Avoid split-brain on Two-Node Clusters: possibly misleading article in knowledgebase

[Patricio Keilty]

Hi folks,

We recently started using Galera Cluster 3 as means to make our MySQL data layer more robust. We initially set up a 2 node cluster, but soon found out we needed to avoid split-brain situations....so we conduct our research and test-drive different approaches, reading through the documentation and knowledge base was very helpful indeed.

We tried the following:

a- 2 nodes using pc.weight=2 parameter on one of the nodes so that quorum is achieved when such node is up and running 

b- 2 nodes running garbd (arbitrator) on both nodes as sugested by article in KB (https://galeracluster.com/library/kb/two-node-clusters.html)

Option a is sub-optimal solution because you may end up with a second node up and no quorum. Option b didn't work at all since you can never achieve quorum having total weight of 2+2.

To the best of our knowledge, the proper approach to avoid split-brain seems to be using a third node with arbitrator. Would appreciate your kind advice on this subject.

Also, wanted to point out that the good quality material in the documentation, tutorials and KB helped us a lot in the process. "Two-Node Clusters" article seems somewhat misleading though.

Greetings,

--patricio

[vidhut singh]

Yes this is a misleading article since I tried and couldn't bring up the galera-arbitrator as it gives following error 

failed to open gcomm backend connection: 98: error while trying to listen 'tcp://0.0.0.0:4567?socket.non_blocking=1', asio error 'bind: Address already in use': 98 (Address already in use)

[Nilesh Ramesh Gavali]

Hi Patricio, you need to run (garbd) Arbitrator process on a  (single) separate (third) node.

On Thursday, 17 December 2020 at 19:41:19 UTC+5:30 patrici...@gmail.com wrote: