Incremental option for wsrep_sst_method

MySQL & MariaDB Percona XtraDB Cluster 5.x
1

Hi
Currently i am using this configuration for setting up a cluster on node1

node 1
binlog_format=ROW
wsrep_provider=/usr/lib64/libgalera_smm.so
wsrep_cluster_address=gcomm://
wsrep_slave_threads=2
wsrep_cluster_name=dev_cluster
wsrep_sst_method=rsync
wsrep_node_name=node1
innodb_locks_unsafe_for_binlog=1
innodb_autoinc_lock_mode=2
wsrep_replicate_myisam=1

node2
wsrep_provider=/usr/lib64/libgalera_smm.so
wsrep_cluster_address=gcomm://10.1.0.1:4567
wsrep_slave_threads=2
wsrep_sst_method=rsync
wsrep_cluster_name=dev_cluster
binlog_format=ROW
default_storage_engine=InnoDB
innodb_autoinc_lock_mode=2
innodb_locks_unsafe_for_binlog=1
wsrep_replicate_myisam=1

and what i understand is in case node 2 goes down and comes backup, it will perform a full rsync from node1 /var/lib/mysql directory in order to get unto date with the data and depemding on the size of data it can take much longer for big db’s
so here are my questions

  1. how do i configure this to only patch all the incremental updates from node1?
  2. how does node 1 know where to store all the increments for node2 to patchup once it is up?
  3. How does this scenario work when node2 is up and node1 goes down and comes back up?

Thanks once again…
i am using
Server version: 5.5.24-55-log Percona XtraDB Cluster (GPL), wsrep_23.6.r341
on ubuntu 10.04

2

Hi,

2 nodes is not the recommended amount of members for an optimal cluster.

Also, if you don’t what that SST blocks the donor, use xtrabackup

All your other questions are detailed in http://www.mysqlperformanceblog.com/2012/07/25/percona-xtrad b-cluster-failure-scenarios-with-only-2-nodes/

regards,

3

lefred wrote on Tue, 14 August 2012 00:56

Hello

I found that using xtrabackup results the same situation.
the cluster sync rate also in 3~4KB/s

or is there a better way to switch MySQL multi-master scenario to Percona cluster?

Thanks a lot