OpenVZ VPS. Fully updated Centos v7 with latest yum updates as of today

For the repo question, these are not MariaDB's, ours are named differently: http://yum.mariadb.org/10.1.16/centos7-amd64/rpms/
We don't have control over the 3rd party repo, whichever it was.
Several previous versions of MariaDB packages can be found at http://yum.mariadb.org/ .

The actual Galera question goes to nirbhay_c, to see whether it's something wrong with MGC. It is also a possibility that the packages themselves are broken.

I can confirm that this bug has existed since the 10.1.15 package release. As to what causes it, I traced it to some sort of file parsing issue in the wsrep_recover_position function in /usr/bin/galera_recovery. Specifically somewhere within this block:

  recovered_pos="$(grep 'WSREP: Recovered position:' $log_file)"

  if [ -z "$recovered_pos" ]; then

    skipped="$(grep WSREP $log_file | grep 'skipping position recovery')"

    if [ -z "$skipped" ]; then

      log "WSREP: Failed to recover position: '`cat $log_file`'"

      exit 1

    else

      log "WSREP: Position recovery skipped."

    fi

  else

    start_pos="$(echo $recovered_pos | sed 's/.*WSREP\:\ Recovered\ position://' \

                    | sed 's/^[ \t]*//')"

    log "WSREP: Recovered position $start_pos"

    start_pos_opt="--wsrep_start_position=$start_pos"

  fi

}

Unfortunately I have not had further time to work on a fix for this; however, hoping Assignee et al can pick this up soon, as I am stuck on 10.1.14 for my clusters now as a result. Thanks.

I ran into the same issue and traced it down to the same wsrep_recover_position() code in galera_recover script.
# Redirect server's error log to the log file.
eval /usr/sbin/mysqld $cmdline_args --user=$user --wsrep_recover 2> "$log_file"

What I think is happening is that this script assumes that when it starts mysqld to perform the position recovery, the output it is looking for "WSREP: Recovered position: " will be directed to stdout, but in my case, the output needed is directed to where I asked mysql error logging to go:
In my /etc/my.cnf.d/server.cnf, I have [mysqld] log-error=/var/log/mariadb/mysqld.log

# tail -1 /var/log/mariadb/mysqld.log
2016-07-22 15:58:04 140172581619840 [Note] WSREP: Recovered position: 00000000-0000-0000-0000-000000000000:-1

Not where galera_recover was expecting it.

I did get galera_new_cluster to work by changing the line at the beginning of the wsrep_recover_position() to:

# Redirect server's error log to the log file.
eval /usr/sbin/mysqld $cmdline_args --user=$user --wsrep_recover --log-error=$log_file 2> "$log_file"

To clarify, I am using the official MariaDB10 repository. I copied the name wrong.

rvlane You got it right.

MDEV-10396