Clustering Galera on Single Stack IPv6
@ Chris Suttles · Sunday, Jul 23, 2017 · 3 minute read · Update at May 24, 2020

While setting up Galera on IPv6 only as part of our move toward multiple controller nodes, I ran into a few challenges. This post explains those challenges and how I dealt with them.

Resources I used and config reference

I found some good references on the web. The most useful ones including the following:

I also looked at:

Getting Started

I got most of what I needed from the blog post I mentioned previously. It's a great reference and starting point. I used the MariaDB downloads page to get the information needed to configure APT (we run Ubuntu 16.04 as our base OS). Once that was done, I installed MariaDB 10.1 with Galera and set up the config as mentioned in the blog post, and started trying to bootstrap the cluster. I got the nodes to join, but replication wasn't working, so I had to dig a little deeper.

For reference sake, here's a diff of my config before and after:

diff -u 50-server.cnf /etc/mysql/mariadb.conf.d/50-server.cnf
--- 50-server.cnf       2017-05-06 15:55:26.813413438 -0700
+++ /etc/mysql/mariadb.conf.d/50-server.cnf     2017-05-06 17:53:30.208914326 -0700
@@ -72,7 +72,7 @@
 # note: if you are setting up a replication slave, see README.Debian about
 #       other settings you may need to change.
 #server-id             = 1
-#log_bin                       = /var/log/mysql/mysql-bin.log
+log_bin                        = /var/log/mysql/mysql-bin.log
 expire_logs_days       = 10
 max_binlog_size   = 100M
 #binlog_do_db          = include_database_name
@@ -113,6 +113,23 @@
 # Also available for other users if required.
 # See
+#mysql settings
+#galera settings
+wsrep_provider_options = "gmcast.listen_addr=tcp://[::]:4567;ist.recv_addr=[2001:DB8::5]:4568"
+wsrep_node_address = "[2001:DB8::5]:4567"
+wsrep_sst_receive_address = "[2001:DB8::5]:4444"

 # this is only for embedded server
IPv6 and SST

After turning up logging and digging a bit, I saw that SST was failing and rsync was throwing an error. I dug into the rsync SST script and found that the line assigning the RSYNC_PORT variable assumed IPv4 syntax. I could have fixed the awk statement by changing $2 to $NF, but I wanted to do a simple fix and by the time I patched this I really just wanted to see replication work, so I was a bit heavy handed. I ended up with a patch like this for /usr/bin/wsrep_sst_rsync:

diff -u wsrep_sst_rsync /usr/bin/wsrep_sst_rsync
--- wsrep_sst_rsync     2017-05-23 15:37:30.596302338 -0700
+++ /usr/bin/wsrep_sst_rsync    2017-05-23 15:37:41.508184365 -0700
@@ -270,7 +270,8 @@
     rm -rf "$RSYNC_PID"

-    RSYNC_PORT=$(echo $ADDR | awk -F ':' '{ print $2 }')
+    #RSYNC_PORT=$(echo $ADDR | awk -F ':' '{ print $2 }')
+    RSYNC_PORT=4444
     if [ -z "$RSYNC_PORT" ]

Once I patched all the nodes rsync_sst script, had all the configs set up, and did the cluster bootstrap, I was finally able to check replication and state via the SHOW STATUS LIKE 'wsrep_%' statement.

Social Links