percona cluster as slave of mysql

Hi folks.

Perhaps some one can help me with my strange problem.

I have master mysql server (M1) with several slaves (S). And i need to move this infrastructure to percona cluster. But it’s impossible to update M1 and S from mysql to percona because of old OS (on M and S) and few other reasons.

So i’ve installed percona node (N1) and made id as slave of M1. After delay between M1 and N1 established to 0, i added one more percona node N2. In few moments after adding i’ve got:
“Last_SQL_Errno: 1062”
“Last_SQL_Error: Error ‘Duplicate entry ‘6812047’ for key ‘PRIMARY’’ on query. Default database: ‘%DB_NAME%’…”

So i do address this error to percona cluster internal data exchange. Same error appears if master-master (or multi -master) mysql replication exists and wrong increment and offset are set up.

Once i detach N2 from cluster and do on N1 “slave stop; SET GLOBAL SQL_SLAVE_SKIP_COUNTER=1; START SLAVE;” as mush as needed, slave process on N1 does not breaks anymore.

Is there any chance to make it working at all? what do i need to tune in my.cnf? I’ts not affordable to stop M1 at all, it’s heavily loaded.

In the end i want to keep N1, N2 cluster only and shutdown M1 and S.

Thank you in advance,
Alex M.

I forgot to mention that no firewall on servers or between of them, no DNS resolve delays - IP addresses used only.
M1, S, N1, N2 - are ESX VMs, all connected to same VDS w/o any restrictions. Also no disk space limitations on each instance.

N1, N2 are 100% equal VMs with OS - CentOS 6.4, kernel 2.6.32-358.6.1.el6.x86_64, latest one.

N1 and N2 - has latest percona packages available from CentOS repository:

$ rpm -qa | grep -i perco







Hi Alex,
Are you perhaps hitting this bug?