innobackupex killed while reading lost+found
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Percona BenchWork |
New
|
Undecided
|
Unassigned |
Bug Description
I am using the Galera Cluster with the xtrabackup synchronization method.
MySQL runs as user mysql, and the script innobackupex is run as user mysql.
The datadir is a partition containing the directory lost+found, owned by root and having permissionis 700 (therefore only root can access the directory).
The my.cnf has an option to ignore lost+found and .ssh (therefore 'show databases' does not show lost+found)
When SST synchronization starts, I see in the logs that it cannot access lost+found and the script is killed.
Is there a way to tell innobackupex to ignore such directory, and to put this option inside my.cnf?
The workaroud is, of course, to delete lost+found (that will be re-created again when fsck is run...).
OtherwiseI should have good will to fix the perl script myself....