[MXS-4202] Master Node rejoin as master after some transactions and there is data loss. Created: 2022-07-14  Updated: 2022-09-01  Resolved: 2022-09-01

Status: Closed
Project: MariaDB MaxScale
Component/s: mariadbmon
Affects Version/s: 6.4.0
Fix Version/s: 6.4.2

Type: Bug Priority: Critical
Reporter: Mohammed Y ALALI Assignee: Esa Korhonen
Resolution: Not a Bug Votes: 0
Labels: None


 Description   

I'm using MariaDB Server 10.6.8

(1) Start a Fresh new cluster Server 1 = master, Server2,3= Slave
(2) Bring Master Down (without having done any transactions)
(3) Server 2 gets promoted to Master
(4) Perform a couple of transactions
(5) Bring Server 2 and 3 down
(6) Bring up Server 1
(7) Bring up Server 2 and 3

  • Auto failover and auto rejoin are enabled
  • I'm using MaxScale

Server 1 became master and Server 2 & 3 are not synced.

How do I deal with this situation?

Is it possible to check the GTID before promoting to master? So I can make sure that the node that was the last master should become master again.

 Comments   
Comment by markus makela [ 2022-07-14 ]

Do you have semi-sync replication enabled?

Comment by Mohammed Y ALALI [ 2022-07-14 ]

@markus

Yes it's enabled

Comment by Esa Korhonen [ 2022-08-29 ]

This is a total cluster fail, and not handled. After step 6, MaxScale must assume that Servers 2 & 3 may never come back, so 1 must be allowed to become master.

Generated at Thu Feb 08 04:26:56 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.