Description:
When trying to create a UNDOFILE on our cluster environment with the following command:
CREATE LOGFILE GROUP lg_1
ADD UNDOFILE 'undo_1.dat'
INITIAL_SIZE 16M
UNDO_BUFFER_SIZE 2M
ENGINE NDB;
We recieve the following Error:
ERROR 1528 (HY000): Failed to create UNDOFILE
show warnings:
*************************** 1. row ***************************
Level: Error
Code: 1296
Message: Got error 1509 'File system error, check if path,permissions etc' from NDB
*************************** 2. row ***************************
Level: Error
Code: 1528
Message: Failed to create UNDOFILE
2 rows in set (0.00 sec)
When executing an ls -la on the ndb_#_fs we see the files present:
[root@hostname /mnt/data/ndb_10_fs]# ls -la
total 16420
drwxr-x---. 9 root root 4096 Dec 21 12:45 .
drwxr-xr-x. 5 root root 4096 Dec 21 12:04 ..
drwxr-x---. 4 root root 4096 Dec 21 12:05 D1
drwxr-x---. 3 root root 4096 Dec 21 12:04 D10
drwxr-x---. 3 root root 4096 Dec 21 12:04 D11
drwxr-x---. 4 root root 4096 Dec 21 12:05 D2
drwxr-x---. 3 root root 4096 Dec 21 12:04 D8
drwxr-x---. 3 root root 4096 Dec 21 12:04 D9
drwxr-x---. 4 root root 4096 Dec 21 12:24 LCP
-rw-r--r--. 1 root root 16777216 Dec 21 12:45 undo_1.dat
The behavior doesn't change when giving chmod 777 access rights to the ndb_#_fs directory.
Here the config.ini for our Cluster
[NDBD DEFAULT]
DataDir=/mnt/log/ # Directory for this data node's log files
FileSystemPath=/mnt/data # Directory for this data node's data files
BackupDataDir=/mnt/backup/ # Directory for this data node's backup files
[NDB_MGMD DEFAULT]
datadir=/var/lib/mysql-cluster # Directory for MGM node log files
[ndb_mgmd]
NodeId=1
hostname=<IP Adress>
[ndb_mgmd]
NodeId=2
hostname=<IP Adress>
[ndb_mgmd]
NodeId=3
hostname=<IP Adress>
[ndb_mgmd]
NodeId=4
hostname=<IP Adress>
[ndb_mgmd]
NodeId=5
hostname=<IP Adress>
[mysqld]
NodeId=6
hostname=<IP Adress>
[mysqld]
NodeId=7
hostname=<IP Adress>
[mysqld]
NodeId=8
hostname=<IP Adress>
[mysqld]
NodeId=9
hostname=<IP Adress>
[ndbd]
NodeId=10
hostname=<IP Adress>
[ndbd]
NodeId=11
hostname=<IP Adress>
[ndbd]
NodeId=12
hostname=<IP Adress>
[ndbd]
NodeId=13
hostname=<IP Adress>
SeLinux is disabled on all Servers.
We've tried to reduce the number of mgm an ndbd nodes but it didn't change the ourcome.
We've also tried to use a different directory or no defined directory at all, both lead to no change in behavior.
The filesystem for all Nodes is ext4
How to repeat:
This is a fresh cluster setup with a bare minimum configuration.
Starting the cluster with the following commands:
ndb_mgmd -f /var/lib/mysql-cluster/config.ini --nowait-nodes 3,4,5 --config-cache --initial
ndbmtd --nostart
ndb_mgm> ALL START
/etc/init.d/mysql start
Then issue the following mysql commands on any mysql nodes in the cluster:
create database ndb;
use ndb;
CREATE LOGFILE GROUP lg_1
ADD UNDOFILE 'undo_1.dat'
INITIAL_SIZE 16M
UNDO_BUFFER_SIZE 2M
ENGINE NDB;
