We tried a couple of days ago to create a slave using an existing slave as the place to copy from, we are using mariadb 10.0.14.
We tried a couple of ways to point that slave to the master and in the end had to use the original slave as the master for this slave, creating a daisy chain instead of two slave directly pointing to the one master.
That slave has log_bin enabled and the log_bin file numbers on it are at a higher number than the current master. This occurred due to this slave having been a master previously and us not disabling log_bin when restarting the slave after switching masters.
Using the notes at http://www.percona.com/doc/percona-xtrabackup/2.1/howtos/setting_up_replication.html to create a slave from an existing slave left us a little short and had to work a couple of things out.
The xtrabackup_slave_info file contained change master to slave_pos for GTID not being fully across GTID I wasn’t sure how the slave was supposed to sync first with a file and position number.
There are a few steps missing for mariadb in that note now pertaining to how mariadb gtid is implemented
My question is what are the steps to take this from a copied sslave to getting it to sync to the master not the other slave as its mater creating a daisy chain?
Should I enable GTID as we are not currently using it?
Was it just a case of entering “Change master to master_host Master_user, Master_password” without master_file and position
followed by Change master to slave_pos
and then Slave Start?
Would that have allowed that slave to join the actual master and rectify the Log_bin file numbers and position being different?
slave 1 mysql-bin.000021
Should I now enable gtid so allowing me to switch from slave 1 as the master for slave two or is this problem of position number being higher a real problem?
I do have the option of turning the bin_log off on slave 1 forcing a change in behaviour but perhaps other don’t and I may not at some stage in the future.
Any assistance is greatly appreciated