 last applied record: 6241525
Could not find information driver 'monitord'
oned is now the leader of the zone

And other case.
When i restarted opennebula service on 2 or 3 node at the same time i get 2 leaders in my cluster.

HA & FEDERATION SYNC STATUS                                                     
ID NAME            STATE      TERM       INDEX      COMMIT     VOTE  FED_INDEX 
 0 oned-preprod-1. follower   41517      6241563    6241563    0     -1
 1 oned-preprod-2. leader     41517      6241563    6241563    2     -1
 2 oned-preprod-3. leader     41517      6241563    6241563    3     -1

Why it happens and how decide the problems?

2

Considering this message

Could not find information driver ‘monitord’

Could it be an out of sync configuration files? This would prevent the proper operation of the cluster and if some ID’s are not properly set it could lead to this multiple leader election.

3

Hi.
Thank you for your reply.
I’ve sync .one directory for all my servers in the cluster. And of course backup and restore the database. ID’s on all server are different.