Details
-
Bug
-
Status: Closed (View Workflow)
-
Trivial
-
Resolution: Not a Bug
-
10.1.12, 10.1.14
-
Ubuntu 14.04, vSphere VM. Three VMs, two are mysqld master nodes and one is garbd. Cluster is in one DC but being prepared to be used in three DCs and at that point will add two more master nodes and move garb to third DC.
All nodes have 2 vCPU cores, 6G RAM, DB on separate vdisk, DB is 51G in total.Ubuntu 14.04, vSphere VM. Three VMs, two are mysqld master nodes and one is garbd. Cluster is in one DC but being prepared to be used in three DCs and at that point will add two more master nodes and move garb to third DC. All nodes have 2 vCPU cores, 6G RAM, DB on separate vdisk, DB is 51G in total.
Description
I just patched from 10.1.12 to 10.1.14 and this issue remained. Discovered recently due to xtrabackup-v2 having problems with binlogs because they had defaulted to the datadir.
So binlogs were using the pattern datadir/log-basename-bin.?. But in configuration I had set /var/db/binlogs/log-basename-bin.?.
show variable like 'log_bin%'; would show the default datadir path described above.
I tried searching all of /etc/mysql for the cause and this was the result.
$ sudo grep -rE 'log-bin|log_bin' /etc/mysql/
|
/etc/mysql/my.cnf:#log_bin = /var/log/mysql/mariadb-bin
|
/etc/mysql/my.cnf:#log_bin_index = /var/log/mysql/mariadb-bin.index
|
/etc/mysql/conf.d/replication.cnf:log_bin = /var/db/binlogs/mydb-bin
|
/etc/mysql/conf.d/replication.cnf:log_bin_index = /var/db/binlogs/mydb-bin.index
|
The only workaround was to explicitly set --log-bin arguments when starting the sysvinit start-script. Which of course is not a good solution for future patching. When using this method the log_bin% variables are set correctly too.
Edit: I'm aware that log_bin is supposed to be a boolean value but passing it as a string on cli resolves the issue. Also if I try to use log_bin_basename in the config file I get an error on startup saying ''unknown variable 'log_bin_basename=/var/db/binlogs/mydb-bin''.