Donor/desynced cluster request timeouts.

Hello. I try switch from master-slave to master-master with minimal downtime.
I prepare 3 servers for m-m cluster. Start m-s replication with original server, wait when replication is complete, stop slave. All is ok, all data in base is valid. Now i switch my frontend web servers to new DB. All work fine.
Now i start add new masters in cluster. I restart my db with enabled wsrep=on and --wsrep-new-cluster , try my web= all fine. Go to another master console and start mysql and all my web go
down, any request go to timeout, from console also timeout for any request …

Primary maser

Second master

How i can work with my cluster in donor/desynced state ?

Try use wsrep_sst_method=xtrabackup and timeouts go away. Help me understand, i do right ?